By Manny Tayas, Chief Revenue Officer

Key Takeaways

  • Contract rates set early in 2026 bid season stopped holding by June.
  • 43% of brokers in a 141-broker survey said gross margin fell in H1 2026.
  • Shippers now reprice broken lanes with mini-bids, so keep carrier pricing current all year.
  • Freight procurement software saves the most time on collecting carrier rates.
  • AI for freight procurement is only as safe as the limits you set.

I miss the days when a carrier gave you a rate in January, and you could reasonably expect it to mean something in June. Now a shipper drops 400 lanes in your lap on Monday, wants your best numbers by Friday, and your pricing team spends the week chasing carriers who apparently consider answering emails a personal favor.

Even when you get the quotes, good luck keeping them fresh. SONAR’s tender rejection index hit 14.17% in October, and Knight-Swift’s CEO is talking about “low-to-mid double digits” in rate increases. Suddenly, winning that big contract can be a pretty expensive way to congratulate yourself.

I’ve always found it funny how much freight procurement software we’ve built to help shippers ask for rates, while the LSPs answering those requests are still buried in spreadsheets and email chains.

So what happens when we flip that around? When carrier sourcing, negotiations, and pricing happen continuously, and the next RFP shows up to find most of the work already done? In other words, “running itself.”

Your Price Stops Expiring in June

You know how everyone goes into January convinced this will finally be the year they stick to their New Year’s resolutions? Annual freight contracts have about the same track record. Everyone agrees on a number, shakes hands, and hopes the market cooperates.

June 2026 was a pretty good reminder of how that usually works out. Executives at several major carriers were already telling investors that contracted rates weren’t holding, forcing some shippers to rebid their freight.

Brokers got caught in the middle. According to a Truckstop and Bloomberg Intelligence survey, 43% reported shrinking gross margins, while 69% had to renegotiate with shippers earlier than usual. Nobody enjoys calling a customer to explain why yesterday’s price is suddenly fiction.

Keeping carrier buy rates current through continuous sourcing means your next bid starts with what freight actually costs today, rather than what everyone hoped it would cost six months ago.

Mini-Bids Get a Lot Less Painful

Of course, shippers aren’t sitting around until next January to fix rates that stopped working in June. They’re reopening lanes whenever carriers start saying no, which means your pricing team gets to enjoy bid season several times a year.

Uber Freight’s Q3 market update gives you an idea of how bad it’s gotten. Primary tender acceptance in its network fell to 78% in August, compared with 90% to 94% in the previous three years. Shippers have been using targeted mini-bids to repair those routing guides, often covering 10 to 50 lanes over two to four weeks.

I suspect we’ll see plenty more of that in 2027, and I’d rather not have my team start another carrier-email marathon every time a shipper reopens 30 lanes. When your carrier rates are already current, a mini-bid becomes a quick pricing review instead of another week spent rebuilding quotes from scratch.

Your Inbox Gets Demoted

All those mini-bids sound a lot more manageable when your carrier rates are already current. Trouble is, somebody has to keep them current, and if your pricing team is still chasing quotes through email, you’ve essentially traded one full-time headache for another.

I’ve seen teams request 30 rates and get back 20 different interpretations of how a spreadsheet should work. Then some poor soul gets to spend Friday afternoon copying numbers into the TMS. What a fantastic use of expensive talent.

There’s real risk buried in that inbox, too. Uber Freight blocked 784,201 fraudulent emails in Q2, with communication-based attacks accounting for half its classified fraud attempts.

Our Connect Once approach handles API and EDI connections directly as AI agents work the email-and-phone crowd and capture rates in your TMS. Your people handle exceptions, including independently verifying suspicious payment changes, instead of spending their careers as human copy-and-paste machines.

Your Buy Rates Stay in the Family

Getting carrier rates into your TMS automatically sounds great, until you ask a question I wish more buyers would: Who else gets to see them?

Brokers guard their buy rates like Grandma’s secret sauce. Except Grandma doesn’t have competitors looking to steal her customers. I’ve seen this question decide software deals faster than price, and I don’t blame anyone for asking.

At LSP44, we’ve put our answer in writing. We’ll never operate as an NVOCC or indirect air carrier, seek brokerage or motor carrier authority, train other LSPs’ models on your data, or resell it to shippers.

We won’t feed it into project44’s shipper products, either. We’re a separate business with our own P&L, and our FAQ spells out the commitments for anyone whose legal department enjoys a good afternoon of reading.

Ask every vendor, including us, to put those promises in writing. Watch how quickly the conversation gets interesting.

Freight Procurement Software Does the Grunt Work

Now that we’ve established your buy rates stay yours, let’s talk about what happens when the software actually shows its worth. A shipper drops a monster lane file in your lap, and instead of assigning three people to spend the week wrestling with it, your team gets a bid that’s already been cleaned up, priced, and checked against the market.

The software matches lanes to your records, works from your rate history and margin floors, and checks current carrier pricing before recommending numbers. That’s important when SONAR is seeing tender rejections anywhere from under 3% to nearly 45%, depending on the lane.

LSP44’s freight procurement agents can then source carriers based on lane performance, gather quotes, and build the response in the shipper’s template. Anything with questionable margins or thin history goes to your people. Everything else moves without the usual email Olympics.

Fast Stops Meaning Reckless

I can practically hear the objections already. Let an AI agent price lanes, negotiate with carriers, and build your bids? Sounds great until someone in the demo says, “Don’t worry, the AI handles everything.” 

I’ve watched that sentence kill a deal faster than a bad rate.

Nobody wants software freelancing with their margins or carrier relationships. Our agents check performance against a network of 280,000+ carriers and 706 million daily events. That scrutiny matters even more after the Supreme Court’s May ruling in Montgomery v. Caribe Transport II.

You also decide how much authority each agent gets:

  • Observe: Flags lanes priced above market.
  • Recommend: Suggests carriers and rates for your approval.
  • Act with approval: Sends RFQs after pricing signs off.
  • Act autonomously: Arranges backup capacity within your dollar limits, with every action logged.

The Bid Keeps Working After You Win It

Giving your agents some freedom doesn’t mean much if they clock out the moment you win the RFP. I’ve seen brokers spend weeks fighting for a contract, celebrate when they land it, and then leave the rates alone until finance discovers they’ve been hauling freight at a loss for three months. 

Some victory lap.

Armstrong & Associates warned about more contracted loads running at negative revenue heading into 2026. Even J.B. Hunt’s brokerage operation saw gross margins fall from 15% to 14.3% in Q3.

I’d rather catch a bad lane before it eats another quarter’s margin. Continuous procurement keeps checking carrier costs against what you’ve promised the shipper, finding backup capacity or flagging trouble while there’s still time to renegotiate.

Then, when that same shipper comes back with a mini-bid, you know exactly what the lane’s earned you. No archaeology required.

Your Pricing Team Gets Its Week Back

There’s one final payoff to catching bad lanes early: your pricing team finally has time to do something besides chase numbers. I’ve yet to meet a pricing leader who hired their best analyst for their extraordinary ability to copy rates between spreadsheets.

Our Carrier Procurement agent benchmarks lanes and sources carriers within your guardrails, doing in seconds what can take hours manually. It can turn around priced, lane-level bids in hours, while Rate & Quote handles spot and contract pricing using live network rates. Your first agent can go live in a week, without replacing your TMS.

I’d much rather have that analyst arguing with sales over a questionable margin than babysitting an inbox.

Your CFO should appreciate the arrangement too. At typical 3PL operating margins of 5-8%, saving $1 can deliver the same operating profit as roughly $16 in new revenue. Try getting that return from another spreadsheet.

What the 2027 Bid Cycle Will Reward

I’ve got a prediction for 2027: the annual RFP gets you invited to the party, but the mini-bids decide whether you get to stay. Shippers aren’t waiting until next January when a lane falls apart in April, and I wouldn’t want my pricing team spending every other Thursday rebuilding rates from scratch.

That’s why we built LSP44. We call it Agentic Infrastructure for LSPs, but the idea is pretty straightforward. Keep carrier sourcing, negotiations, and pricing running between bids, give your people control over the decisions that matter, and keep checking your margins long after you’ve won the freight.

We’re building on infrastructure already used by nine of the world’s ten largest LSPs, and our agents have handled over 1 million automated carrier communications. Worldwide Express President Joel Clum put it nicely: “It’s the difference between software you adopt and infrastructure you build on.”

If you’ve got a 400-lane spreadsheet sitting in someone’s inbox, I’d love to see it. Book a mission briefing, give us 30 minutes, and we’ll show you which parts of your procurement process agents can take off your team’s plate.  

Freight Procurement Software FAQs

What Is Freight Procurement Software?

Freight procurement software helps a broker or 3PL answer shipper RFPs and keep carrier pricing current between them. It cleans up the shipper’s lane file, prices each lane from history and live market rates, collects carrier rates over API, EDI, email, or phone, and builds the response. Your pricing team reviews the exceptions and makes the final calls.

How Does Freight Procurement Software Automate the RFP Process?

It automates the slow parts of a bid. The software normalizes the lane file and prices each lane from your history and margin floor, checked against live rates. Then it requests rates from carriers ranked by how they’ve performed on each lane and builds the bid in the shipper’s template. Thin-margin or no-history lanes go to a person.

Dynamic Freight Procurement vs. Annual Bid Cycles: Which Should Brokers Run?

Brokers end up running both, because shippers set the cadence and many now pair an annual RFP with mini-bids on the lanes that break. The practical move is to price your carrier side every week, so you’re bidding off current rates whether the shipper sends a full RFP in January or a 40-lane mini-bid in June.

Email Tendering Is Slowing Down Our Freight Procurement. What Do We Do?

Connect the carriers that already support API and EDI first, since they often carry a big share of your volume, then put an AI agent on the carriers that still work by email and phone. The agent reads emailed rates and writes them into your TMS. Anything involving payment changes still goes to a person for verification.

Can an AI Agent Award Freight Without Human Approval?

Only if you set it up that way. LSP44 agents run at four autonomy tiers: Observe, Recommend, Act with approval, and Act autonomously. You choose the tier by customer, lane, and dollar amount, so an agent might line up backup carriers on its own for small lanes. At the same time, every award on your biggest account waits for a person.

By Kait Peterson, Head of Marketing at LSP44

The clearest announcement related to AI agents in logistics I read all quarter was a shutdown notice. On Tuesday, September 1, TruckSmarter emailed its users: the company was being acquired, and the app would shut down that Friday.

More than 500,000 carriers had used it, AI dispatch agents and all, and they got three days’ notice and a thank you from the CEO for “five years of trust.”

That’s vendor risk at full speed, and it’s why “will you still be here in 18 months?” now comes up before the demo’s over.

The slower risk is slideware. Launch releases read the same whether the agent runs on live freight or a demo account, and I should know, because I write them. Gartner calls the worst of it “agent washing.”

Agent Watch is our quarterly audit of public AI agents in logistics news, checked like a freight bill.  

And fair warning: LSP44 sells agents too, so our own news gets the same audit.

Which AI Agents Went Live on Customer Freight This Quarter?

Two showed their work: C.H. Robinson’s ocean booking agents, live in China since August, and a freight audit agent from Evos, a $9 million pre-seed startup, working a brokerage’s LTL invoices.

Robinson Brought Receipts

CTO Mike Neill says a decade of digital forms automated about 5% of ocean and air booking requests, since “a customer might have 50 suppliers.” Now, he says, those requests “no longer need to be handled manually.” Robinson’s quoting agents, live since 2025, have sent 80,000+ quotes to 3,600+ customers, in minutes instead of a day or more.

My favorite detail: one agent’s whole job is checking terminal websites daily and watching email for fee notices.

Pre-Seed Money, Real Invoices

The receipt: nearly 3,500 audit decisions at 97% accuracy in two months, per Evos. Its agent learned by watching experienced staff decide, the way most of us learned freight.

Both receipts started from context. Our CEO, Jett McCandless, put it bluntly in September: agents without context are expensive theater.

Which Agent Announcements This Quarter Were Still Pilots or Plans?

Plenty, including parts of the quarter’s best release. C.H. Robinson is still piloting quoting in China, with global ocean booking penciled in for Q1 2027. Freight Technologies’ negotiation agent and our own Carrier Logistics API went out with no customer on record.

50 Negotiations, No Names Yet

Freight Technologies’ Autopilot Negotiator, launched October 27, haggles with 20 or 50 carriers at once over WhatsApp and email, under a rep’s target and maximum price. It waits for one-click approval, which is undoubtedly good freight procurement design.

That said, the release names no customers and no usage count, so for now, you’re the pilot.

Our Own IOU

Carrier service teams are buried in bot emails and calls chasing bad tracking numbers. Carrier Logistics calls it AI bot fatigue, and on October 14 we teamed up on an API that finds the shipment without a clean one.

Nick Ruggiero, our Head of Product, says it removes “an entire category of friction overnight.” Overnight starts when a connected 3PL runs it.

Which Agent Vendors Raised Money or Changed Hands This Quarter?

HappyRobot raised the most we saw: $150 million, at a $1.2 billion valuation, on August 4. Control moved twice: Descartes bought broker TMS provider Tai for $100 million on August 24, and TruckSmarter, which we already mentioned in the intro.  

A Lot of Runway

HappyRobot proved its platform in logistics, and its Series C release counts 150+ enterprise customers, including DHL, Kuehne + Nagel, and Uber. The raise buys a lot of runway, and the release says where it’s flying next: across the supply chain and into insurance, energy and utilities, telecom, and airlines.

Good for them. If your brokerage runs on its agents, ask where freight lands on the roadmap once the airlines call.

Soft Landings and Hard Stops

Tai’s broker customers landed inside Descartes, which paid cash. TruckSmarter’s Dispatch subscribers got three days’ notice and a refund.

Either way, the roadmap changed hands without a vote. Put two lines in every agent contract: change-of-control terms with a data-export clause and a minimum sunset notice.

Who Controls the Data and Pricing Under an AI Agent?

Often, contracts you never signed. In late August, Highway said Motive had restricted its access to Motive ELD data and wanted payment. In October, Citi said WiseTech will charge separately for future CargoWise agents.

Your Visibility, Their Contract

The carriers had authorized the data. The vendors fought over the bill. Highway pulled its Performance Guarantee on loads hauled by Motive-equipped carriers, and a joint statement restored access that week, terms still being haggled over.

Any agent that vets or tracks carriers inherits the contracts underneath it, and so does your supply chain visibility. Ask your vendor to list them.

Read It Like a Rate Con

The six agentic capabilities WiseTech launched last December stay in its Value Packs. It hasn’t said what the next ones will cost. So check which agents today’s contract covers and who prices the next one.

On data, the FAQ has our answer: we never train models for other providers on one customer’s data or resell it to shippers.

How Many 3PLs Actually Run AI Agents in Production?

No one has a clean count, but about half say they’ve started, and 1-in-8 it’s ready. The 2027 Third-Party Logistics Study, out October 6, found 53% of 3PLs have implemented autonomous or agent-based decision support, while 12% feel highly prepared for AI.

The 94% That Made Me Wince

In the study, 94% of providers believe customers are happy with their IT. Only 70% of shippers agree. As a marketer, I know grade-yourself optimism when I see it, and 24 points is a lot of customers who’d take a faster competitor’s call.

The study’s top AI barrier: data quality, cited by 61%.

Gartner Says We’re at the Peak

Gartner’s 2026 CIO survey puts cross-industry agent deployment at 17%, and the firm’s Hype Cycle parks agentic AI at the Peak of Inflated Expectations. Gartner also says fully autonomous agents aren’t ready for most enterprise uses.

Take one test into every demo: ask where a human approves, and where the log lives.

What Should an LSP Do With This Quarter’s Agent News?

Before you sign with a vendor, ask who runs its agent on live freight, and get a clause that lets you leave with your data if the company sells. TruckSmarter’s agents were real, and its carriers still got three days’ notice.

And your own buyers will probably ask the same, since C.H. Robinson’s planned $5.8 billion RXO deal counts Lean AI among its savings levers.

You knew a plug was coming (I run marketing after all). We built LSP44 as Agentic Infrastructure for LSPs so your systems can keep up, with agents like these:

  • Freight Procurement: Sourcing takes hours. Our agent does it in seconds.
  • Carrier Relations: 1 million automated carrier communications, so your reps keep the relationships.
  • Visibility & Exceptions: Agents act at the autonomy tier you set, with every move logged.
  • Freight Audit & Finance: Invoice discrepancies caught before they become disputes.

LSP44 is powered by the same global data graph as project44, connecting 280,000+ carriers, as do nine of the 10 largest logistics providers on the Armstrong & Associates Top 50. We launched profitably, and we’ll never compete with our customers.

Contact us today to book a mission briefing.

By Vinay Mathur, Founding CTO, LSP44

Key Takeaways

  1. Agentic AI for logistics works when each freight step gets the right tool.
  2. APIs still win tracking, rating, tendering, and booking.
  3. Agents earn their keep on exceptions, documents, carriers with no feed, and judgment calls.
  4. Bots that ask carriers what an API already knows add cost and bot fatigue.
  5. Trigger agents on API exceptions and finish the work through APIs.

My favorite agentic AI demo of 2026 is one I’ve seen about a dozen times. An agent calls a carrier, asks where the truck is, gets an answer, and updates the shipment. Everyone seems impressed. Nobody mentions that the truck reported its location through an API five minutes earlier, without making small talk with a dispatcher.

Apparently, we’ve invented a more expensive way to ask a question we already know the answer to.

I spent years building enterprise software, most recently at Stripe, and watching some of these demos makes me wonder when we decided every problem needed a conversation. Sometimes you need an agent to sort through conflicting information, chase down an exception, or make a judgment call. Sometimes you need an API to move data from one system to another. Knowing the difference is central to how we’re building agents and APIs at LSP44.

Gartner has put agentic AI at the Peak of Inflated Expectations, which feels about right. Everyone wants an agent, preferably everywhere, and preferably yesterday.

I think agentic AI for logistics has enormous potential. I also think freight operators who’ve survived two recessions are right to ask whether we’re solving their problems or giving old problems a very expensive personality.

Draw the Line One Step at a Time

The easiest way to avoid giving every freight problem an expensive personality is to ask two questions: Is there one right answer, and does a system already have it?

If both answers are yes, use an API. Nobody needs an AI agent pondering a truck’s GPS coordinates.

Follow a load from quote to invoice, and you’ll see why. Rate lookups, location updates, and routine status checks already have reliable ways to get done. Agents become useful when the information doesn’t add up, something goes wrong, or the next move isn’t obvious.

Gartner made a similar point in June 2025, warning that plenty of supposedly agentic use cases don’t need agents at all. It also predicted over 40% of agentic AI projects would be canceled by 2027.

I’d rather make that distinction before buying anything. Otherwise, you end up collecting agents from different vendors, each bringing another integration to maintain. We call it the Integration Tax. Funny how nobody puts that on the demo slide.

Where APIs Still Win

Nobody should be paying an AI agent to chase information an API can deliver in seconds. Yet that’s exactly what happens when companies put agents in charge of tracking, rating, tendering, and booking. 

APIs do that work faster, more accurately, and at a fraction of the cost.

Take tracking. LSP44’s network pulls carrier updates through APIs roughly every seven minutes. An agent calling dispatch might get through once an hour, assuming someone picks up. And while a dispatcher might optimistically describe a late truck as “making good time,” GPS coordinates have no such diplomatic instincts.

Speed matters, but so does feedback. A bad API request returns an error almost immediately, giving the system something concrete to fix. That feedback loop also helps power our agent factory.

At scale, API-based tracking costs roughly a tenth of call-based coverage. Even smaller carriers can connect through EDI, telematics, or mobile apps.

Of course, getting fresh data and getting a useful answer aren’t always the same thing. That’s where agents get interesting.

Where Agents Belong

The interesting work starts when the API comes back empty, the paperwork makes no sense, or the carrier’s version of events doesn’t match the GPS. Someone has to figure out what’s happening, and I’d rather have an agent chase the answer than another dispatcher lose half an afternoon.

ARC’s Jim Frazer makes a similar case for starting with exception management. We’re already seeing it work at LSP44. Lineage Logistics uses our Dispatch Agent on FTL cold-chain shipments to contact carriers and recover missing shipment identifiers. A small problem, perhaps, until you’re handling thousands of loads.

Paperwork offers another opening. Armstrong & Associates found that only about half of freight forwarders had automated documentation and invoicing by June 2026. That’s a lot of people still babysitting PDFs.

Agents can also help weigh cost against service when choosing a backup carrier. We’ve mapped these opportunities in our guide to agent capabilities and Agent Job Board.

The trick is knowing when the agent should hand control back.

When an Agent Is the Wrong Tool

My patience runs out when an agent starts doing work we’ve already solved with perfectly good software. Calling a dispatcher for GPS coordinates the truck reported five minutes ago is bad enough. Do it across thousands of shipments, and you’ve built an expensive robocall operation that carriers will quickly learn to ignore. 

Bot fatigue is very real and a lousy way to thank people for hauling your freight.

I feel even more strongly about rates and transactions. A language model has no business guessing its way through a tariff table when a rate engine can calculate the exact charge. And tenders, bookings, payments, and settlements need exactly once execution. At Stripe, we used idempotency keys to keep retries from charging customers twice. Freight deserves the same discipline.

Reliability also gets ugly when you string agents together. Twenty steps at 95% accuracy each give you roughly a 36% chance of finishing cleanly.

I’d trust an agent to choose a backup carrier. I’d trust an API to tender the load. Getting that handoff right matters far more than another impressive demo.

Trigger on the API, Finish on the API

Getting that handoff right starts with a rule: let the API spot the problem, give the agent room to solve it within agreed limits, and send the result back through an API. Simple enough, although you’d be surprised how many vendors manage to complicate it.

Take proof of delivery. A poorly designed agent calls every carrier four days after pickup asking for a POD, regardless of whether one exists. Congratulations, you’ve automated annoying your carriers.

I’d rather trigger the workflow when the API confirms delivery, then check what’s missing. Some LTL carriers batch paper PODs days later, while drivers with digital handhelds capture signatures immediately. The agent should know the difference, call only when necessary, and return the document through the API.

At LSP44, customers control that authority through four tiers: Observe, Recommend, Act with approval, and Act autonomously. Permissions vary by customer, lane, and dollar threshold, with every action logged and SLAs tracked through our Observability layer.

Every box in that above diagram has an owner, and the ones your customers see should be yours.

Own the Rules, Embed the Connectivity

Ownership matters because an LSP shouldn’t have to surrender its customer relationships or operating playbook to get better technology. That’s the thinking behind Agentic Infrastructure for LSPs. You own the rules, the decisions, and the customer experience. We provide the carrier connectivity and event data through APIs you embed in your own products.

BlueGrace CEO Bobby Harris said it well at our July launch: “Nobody should be rebuilding carrier connectivity from scratch.” I agree. Connect Once lets you start with the connections your first workflow needs and expand from there. Half of carrier connections go live within hours, and the first agent can deploy in a week.

Of course, trusting us with your rates and lanes deserves more than a nice architecture diagram. We commit to never:

  • Operate as an NVOCC
  • Operate as an indirect air carrier
  • Seek freight brokerage or motor carrier authority
  • Train models for other LSPs on your data
  • Resell your data to shippers or feed it into project44’s shipper products

The principle behind all five is pretty straightforward: we’ll never compete with our customers. You have enough competitors already.

Where Agentic AI for Logistics Goes Next

I hope some of the agents we’re building today are out of a job in two years. Seriously. If an ocean carrier finally offers an API for rates and bookings, why would I keep paying an agent to chase that information? Give the routine work to the API and let the agent handle the next problem nobody anticipated.

That’s where I see agentic AI for logistics heading: fewer agents doing unnecessary work, more intelligence coordinating the parts of freight that refuse to cooperate.

Having said that, the industry has some catching up to do. BCG surveyed 30 major logistics companies in August 2026 and found only 13% seeing measurable financial returns from AI. Even more telling, 80% named a standalone tool as their most impactful AI deployment.

We’ve gotten pretty good at buying clever tools. Connecting them into an operation that actually runs better is another matter.

My bet is that the winners won’t count how many agents they’ve deployed. They’ll know exactly when to retire one.

Build for the Freight, Not the Demo

When I joined LSP44, I said project44 had already done the hardest part of making AI useful: building the data, network, and context to run it in production. I stand by that. We’re talking 280,000+ carriers, 706 million events a day, and infrastructure supporting nine of the world’s 10 largest logistics providers. 

Good luck replicating that with a clever prompt.

Everything I’ve argued here comes back to putting that infrastructure to work intelligently. Let APIs handle reliable data and transactions. Give agents the exceptions worth investigating and the judgment calls they’re authorized to make. Keep the rules and customer relationships yours.

The economics deserve attention, too. At a typical 3PL operating margin of 5% to 8%, saving $1 can deliver the same operating profit as roughly $16 in additional revenue. Our Integration Tax Calculator helps put a number on it.

Better yet, bring us one exception your team still handles manually. We’ll show you where an agent earns its place and where an API can do the job without the theatrics.

Frequently Asked Questions

What Is Agentic AI for Logistics?

Agentic AI for logistics means AI agents working toward a freight objective, like recovering a late load or a missing POD, across your systems and within limits you set. The agent is the worker, and it gets things done by calling APIs.

Should Logistics Automation Use AI Agents or APIs?

Both, split by step. Use an API wherever a carrier system can answer exactly: rate lookups, status checks, tenders, bookings. Give agents the steps that need interpretation, and have them execute through the same APIs so every action is logged.

When Do AI Agents Make Sense in Freight Operations, and When Do They Not?

They make sense on exceptions flagged in API data, emails, and PDFs, carriers with no feed, and calls that trade cost against service. They don’t when a feed already knows the answer, the math is fixed, a transaction must run exactly once, or context is missing.

Do AI Agents Replace a TMS or EDI Connections?

No. Your TMS stays the system of record, and EDI and APIs stay the pipes. Agents read events from those systems and write results back through them. Your TMS manages the record. LSP44 executes the work.

How Should an LSP Measure Whether an AI Agent Is Working?

Count finished work, like exceptions resolved and loads recovered, against exceptions flagged. ARC’s Jim Frazer suggests watching that ratio. If alerts climb while resolutions stay flat, you’ve bought visibility without control. Every agent action should also land in an audit log.

About LSP44

LSP44 is the AI-native agent and API infrastructure business for logistics service providers, built on project44’s data graph, profitable, and headquartered in Chicago.

About the Author

Vinay Mathur is Founding CTO of LSP44 and leads engineering, platform infrastructure, and applied AI for LSP44 and project44. He joined from Stripe, where he led product engineering for its revenue suite and its AI data products and pipelines, after engineering leadership roles at SAP Ariba, GEP Worldwide, and Microsoft Dynamics 365.

By Jett McCandless, Founder and CEO, project44 and LSP44

I get asked this constantly: project44 versus LSP44, which company do you actually run?

I probably deserve the question. My email still ends in project44.com. A September byline in Global Trade Magazine called me “Founder + CEO, project44 + LSP44.” I figured that covered it. Freight people, as usual, wanted a little more detail.

The short answer is both. I’ve run project44 and LSP44 side by side since we split the businesses on July 14.

The longer answer is the one I usually use for brokers, 3PLs, and forwarders after we’ve talked for five minutes. They know project44 works with shippers. They know LSP44 is building AI infrastructure for LSPs. What they really want to know is where the wall sits between the two, because nobody in this business wants to hand over rate data and then wonder who else can see it.

Fair question. It’s also the question that explains the whole setup.

I wrote the clean version in Two Businesses, Two Buyers. This is the version I’d give you over a drink, including why we split the companies and what that means when you’re trying to get an AI business case approved inside your own shop.

What Is LSP44, and How Is It Different From project44?

LSP44 is the company we built for brokers, 3PLs, 4PLs, freight forwarders, and carriers. project44 serves enterprise shippers through Movement, so the cleanest way I know to explain project44 versus LSP44 is to start with the freight itself.

If you own the freight, project44 is built around your operation. If your business is moving freight for somebody else, LSP44 is built around yours.

That distinction sounds obvious once you say it out loud, but it changes almost everything about the product. An LSP already has a TMS, carrier relationships, customer workflows, and its own way of running the business. We built LSP44 to fit into that environment, which is why our agents and APIs sit underneath the operation instead of asking the LSP to rebuild around us.

That’s also what we mean by agentic infrastructure for LSPs. The technology works inside your business, while your customer keeps dealing with you.

Are project44 and LSP44 the Same Company?

No. project44 and LSP44 are separate businesses with their own leadership, engineers, sales teams, and roadmaps. What they share is infrastructure: the largest AI-native agent network, carrier API infrastructure, and logistics data graph connecting more than 280,000 carriers and processing 706 million carrier events a day.

I dispatched trucks back in 1999, so I think about it like a set of doubles. There’s a common tractor underneath, while each trailer has its own freight, customer, and seal. Sharing the tractor doesn’t give anybody permission to break the other seal.

The same rule applies here. project44 doesn’t share shipper data with LSP44, and our commitments to LSPs are in writing. LSP44 won’t become a broker, carrier, NVOCC, or indirect air carrier, use one customer’s data to train models for another, or feed LSP data into project44’s shipper products.

If you build your business on LSP44, we’re not coming after your freight.

Why Did project44 Split Into Two Businesses?

project44 split into two businesses because shippers and LSPs needed different things, and over time we built the company around the shipper side.

The irony is that LSPs were there first. They made up 82 of our first 100 customers, and Worldwide Express wrote our first check. Then shippers started pulling us toward visibility software they could log into, and COVID poured gasoline on that demand.

We followed it. By the time the market cooled and LSPs were dealing with a freight recession, we had a very good shipper business and an LSP customer base waiting behind shipper priorities.

I don’t have a clever way to dress that up. We got better at serving one buyer and worse at serving the other. The fix was to stop asking LSPs to share the line and give them a company built around them.

Does My project44 Relationship Change?

For most customers, no. If you’re an LSP already connected to project44, those integrations and carrier connections keep working, and your data stays yours. 

What changes is the team on the other side. LSP44 has its own R&D organization and forward-deployed engineers working directly with LSP developers on the problems sitting inside their operation.

You can see that in the work we announced with Carrier Logistics on AI bot fatigue. When a tracking number is missing or wrong, automated calls and emails can hammer a carrier’s service desk. We’re building an API that can find the shipment anyway, which is exactly the kind of difficult, specific LSP problem I wanted this team waking up to every morning.

If you’re a shipper, even less changes. Your contract, account team, and project44 relationship stay where they are. I split my calendar between two companies. The project44 team doesn’t, and the business has kept growing with them focused on that customer.

How Should an LSP Build the Business Case for AI Agents?

Start with revenue and customer service. Put the efficiency deal a few slides later.

When I dispatched trucks, people on my team got kidney stones because nobody wanted to leave the two-way radio long enough to use the bathroom. That sounds insane now, but the basic constraint hasn’t changed much: an LSP can only move as much work as its people can absorb.

I wrote about this in August. AI agents raise that ceiling, which means you can take on more freight without making every growth spurt a recruiting exercise.

Then show the cost case. At a typical 5% to 8% operating margin, $1 of cost removed is worth roughly $16 of new revenue.

Still, the part I’d lead with is service. The 2027 Third-Party Logistics Study says shippers value trust and open communication most in a partner. For an LSP, protecting that relationship is how the revenue stays.

Finally: Which LSPs Come Out Ahead Over the Next 12 Months?

The LSPs that come out ahead over the next 12 months will be the ones that can take on more freight, respond faster, and protect the customer relationship without adding headcount every time volume moves.

Capacity is already getting tighter. SONAR’s rejection index has been running near 14%, roughly triple its 2023 to 2025 range, which means more exceptions, more touches, and more pressure on the same teams. Meanwhile, more than 40% of shippers already consider a provider’s AI capabilities when choosing who gets the freight.

That’s why I’d build the AI case around growth and service first.

It also brings us back to project44 versus LSP44. I run both. If you own the freight, project44 is built for you. If your business is moving it for someone else, LSP44 is the company I built around your side of the table.

Bring us the workflow your customers feel most. That’s where I’d start.

Freight tech has never met a promising new term it couldn’t overwork, and right now “agent” is doing double shifts.

AI agents for logistics operations now apparently include everything from software that can spot an exception, understand the rules around it, make a decision, work across systems, and chase down the next move, to a chatbot sitting beside your TMS looking extremely employable in a demo.

We build and deploy these things too, so we’re participating in the same market we’re criticizing. We also know how much plumbing sits underneath an agent that can be trusted with live freight, which makes the current agent-washing boom hard to miss. Gartner looked at thousands of vendors making agentic claims and found roughly 130 with credible products.

But that small group is what makes this whole conversation worth having. The real systems are already doing useful operational work, and Inbound Logistics found that 93% of 3PLs expect AI to have the biggest innovation impact ahead.

So forget the label for a minute and look at the work and fundamentals behind the scenes. 

What Are AI Agents for Logistics Operations, Minus the Hype?

An AI agent for logistics operations is software that can own a defined freight task from start to finish: pull the right context, make decisions within your rules, act across systems and people, handle routine exceptions, and close the loop without someone steering every move.

Take appointment scheduling. A real agent doesn’t stop after reading the email or identifying an open slot. It has to know the customer’s rules, the facility’s quirks, the carrier’s constraints, and what to do when the first option falls apart. Then it books the appointment, updates the TMS, and leaves the workflow finished.

That’s where a lot of so-called agents fall apart. They can interpret the work, summarize it, and recommend what should happen next, but the operator still has to do the job.

A real digital worker earns the name when the work actually leaves the human queue.

Isn’t an AI Agent Just EDI With a Better Haircut?

No. EDI follows a predefined message, format, and route. An AI agent can interpret unstructured information, decide what changed, choose the next action within your rules, and keep the freight workflow moving when the original path goes astray.

If a clean 204 can turn into a tender without a person touching it, keep EDI exactly where it is. It’s cheap, dependable, and doesn’t need a new job title every six months to prove it’s useful.

Freight rarely stays that polite. A carrier emails that the driver is two hours late, or a customer revives a six-week-old thread to move Friday’s delivery to Monday, and now somebody has to interpret the message, find the load, check the rules, update the system, and decide what happens next. An agent can do that work without first forcing the problem into a neat little box.

People still spend far too much time acting as that box. Deep Current found that roughly half of logistics operators rekey shipment data across systems, which is part of what we call the integration tax.

What Does an AI Agent Actually Do in a Freight Operation?

An AI agent does the operational work a person would otherwise have to pick up, chase down, and close out. In freight, that can mean quoting a lane, tendering a load, booking the dock, finding a POD, checking an invoice against the rate con, or figuring out why a truck suddenly stopped looking like it was going to make delivery.

Check calls are a good example of how low the bar has been for too long. Freight has spent decades paying people to ask carriers where the truck is, wait for an answer, then copy that answer into another system. Nobody built a career around being spectacular at check calls.

The same logic applies across the work from quote to settlement. If someone on your team owns a repetitive operational task today, there’s a decent chance an agent can own at least part of it tomorrow, including the annoying exceptions that usually send automation running for cover.

Lineage Logistics already uses our dispatch agent to hunt down missing shipment IDs on FTL loads, and across our customers, agents have handled roughly 1 million carrier communications.

What Won’t an AI Agent Do?

An AI agent won’t own the business around the workflow. 

It won’t hold your operating authority, replace your TMS, carry the customer relationship, or take the furious phone call when a load misses its window. Your people still own the judgment, the liability, and the customer, which is why AI isn’t about to replace freight brokers who are actually good at the job.

It also doesn’t wake up knowing everything your team learned the expensive way. If a carrier looked great on paper and burned you three times in six months, the agent needs access to that history or the rules built from it. Otherwise, it can make a perfectly logical decision your ops team already knows is a bad one.

The vendor has limits too. Ours are spelled out in our FAQ: we don’t hold brokerage or carrier authority, train other providers’ models on your data, sell your data to shippers, or compete with our customers.

Every vendor should be willing to say that part out loud. 

Can AI Agents Actually Run Freight Workflows Without Babysitting?

Yes, provided you decide in advance what the agent is allowed to do on its own.

Nobody wants to wake up and discover some enthusiastic piece of software re-tendered a key account’s freight at 2 a.m. because it technically found a cheaper truck. Autonomy only works when the boundaries are explicit.

Every agent we run can operate at one of four levels, with different rules by customer, lane, workflow, and dollar amount:

  • Observe: Watch the work and flag problems.
  • Recommend: Suggest the next move and wait.
  • Act With Approval: Do the work once a person signs off.
  • Act Autonomously: Handle it end to end and record what happened.

So the ugly, repetitive freight can run while your team sleeps, while a touchy account stays firmly behind human approval. You don’t have to choose between babysitting every move and handing the keys over completely.

The whole point is deciding where you’re comfortable letting go.

Why Do Most AI Agent Pilots Go Nowhere?

Most AI agent pilots go nowhere for three pretty unglamorous reasons: the agent doesn’t know enough about the operation, can’t reach enough of the systems, or isn’t allowed to do enough once it gets there.

That’s why the demo can look terrific, and the rollout dies six weeks later. Sample data doesn’t know Carrier A burned you twice, Customer B has a rule buried in an inbox, or Facility C treats its own appointment policy as optional. Your best people know all of that because they’ve been collecting those scars for years.

Then comes access. If the agent can’t write back to the TMS, contact the carrier, update the appointment, or act without waiting for three approvals, you haven’t automated much. You’ve built another screen for somebody to watch.

BCG found 97% of logistics executives call AI strategic, while only 13% have seen meaningful financial returns.

Our agents start on the project44 data graph already used by nine of the 10 largest logistics providers, because context and access are what turn a pilot into actual work.

Do AI Agents for Logistics Operations Pay for Themselves?

The good ones do, and the math is rigged in your favor. At a 6% operating margin, every dollar of manual work you cut is worth about as much profit as $16 of new revenue.

Cutting costs is also the one lever that works without winning a bid or surviving a procurement cycle. All it takes is knowing what the manual work costs you today.

Our Integration Tax Calculator turns your headcount and the hours they spend moving data by hand into one annual number. That’s the baseline an agent has to beat, and the first one can be working against it in about a week.

The question we hear most from ops leaders is how to get off the phones without hiring another offshore shift. Start with whichever job your team would pay you to take away.

How Do You Catch a Chatbot Pretending to Be an Agent?

Ask anyone selling AI agents for logistics operations these five questions, and watch the sales engineer’s face:

  1. What did the agent know about this carrier before it acted?
  2. Which system does it write back to, and who keeps that connection alive?
  3. Which autonomy tiers does it support, and who sets them?
  4. What happens to the data it collected if you walk away?
  5. What does the audit log show when something goes wrong?

If “roadmap” comes up before question three, end the meeting early and go get lunch. Anyone who survives gets one last question: is this built for the LSP, or retrofitted for you?

We hold ourselves to the same test. Our head of product graded five agent jobs in The Agent Job Board and admitted where a specialist might beat us. Our guide to AI agents for freight covers the rest of the field.

Where Are AI Agents for Logistics Operations Headed?

Into your budget, whether you invite them in or not. 

Gartner research cited by Logistics Management has agents taking almost half of supply chain software spending by 2029, up from about 11% this year. BCG and Alpega found more than 40% of shippers already weighing an LSP’s AI when they decide who gets the freight.

Our bet is that most agent-only startups get bought or buried within 18 months. The LSPs that come out ahead will be the ones that spent this year wiring up real infrastructure while everyone else collected pilots like trading cards.

We were built for that job at LSP44. We’re the AI-native agent and API infrastructure business for logistics service providers, run by people who came up inside these operations. Our agents show up on day one already knowing a decade of carrier history, and they work under your brand on whatever leash you choose.

Your TMS manages the record. LSP44 executes the work.

Contact us to book a mission briefing. Bring the job your team hates most, and we’ll show you what a real agent can do with it.

We put “agentic” on the LSP44 homepage when we launched in July, which means we don’t exactly get to stand outside the circus and laugh at the tent.

We’re in it, and we used the word on purpose. The trouble is freight tech got hold of “agentic” and saw it like free merch at a trade show. Suddenly old automation had a new name, workflows became agents, and software that still needs a human checking its work started showing up in decks as autonomous. The industry can survive a little vocabulary abuse, but this one is getting expensive.

Gartner already has agentic AI at the “peak of inflated expectations,” and LSPs are starting to put real budgets behind products carrying the label. At that point, arguing over the word gets boring. You need to know what the software can do, what it knows, what it can reach, and what happens when the clean demo meets the reality of a chaotic afternoon in freight.

That’s where agentic infrastructure for LSPs comes in. 

Our Logistics Language Guide gives you the short definition, but now we’re getting into the part a homepage rewrite can’t fake: the foundation of the agent for any of this to work.

What Is Agentic Infrastructure?

Agentic infrastructure is everything around an AI agent that lets it keep working once freight stops behaving like the demo. 

Think of it this way. An AI agent is a worker. Agentic infrastructure is the building the worker shows up to.

The agent itself might book an appointment, chase a carrier, or update a system. Useful, sure. But the real job gets harder the minute it needs context from somewhere else, access to another platform, help from another agent, or a memory of what happened three steps ago. Freight has a habit of producing those moments all day long.

In a nutshell, that’s why the infrastructure matters. It’s the data, connections, models, memory, and audit trail that let agents work together across an operation instead of living as a bunch of isolated tricks.

What Does Agentic Mean in Freight Software?

In freight software, “agentic” means AI can take a goal, figure out what needs to happen next, and keep the work moving without a person directing every step.

That’s the important part: the software isn’t waiting for someone to click through a preset workflow. It can read what’s happening, make a decision, take action, and adjust when the situation changes. The AI agent is the worker doing that job, while “AI-native” describes the company or product built around AI from the start.

The confusion starts when every chatbot or automation tool gets called agentic anyway. Gartner calls that “agent washing,” and in June 2025 it found only about 130 vendors with real agentic capability among thousands making the claim.

What Is Agentic Infrastructure for Logistics Service Providers?

Agentic infrastructure for LSPs gives providers the brains and connective tissue for AI agents without asking them to hand over the customer experience.

That distinction is bigger than it sounds. LSPs already have a TMS, a portal, a brand, and plenty of logins. They don’t need another software company elbowing its way into the relationship. The technology should plug into the operation they already run, so their agents can work across systems while the customer keeps dealing with the LSP they hired.

It also means the infrastructure provider has to know its place. We don’t broker freight, hold NVOCC or IAC authority, train on your data for somebody else, or resell it to shippers.

What Are the Five Layers of Agentic Infrastructure?

Once you get past the label, this is where agentic infrastructure either holds together or falls apart. LSP44 breaks it into five layers, and the agent itself is only the top one that everyone notices in a demo. The other four layers are what give it enough context, reach, judgment, and accountability to do anything useful once real freight gets involved.

From the bottom up, it looks like this:

  • Network: The live carrier graph: 280,000+ carriers, every mode, always running, always learning.
  • Observability: Agent health monitoring, SLA tracking, alerting, and audit logs. Every action recorded.
  • Intelligence: Geofencing, truckload autocorrect, ETA logic, milestone inference, and timestamp correction, trained on a decade of real freight.
  • Integrations: API, EDI, AI voice and messaging, video OCR, and document integration, in whatever dialect your carriers use.
  • AI Agents: Carrier Procurement, Rate & Quote, Tender & Booking, Ocean Execution, Dispatch & Appointment, Carrier Onboarding, Visibility, Exception & Disruption Recovery, Documents & eBOL, and Freight Audit & Settlement.

None of this asks the TMS to pack its bags. It still owns the system of record, while these layers give agents what they need to handle the work around it. They’re doing different jobs, which is why replacing the TMS was never the point.

Why Do AI Agents Need Operational Context in Freight?

AI agents need operational context in freight because good decisions depend on history, not just whatever happens to be on the screen right now.

A carrier that’s great in one lane can be a problem in the next. One receiver treats an appointment like gospel; another treats it as a rough suggestion. Even a stale GPS ping only means something if the agent knows the shipper, the lane, the facility, and what usually happens next.

At LSP44, we bring that kind of context into the decision. Our agents draw from 11 years of freight history across 1.5 billion shipments and 706 million carrier events a day. CEVA Logistics co-developed with us after evaluating agents and context separately, then seeing how much more useful they became together.

Without shared context, point agents handle pieces of the job while people still carry the operation in their heads.

How Much Autonomy Do the Agents Actually Have?

Once the agent has the context, autonomy comes down to how much authority you’re comfortable giving it. You set that by customer, lane, dollar threshold, or whatever else carries risk in your operation.

There are four levels:

  1. Observe: Watches and reports, but doesn’t act.
  2. Recommend: Suggests what to do and waits.
  3. Act with Approval: Takes the action once a human signs off.
  4. Act Autonomously: Executes inside the guardrails you set, with every action logged.

We built it this way because nobody wants an AI freelancing on a customer’s freight at midnight. Transport Topics found the same thing in its 2026 Top 100 survey: 3PLs want automation inside guardrails, with people still involved where judgment carries weight.

Start conservative, watch how the agent behaves, then give it more room as it earns it.

So Where Does Agentic Infrastructure for LSPs Go From Here?

Our bet is that nobody will be arguing about the definition of “agentic” 18 months from now. By then, enough of these projects will have hit real freight for buyers to know the difference between an agent that demos well and an operation that can trust one.

Some of them are going to fail. Gartner already expects more than 40% of agentic AI projects to be canceled by the end of 2027, while Forbes’ July analysis points to the reasons: fuzzy value, weak governance, bad controls. Read back through the five layers, and you can see the problem. Plenty of companies bought the agent before figuring out what had to exist around it.

LSPs don’t have much room for that experiment. The Tabi Pricing Pressure Index had broker-awarded margin fall from 16.5% to 14.7% in a single month. At 5% to 8% operating margins, removing a dollar of cost can do the work of roughly $16 in new revenue. Suddenly, the conversation gets a lot less philosophical.

LSP44 was built for that version of the conversation. Five layers of agent and API infrastructure, already running inside nine of the world’s 10 largest logistics providers, with no brokerage arm, no freight to steal, and no interest in getting between an LSP and its customer.

The circus can keep arguing about who gets to use the word “agentic.” If you’ve already sat through four demos, bring us the workflow your team is still doing by hand. We’ll start there.

By Jett McCandless, Founder and CEO of LSP44 and project44

I’ve had some version of this conversation over and over since July: What’s the difference between project44 and LSP44?

So let me save everyone a few emails.

On July 14, 2026, we split project44 into two businesses because we were serving two buyers who wanted very different things from us. project44 is built for shippers. LSP44 is built for logistics service providers.

I run both, which is probably why some of the confusion has stuck around. They share the carrier network we spent 11 years and roughly $1 billion building, and there’s plenty of technology under the hood that connects them. But a shipper running a global transportation network and an LSP trying to serve 200 shipper customers do not have the same job.

We finally built around that reality.

These are the five questions I keep getting about project44 and LSP44, and the straight answers I’ve been giving.

What is LSP44?

LSP44 is the business we built for 3PLs, 4PLs, freight brokers, and forwarders.

The cleanest way to understand it is this: if you move freight for somebody else, LSP44 is built for you.

Your job is different from a shipper’s. You’re serving multiple customers, working across different systems, and trying to make all of it look seamless from the outside. LSP44 gives you the AI agents, APIs, and carrier connectivity to do that inside your own TMS and customer experience, under your own brand.

We call it Agentic Infrastructure for LSPs because the agents are only part of the story. Connect Once means we connect the systems a workflow needs, then keep building on the same foundation instead of starting over every time.

That foundation has 5 layers: 

  1. Network
  2. Observability
  3. Intelligence
  4. Integrations
  5. AI Agents 

What is project44?

project44 is our business for shippers. If your company owns the freight and you’re responsible for getting it where it needs to go, this is your side of the house.

More than 1,000 brands use Movement to manage transportation, track freight across modes, run yards, and handle the last mile. Since we spun off LSP44 in July, there’s no ambiguity inside the company about who we’re building project44 for: the shipper.

That’s a meaningful change from where we started. project44 was founded in Chicago in 2014, and some of our earliest customers were LSPs. Over time, the shipper business became a very different animal. Our Intelligent TMS launched in August 2025, and shipper revenue grew 57% year over year in fiscal Q2 2027, faster than any other project44 product line.

The outside validation followed. IDC named project44 a Major Player in worldwide TMS applications, and G2 gave us our 22nd straight season of recognition.

If you already use project44, nothing got tossed in the air when we split the businesses. Your contract, pricing, and account team stay the same. We just made it much clearer who gets our full attention.

Are project44 and LSP44 the Same Company?

No. project44 and LSP44 are separate businesses.

They have their own leadership teams, engineers, sales organizations, and product roadmaps. I’m Founder and CEO of both, and Vinay Mathur is CTO of both, but the teams under us run independently because they’re building for different customers.

What they share is the infrastructure underneath them: the carrier network, carrier APIs, logistics data graph, and agent stack. That foundation reaches more than 280,000 carriers and processes 706 million carrier events a day.

There was no reason to rebuild all of that twice. The important distinction is what sits on top of it. project44 builds for shippers. LSP44 builds for logistics service providers.

Does Customer Data Move Between project44 and LSP44?

No. Shared infrastructure does not mean shared customer data.

I know that’s the first place a broker or forwarder’s mind goes. If project44 serves shippers and LSP44 serves LSPs, you want to know whether your rates, lanes, customers, or margins can somehow cross that line.

They can’t. LSP44 put the commitment in writing. It will never:

  • Operate as an NVOCC
  • Operate as an indirect air carrier
  • Seek freight brokerage or motor carrier authority
  • Train models that serve other logistics providers on one customer’s data
  • Resell customer data to shippers, or put it into project44’s shipper products

LSP44 won’t compete with its customers. project44 makes the same commitment in the other direction: shipper data doesn’t flow to LSP44 or to the providers running on it.

Why Did project44 Split LSP44 Into a Separate Business?

Because trying to serve shippers and LSPs from the same product organization eventually started making both sides worse.

Shippers want applications they can log into and run. LSPs need infrastructure they can build into their own operation: APIs, SDKs, data feeds, and agents that work behind their brand.

Those are different products, different buying motions, and increasingly different roadmaps.

COVID pushed project44 hard into shipper digitalization. At the same time, the freight recession crushed LSP technology budgets, and I’ll own this part: we stopped understanding that customer as well as we should have. I tell the longer version on our company page.

Then LSPs started adopting agents faster than anyone expected, which widened the gap even more.

So we stopped pretending one roadmap could serve both equally well.

LSP44 now gets one test for every product decision: Was this built for an LSP from the start, or are we trying to make shipper software behave like it was?

Does My p44 Relationship Change?

Yes, but not in the way that makes anyone nervous.

Your integrations, carrier connections, and underlying infrastructure keep working because LSP44 runs on the same foundation. Your data stays yours. You’re not being asked to rip anything out or start over.

What changes is who wakes up thinking about your problems.

LSP44 now has an R&D team focused only on LSP workflows, plus forward-deployed engineers who can write production code alongside your developers. Andy Simmons, who spent four years building at project44, now leads engineering. Robert Zehentbauer, a longtime Kuehne+Nagel operator, leads sales in EMEA North.

Customers are already leaning further in. Net retention rose 393 basis points last quarter, and most customers now run three or more agents at once, according to Q2 FY27 results.

If your request used to sit in a shared queue, that queue is gone.

The Split Only Matters If It Makes LSPs Better

I think the next 12 to 24 months are going to be pretty unforgiving for LSPs still carrying too much manual work.

Armstrong & Associates sees capacity tightening as the freight recession finally burns off. At the same time, the 30th Annual Third-Party Logistics Study says shippers still expect more from 3PL technology than many providers are giving them.

I’ve spent enough time around this business to know what happens next. When the market gets tighter, nobody suddenly finds extra hours in the day.

At a 5% to 8% operating margin, taking $1 of cost out can be worth roughly $16 of new revenue. That’s why I care less about flashy AI demos and more about whether an agent can take a tender, exception, document, or audit off somebody’s desk for good.

LSP44 runs on infrastructure already used by nine of the world’s 10 largest logistics providers, based on the Armstrong Top 50, and we publish a one-week time to first deployment.

If you want to know where to start, don’t start with AI. Start with the workflow your team complains about every week. Use the Agent Job Board to pressure-test it, then bring it to us.

By Jett McCandless, Founder & CEO, LSP44

In 1999, I dispatched trucks out of a terminal in Portland, Oregon, for $12 an hour. My career plan was pretty simple: get promoted, make another $2 an hour, and use the windfall to upgrade my Friday nights at the sticky-floored bars downtown.

Dispatch had a way of making those well drinks feel medicinal.

We were tied to two-way radios for entire shifts. Bathroom breaks were enough of a luxury that dispatchers getting kidney stones was an actual occupational hazard. Load updates moved through the building one phone call, radio transmission, and desk at a time. The integration layer was basically a kid carrying information around, and most days, I was the kid.

Everyone on that floor knew where the operation fell apart because we spent all day patching it back together by hand. What we didn’t have was anyone who could build something better.

In fact, whenever I suggested a different way to run the desk, the answer was some version of, “That’s an engineering problem.” And with engineering working at corporate, I might as well have been yelling at the clouds.

Sure, every once in a while, somebody from headquarters would visit, ask 100 people what they needed, and get 100 answers. But of course, nothing ever really changed. 

Twenty-seven years, one 3PL, and a little company called project44 later, I still walk into LSPs that have spent millions on software and discover something strangely familiar.

The integration layer is still a person, and thanks to that, you have an integration tax.

What Is the Integration Tax?

If the integration layer is still a person, the integration tax is everything you pay because that person still has to be there.

It’s the engineering hours spent building and maintaining connections between systems that were all sold as solutions. It’s the analysts mapping EDI, the IT team babysitting APIs, and the operations people filling whatever gaps are left when those connections don’t quite work.

Those last costs are the easiest to miss because they look like normal work.

You won’t find a line item called “integration tax” on your P&L. You’ll find it scattered across payroll, IT, implementation costs, support tickets, and thousands of little manual fixes your team has gotten very good at treating as part of the job.

Who Pays the Integration Tax Today?

The people paying the integration tax are the people doing the work your integrations were supposed to eliminate.

In 1999, that was me with a two-way radio, carrying information from one part of the operation to another. Today, it’s the coordinator copying data from a customer portal into the TMS, the offshore team digging shipment details out of PDFs, or the engineer keeping an API alive because somebody changed a field and neglected to tell anyone.

We’ve spent 27 years putting better software around the problem, but the bill never went away. It just got spread across operations, IT, billing, customer service, and engineering, which makes it remarkably easy to underestimate.

What Does One More System Really Cost an LSP?

Once the integration tax is spread across five departments, adding one more system rarely looks expensive. That’s the trick.

You pay for it twice: once for the license, and again for the people keeping it connected.

The second bill starts at go-live:

  • An analyst maps the EDI.
  • Engineers lose weeknights when a carrier renames a field.
  • Coordinators rekey tenders that arrive as PDFs.
  • Billing hunts for PODs that never attached.
  • An account manager explains why the portal says “in transit” on a load that already delivered.

Procurement can spend a month grinding a few points off the license. I’ve sat on both sides of that table. The higher cost usually shows up later, buried in payroll, support, and hours your team never expected to spend keeping another piece of software alive.

That’s the number I’d want before I signed anything.

How Do You Calculate the Integration Tax at Your Company?

Grab a napkin. Multiply your operations headcount by the fully loaded cost per person, then by the share of their day spent moving data or chasing status. The result is the integration tax your P&L already pays.

My napkin, in round numbers:

  • 50 people in operations, each paid $50,000.
  • Benefits and payroll taxes are 30.1% of private-industry compensation costs, per BLS data for March 2026, so each person costs about $71,500.
  • 20% of their day goes to rekeying, portal checks, status calls, and cleanup.
  • 50 × $71,500 × 20% = $715,000 a year.

Don’t trust my 20%? Good. Spend a day on the floor with a notepad and count it yourself.

Calculate Your Integration Tax

Anyone who moves information between systems as part of the job. Dispatch, track and trace, carrier onboarding, billing.

Salary plus benefits, payroll taxes and overhead. Your finance team has this number.

Rekeying a load into a portal, copying a rate out of email, chasing a status a system already has.

Fill in all three fields. The number appears here as you type.

Your integration tax

$0

every year

  • 0full-time people, doing this and nothing else
  • 0hours a week, across the team
  • $0over three years, undiscounted

Revenue you would have to add to earn the same money

$0

At a 5 to 8 percent operating margin, a dollar of cost removed does the work of about sixteen dollars of new revenue. Sixteen is the middle of that band. At an 8 percent margin it is twelve and a half. At 5 percent it is twenty.

Run your numbers through the Integration Tax Calculator and compare the result to every license price on your short list.  

How Much New Freight Does It Take to Match a Dollar Saved?

About $16 of it, at the 5% to 8% operating margin a typical 3PL runs. Cut a dollar of integration tax and the whole dollar lands in operating profit, while a new dollar of revenue only brings its margin along. 

At 6.25%, you’d need $16 of new business to earn the same dollar ($12.50 at 8%, $20 at 5%).

Those numbers come from our May 2026 strategy work. With real assumptions behind them, it’s game over.

Armstrong & Associates’ 2026 market report puts domestic transportation management gross margin at 15.3% for 2025, barely down, so the fight is over cost per load. I think the next 24 months will sort LSPs into haves and have-nots faster than the last 10 years did. Whoever sits it out has close to zero chance of winning.

What Is $715,000 of Integration Tax Worth in New Revenue?

Roughly $11.4 million of new freight at a 6.25% margin, which is a lot of napkin. Depending on your margin, call it $8.9 million to $14.3 million.

You won’t capture all of it, and that’s fine. Even half is the profit on about $5.7 million of new business, collected by absorbing the next wave of volume without the next 10  hires.

At a 5% to 8% margin, every dollar of integration tax you remove is worth roughly $16 of new revenue. Ask your sales team which one is easier to find.

Do Single-Purpose AI Agents Lower the Integration Tax?

Sometimes. But a single-purpose agent can just as easily create more integration work if it doesn’t understand how your operation runs in real life.

Take POD collection. On paper, it sounds easy: once a load delivers, go get the POD. But freight is rarely that clean.

Some carriers have the signed POD immediately. Others scan paperwork a day or two later. Some terminals still send documents to a larger facility before anyone uploads them. A load may also sit for days waiting on an appointment, which makes a simple “four days after pickup” rule useless.

Now your engineers have to teach the agent every one of those exceptions.

That’s the problem with narrow AI built without operational context. It may automate one task, but you can end up spending a lot of time and money teaching it how your business works before it saves you anything.

So Then What Happens When You Stack Single-Purpose Agents?

When you start stacking integration work right along with the agents, your AI budget turns into an integration budget before you know it. Each single-purpose agent takes one task off the job board and hands your people one more connection to babysit.

Some vendors call the result “mostly automated.” I call it the most expensive way to run anything, since you pay for the software and the person who finishes its job.

Fair warning, I’ll take some heat for this. A whole crop of startups is raising money to build a worse version of what we already run at global scale. I’d bet a good number of them get acqui-hired or fold within 18 months, leaving their connection in your TMS for your team to inherit.

Why Do AI Agent Pilots Drag On for Months?

Because the pilot keeps discovering the business in real time, and engineering has to teach the agent every new lesson.

The first few weeks feel great. The agent collects some PODs, the dashboard starts moving, and the project sponsor is understandably fired up. Then the exceptions start arriving. One carrier scans paperwork two days late. Another terminal does it differently. A customer has its own rule. Each one turns into another ticket, another fix, and another round of testing.

Six months later, somebody proudly announces that the agent collected 1,000 PODs.

Great. What percentage of total volume is that?

Sometimes it’s 4%.

That’s how these pilots drag on. The team celebrates activity while coverage inches upward, and by month eight you may have six engineers and $250,000 to $500,000 tied up in one very narrow job.

Where Does an AI Agent Get Its Context?

From freight it’s already seen. That’s why the network behind an agent matters more than the agent. Poor data leads to bad decisions, whether a physical worker makes them or a digital one.

At LSP44, collecting a POD is a button, because we’ve spent 11 years learning how carriers and terminals handle paperwork. Your engineers paid for everything the agent didn’t know in that pilot. Context is the product, and it can’t be raised in a funding round.

People’s eyes glaze over when I say our data graph covers 1.5 billion shipments a year, or close to 48 every second. No single LSP will ever see that much freight on its own, and I don’t care how big you are.

What Does Connect Once Actually Change?

Connect once lets the second system start where the first one left off. The systems and channels your first workflow needs get connected one time, and later workflows build on them. Your TMS and existing integrations stay put.

A second agent becomes a configuration job instead of another integration project, and your operators finally stop being the integration layer. I’d have killed for that in 1999.

The agent is the part you see. The other four layers are why it works: network, observability, intelligence, and integrations. Half of our new carrier connections go live within hours through the Connection Center, and the first agent deploys in a week.

What Is LSP44 Built On?

LSP44 is built on the network LSPs helped us create starting in 2014.

Back then, freight still ran on a heroic amount of phone calls and EDI, and Worldwide Express was the first LSP to write us a check. Since then, that network has grown to 280,000+ carriers and 706 million carrier events a day. Nine of the 10 largest logistics providers on the Armstrong & Associates Top 50 already run on it.

LSP44 takes that foundation and turns it into agentic infrastructure built specifically for LSPs. BlueGrace founder and CEO Bobby Harris put it better than I could: It lets his team put engineering into what actually differentiates BlueGrace.

That’s really the point. Your engineers should be building your advantage, not rebuilding freight connectivity.

Finally: Can You Trust One Provider With Every Connection?

You can, but only if that provider has no reason to turn around and compete with you.

That line matters more once you start connecting your TMS, customers, carriers, workflows, and agents through one infrastructure layer. LSP44 is built for LSPs, and we intend to stay on your side of the table. We will never:

  • Operate as an NVOCC.
  • Operate as an indirect air carrier.
  • Seek freight brokerage or motor carrier authority.
  • Train models that serve other logistics providers on one customer’s data.
  • Resell customer data to shippers, or make it visible inside project44’s shipper products.

We also built LSP44 as a business, not as a venture experiment we hope someone acquires before the runway ends. We’ve been profitable since day one.

After 27 years in this industry, I have very little interest in giving LSPs another connection to babysit. The whole point is the opposite: connect once, put agents to work across the operation, and let your people spend less time holding the stack together.

Agents deployed. Exceptions handled. Book a mission briefing and bring your calculator number.

We named ourselves LSP44 because “logistics service provider” describes exactly who we build for. Then we started talking to more people in American freight and discovered a small caveat: plenty of them don’t call themselves LSPs.

A broker in Chicago says broker. A 3PL says 3PL. A forwarder says forwarder. A warehouse operator definitely isn’t introducing himself at dinner as a logistics service provider. In Europe, though, LSP is normal lingo for the whole group.

Freight already has enough acronyms to qualify as its own dialect, so normally we’d leave this alone. But the terminology hides a useful point: these companies may all move or manage freight, yet they can run very different businesses.

Armstrong & Associates breaks the market into four major segments, each with its own economics, operating model, and technology needs. Since we build for this industry, those differences are a lot more interesting to us than arguing over vocabulary.

So what actually counts as an LSP, how big are the four segments, and why does an American 3PL in Chicago sound like something completely different in Hamburg?

What Does LSP Mean in Logistics?

LSP stands for logistics service provider, a company that moves, stores, arranges, or manages freight owned by someone else. The shipper owns the goods. The LSP sells the work of getting them where they need to go, using its own trucks and buildings or capacity it buys from carriers.

The term is also an umbrella. Brokers, freight forwarders, 3PLs, third-party warehouses, and 4PLs all sit under it, along with plenty of companies that are all of those at once.

If the goods belong to your customer and your company gets paid to handle them, you’re an LSP, whatever your letterhead says.

It’s also important to note that the overlap with 3PL is large but incomplete. 3PL names the provider by its seat on the contract, the third party next to shipper and carrier. LSP names what the provider sells. 

Which is why it stretches to cover 4PLs and, in most European usage, carriers too.

Which Businesses Count as a Logistics Service Provider?

The definition holds five kinds of business, and big providers often run three or four under one logo.

  • Third-Party Logistics Providers (3PLs): Run outsourced transportation and warehousing programs across segments. In the U.S., it’s also the umbrella term.
  • Freight Brokers: Arrange truck capacity they don’t own and earn the spread between shipper rate and carrier cost, under FMCSA broker authority.
  • Freight Forwarders: Arrange international air and ocean moves, including documents and customs. U.S. ocean forwarders and NVOCCs need a Federal Maritime Commission license.
  • Third-Party Warehouses: Store and handle inventory for other companies under contract. Cold storage operator Lineage ranks 27th on A&A’s global 3PL list at $5.4 billion.
  • Fourth-Party Logistics Providers (4PLs): Orchestrate the other providers for a shipper, usually asset-light. Andersen Consulting coined and trademarked the term in 1996.

Carriers are the edge case. A carrier sells its own capacity, though the line between the two can run through a single company. For example, Carter Logistics tracks loads as an LSP while its own fleet connects to the same network as a carrier.

How Big Are the Four Segments of the LSP Market?

On paper, the U.S. LSP market is enormous. Armstrong & Associates puts 2025 U.S. 3PL revenue at $323.4 billion, up 5% from the year before.

The interesting part is what’s hiding inside that number. A&A divides the market into four very different businesses, plus another $4.5 billion in contract logistics software.

  • Domestic Transportation Management (DTM), $128.3 billion: The biggest piece of the market includes freight brokerage, managed transportation, intermodal, and last-mile. It grew 4.5%, and brokerage alone makes up 83% of the segment.
  • International Transportation Management (ITM), $85.9 billion: Air and ocean forwarding, customs, and trade compliance. It grew 7.7%, making it the fastest growing of the four.
  • Value-Added Warehousing and Distribution (VAWD), $72.7 billion: Contract warehousing and distribution center management, up 4.4%. Short-term public warehousing sits outside this category.
  • Dedicated Contract Carriage (DCC), $32 billion: Trucks and drivers dedicated to a single customer under contracts that typically run one to seven years. Revenue grew 1.6%.

Zoom out, and A&A puts the global 3PL market at $1.3 trillion, with Asia Pacific accounting for 37.5% and North America 28.9%.

Useful numbers, sure. However, they can also fool you into seeing one giant logistics market. A brokerage earning a spread on a load has very little in common economically with a warehouse running a multiyear contract. 

The real story starts when you look at how each of these businesses earns its money.

Why Do LSP Segments Make Money So Differently?

Because they’re selling different things.

A freight broker largely resells transportation capacity it buys from carriers. A warehouse sells its own space and labor. A dedicated fleet puts its own trucks and drivers to work for one customer. Those business models can all live under the LSP umbrella, but the money moves through them very differently.

A&A’s net revenue figures make that obvious. DTM is the biggest segment by gross revenue at $128.3 billion, yet only $19.6 billion remains after purchased transportation is removed. Warehousing, meanwhile, leads the market in net revenue at $56.1 billion. DCC reports the same $32 billion in gross and net revenue because the fleet itself is what the customer is buying.

That leaves brokers operating on a much thinner slice. DTM kept about 15 cents of every gross dollar in 2025, compared with 35.4 cents for ITM.

When your economics look like that, manual work gets expensive very quickly. At a typical 3PL operating margin of 5% to 8%, cutting $1 of cost can have roughly the same profit impact as adding $16 of revenue.

So yes, a warehouse and a brokerage can both call themselves LSPs. They just shouldn’t be buying the same software for the same reasons.

Why Is LSP the Standard Term in Europe?

Because Europe never chopped the industry into as many separate boxes as the U.S. did.

In Germany, logistikdienstleister literally means logistics service provider. It’s a broad, everyday term that can cover a forwarder, contract logistics operator, road carrier, ocean line, or parcel network. Fraunhofer SCS uses it exactly that way in its TOP 100 study of the European logistics market.

That broader definition explains why the European market looks so huge on paper. Fraunhofer says businesses spent nearly €1.6 trillion on logistics in 2023, with about 55% going to outside providers.

Compare that with A&A’s much narrower European 3PL estimate of $229.7 billion for 2025. Different year, different currency, no doubt. But the bigger difference is simple: Europe counts more kinds of logistics companies under the same roof.

That’s also why “LSP” lands naturally with many of our EMEA customers. It’s the language the market already speaks, which is one reason we added a Head of Sales for EMEA North in August.

Why Do U.S. Operators Say 3PL Instead of LSP?

U.S. operators say 3PL instead of LSP because 3PL became the American industry’s catch-all term decades ago, while regulators pushed companies to identify themselves by the specific job they perform.

A broker is a broker. A freight forwarder is a freight forwarder. A carrier is a carrier. FMCSA makes companies choose their type of operation when they register, and ocean freight has its own licensing structure. Even the FDA uses “third-party logistics provider” in its drug supply chain rules.

Meanwhile, 3PL became the broader commercial label. A&A has tracked the U.S. 3PL market since 1994 and still publishes its rankings under that name. Thirty years of reports, conferences, sales decks, and company bios will make a term stick.

We kept LSP44 broader because logistics companies have a habit of refusing to stay in one box. Brokers add warehousing, carriers start brokering freight, forwarders add managed transportation, and so on. 

The specific label still tells you what a company does today, but LSP leaves room for what it becomes next.

So Why Did LSP44 Put the Acronym in Its Name?

After all of that, the name LSP44 is pretty literal.

We build for the whole LSP world: brokers, forwarders, carriers, warehouses, and the companies that have become some combination of all four. That goes back to the earliest days of project44, when logistics service providers like Worldwide Express, BlueGrace, and DSV helped build the carrier network underneath what came next.

LSP44 brings that focus back to the operator, this time with AI-native infrastructure. Connect the systems and channels once, then let AI agents work across them with context from 280,000+ carriers and 706 million carrier events every day.

Nine of A&A’s 10 largest global providers already run on that infrastructure. We also hold no broker authority and no NVOCC license because competing with our customers would be a strange way to build for them.

So yes, we know Americans usually say 3PL. We kept LSP because the industry is bigger, messier, and more interesting than one acronym can capture.

If you run one of these businesses, bring us the workflow your team is tired of doing by hand. That’s a much better conversation than arguing over what to call you.

By Nick Ruggiero, Head of Product, LSP44

In August, I got on a call with a VP of operations at a brokerage doing about $300M a year. He shared his screen, and there it was: the spreadsheet. Five rows down the side, one per job he wanted off his desk. Seven vendors across the top. Every cell a green checkmark, which is the freight tech equivalent of a dating profile where everyone likes hiking.

He’d started out comparing freight analytics platforms, the way everybody does, and three months later he was looking at agents and asking me which column to pick.

I told him I couldn’t, because his spreadsheet contained no information. He’d asked seven sales teams if they could do a job, and seven sales teams said yes. I spent most of my career building platforms for shippers, so I’ve drawn that checkmark plenty. It’s an honest answer and a useless one.

So I asked if I could redo the spreadsheet with him on the call, and he said sure, in the voice of a man who’s been on many vendor calls. What follows is that hour: two definitions, his five rows, the five questions that weren’t on the grid, and the two rows where I told him to spend his money somewhere else.

First and Foremost: Two Starter Questions

ARC Advisory said something this spring I keep repeating: supply chain leaders have plenty of AI claims and hardly any proof. So before we touched a row, I answered two fundamental questions for him.

What Is a Freight Analytics Platform in 2026?

A freight analytics platform takes your shipment, rate, carrier, and event data and turns it into decisions about cost, capacity, and service. That definition hasn’t moved in a decade. 

What moved, however, is what sits on top. Through 2024, the output was a dashboard somebody squinted at before going and doing something. Now the output feeds an agent that does the something.

He had DAT iQ and rate intelligence bolted onto his TMS, both fine. They tell you what a load should cost, and they’ve never moved one, and most buyers discover those are two different products after the contract is signed. A directory lists options. An operating system decides, executes, and learns from every load.

Why Has the Freight Analytics Platform Question Turned Into an Agent Question?

Because an agent is exactly as smart as the data underneath it, and no prompt is clever enough to fix that. Hand two agents the same instruction with different context and you get two different answers, and I’d bet one of them just re-tendered a load to a carrier with a conditional safety rating and felt great about it.

Gartner’s 2026 Hype Cycle for Agentic AI parks the category at the Peak of Inflated Expectations. Only 17% of organizations have deployed agents, more than 60% say they will within two years, and Gartner’s read is that the foundation matters as much as the agent. 

Three months in, he hadn’t asked one vendor what their agent would know about his carriers before it touched a load.

What Are the Five Jobs Freight Teams Hire AI Agents to Do?

The five jobs are check calls, POD and document retrieval, appointment scheduling, exception management, and inbox triage. It’s the same list at every brokerage and 3PL I’ve walked through this year, because those are the five places a desk burns hours on work that doesn’t need a person.

They were his five rows too. We published a polite version of this list in July. This is the other one.

Job One: Check Calls and Status Cadence

A check call is a rep phoning a carrier to ask where the truck is, and it eats two hours of a rep’s day. His row was seven for seven, and he’d piloted a voice specialist he liked.

On the call itself, a voice agent beats any platform’s voice module, ours included, and HappyRobot’s $150M raise on August 4 at $1.2B says the market agrees.

Then we checked how many of his loads ran on carriers we had API connections to. Most of them. On an API network, a position lands every seven minutes, and a ping is a fact. A call gets a dispatcher once an hour, and a guess. His agent was dialing because dialing was all it had.

Ask what yours knows before it picks up the phone.

Job Two: POD and Document Retrieval

A POD is the signed proof the freight arrived, and you can’t invoice without it, so billing spends its day chasing carriers for paperwork by email and portal. This was the first row I told him a specialist might beat us on. Shortlist two.

The payback is cash; it hits DSO inside a billing cycle, and the good ones read a 2 a.m. truck stop BOL photo better than most humans. The catch is that retrieval is half the job, since the document still has to match a shipment and write back.

His team had hundreds of classified PDFs for loads no one could match. That’s why we built LTL PRO Resolution. A shipment with no valid PRO is a ghost, so the agent gets the number from the carrier and writes it back. Ask what theirs does when the PRO is wrong.

Job Three: Appointment Scheduling and Rescheduling

Scheduling is booking the dock window, and rescheduling is moving it when the truck’s going to miss, and a desk does both by phone. I told him to buy dock scheduling software, and he already had it, which made me like him more. Second row I’d hand to somebody else.

So why was the row still there? Detention. ATRI’s survey of 587 drivers and 245 carriers found drivers held at 39.3% of stops in 2023, up to 209 hours a year each, $11.5B in lost productivity and $3.6B in direct cost.

Rescheduling is an ETA problem before it’s a calendar problem. His dock software owned the slot, with no idea the truck was 90 minutes down outside Effingham. Something has to see the slip and move the slot first. That’s Dispatch & Appointment. Ask what triggers a reschedule, and how early.

Job Four: Exception Management and ETA Validation

An exception is any load that’s off plan, late, stuck, missing a status, and someone has to notice, figure out what’s true, and fix it. He got annoyed here because he’d watched four exception demos that all looked great.

Every exception demo is the same: 90 seconds, red row, click, green row, rehearsed so you don’t ask where the red row came from. Ask anyway. Detection quality is the data underneath, and a tool on your TMS’s feed inherits every gap your TMS has, so it finds what you knew and misses what costs money.

Then there’s the fix, which is a network action. Re-tendering means knowing who’s next best on that lane this week and how to reach them, which ARC Advisory calls tracking to intervention and we call the Intelligence layer. Make every vendor show you an exception your TMS never flagged and what they did. Two of his four couldn’t.

Job Five: Inbox Triage and Inbound Routing

Inbox triage is working the shared ops mailbox, where carrier updates and customer “where’s my freight” emails pile up, and somebody reads each one to figure out which load it’s about. I told him this was the easiest row and to take the win, because sorting email is close to what these models do naturally. What worried me was the write-back.

I made him read CargoNet’s Q2 report from August 6. Theft incidents were down 26% year over year. Yet, losses more than doubled to $304.6M, and the schemes that held steady were business email compromise and shipment misdirection. The physical stuff got harder, and the email stuff kept working.

He was about to let software answer carrier emails without asking what it checks before it hits send. Ask what the agent verifies about a sender before it acts, and at what dollar amount a person has to look first.

Five More Questions Before Putting Pen to Paper 

By then his five rows had answers, and the spreadsheet had run out of columns, which was the point. The questions that decide whether any of this works don’t fit in a grid, and none of his seven vendors had volunteered them.  

Should You Buy a Point Tool and Go Home?

Halfway through, he asked when he should just buy a point tool and be done. Fair question, and there are three cases where the answer is yes.

  1. You’re a Shipper: LSP44 is built for brokers, 3PLs, and forwarders moving someone else’s freight. If you’re buying for your own, you want project44, and I’ll walk you over myself.
  2. One Desk Is Underwater, and the Rest of the Operation Is Fine: Buy the specialist for that desk. Infrastructure bought in a panic gets ripped out 18 months later by whoever inherited it.
  3. Your Dock Calendar Is the Constraint: Covered above, and repeated because it’s the miss I see most.

He didn’t fit any of the three. He had three of five rows on fire at once, and that’s where I stopped being polite.

What Did the Supreme Court Do to Your Audit Log?

On May 14, the Supreme Court ruled 9-0 in Montgomery v. Caribe Transport II that negligent-hiring claims against brokers survive FAAAA preemption, and a decade-old shield was gone. Four days later, the 4th Circuit vacated Echo Global’s summary judgment.

He hadn’t connected that to the spreadsheet. Any agent that picks a carrier or re-tenders a load leaves a discoverable record of a safety decision, and every late-night re-tender is a document a plaintiff’s attorney will eventually read to a jury. If your vendor’s audit log is a CSV export, congratulations, you’ve prebuilt the exhibit. Ask if you’ll still have that log in two years.

Which Autonomy Tier Are You Actually Buying?

He asked this one himself, so the ruling had landed. You set the autonomy by customer, by lane, and by dollar threshold, and a vendor who hands you an on/off switch has never sold to a burned broker. 

Four tiers, and make every vendor name which one you’re buying:

  1. Observe: Watches and reports; touches nothing.
  2. Recommend: Proposes and waits for you.
  3. Act With Approval: Executes after a human signs off, inside your thresholds.
  4. Act Autonomously: Executes inside your guardrails, with every action logged.

This matters more than any feature list. Gartner expects over 40% of agentic AI projects will be canceled by late 2027 over cost, value, or controls, and MIT’s NANDA group found 95% of 300 enterprise AI projects had no measurable P&L impact. These projects die in procurement, and they die over control.

What Does Buying Five Tools Really Cost?

We crunched numbers on the call. Five jobs bought separately are five contracts, five integrations, five write-back paths, five audit trails, and five vendors who all have to survive until 2029. Not to mention, the license fees were the smallest number on that list and the only one in his business case. The real cost is the evidence.

ATRI found 94.5% of fleets charge detention and collect on fewer than half the invoices: the work got done and the paperwork couldn’t prove it. Every vendor on his grid priced the task and left the evidence as somebody else’s problem. Ask each one what happens to the record when their tool is wrong, and count the shrugs.

What Happens When You Add the Second Agent?

None of his seven vendors wanted this question, because it separates a pilot from a purchase. Anyone can get one agent live on one workflow. With five-point tools, the second agent is a second integration project and a second vendor to keep alive. With one data graph, it’s a configuration, and the difference compounds with every agent you add.

As of our quarter ending July 31, most customers run three or more agents; the next one adds no integration work, and new ARR grew 76% quarter over quarter with net retention up 393 basis points. 

I’d rather you check it than take my word. Ask every vendor what the second agent costs to stand up, and count the statements of work.

Buy the Layer Underneath the Agent

He went back to his spreadsheet and replaced every green checkmark with the answer to the question for that row. Most turned yellow, a few turned red, and it was the first honest version of the grid he’d seen in three months.

So score the jobs, then buy the data layer. The specialists will keep getting better at single jobs. What compounds is the context underneath, and there’s exactly one place to get 11 years of it.

LSP44 runs on 280,000+ carriers and 706 million carrier events a day, roughly $1B in the making. Nine of the 10 largest logistics providers on the Armstrong & Associates Top 50 run on it; we’ve been profitable since day one, and most of the companies on your grid can’t say either. Your TMS manages the record. LSP44 executes the work.

Sometime in 2028, an attorney is going to ask your software to explain a carrier selection. Buy for that conversation.

If you’ve got a spreadsheet like his, book a 30-minute mission briefing and bring it. We’ll go row by row, and I’ll tell you to your face which rows we lose.