Descartes Bought Tai. Your Only Lever Is 60 Days.

If you run Tai TMS, the leverage you assumed you had over this acquisition does not exist, and the data clause you would have negotiated was signed years ago. Descartes Systems Group announced on August 24 that it acquired Tai, the AI-powered transportation management system for freight brokers, for approximately US$100 million satisfied from cash on hand. The deal is done — the 6-K describes a completed transaction, not a pending one, with no regulatory condition to wait out.

What makes this worth twenty minutes of a broker's Monday is not the price. It is that Descartes named the data in its own press release. Andrew Wimer, Associate General Manager for Transportation Management, said the acquisition "adds valuable transaction, carrier and shipment execution data to the Descartes Global Logistics Network." That is your quoting history, your carrier list and your lane margins, described as an asset of the buyer. Read your own agreement before you read the roadmap.

---

What $100 Million Bought, and Why the Press Release Named Your Data

Descartes bought a full brokerage operating system, and the operating system is where the commercially sensitive data lives. Tai's platform "unifies quoting, carrier sourcing, load execution, billing, and customer engagement" across truckload, LTL, drayage and cross-border, per Descartes' release. Tai's own FAQ is more specific about what sits in the database: shipment records and history, customer and carrier databases, invoices, payments, margins, lane pricing history and activity logs.

Lane pricing history and margins are the two fields a competitor would pay for. In April 2026 Tai also launched an autonomous AI voice agent — built on CloneOps.ai — that phones drivers for check calls and writes ETAs, delivery outcomes and call summaries straight into the load record. So the dataset now includes machine-generated observations about carrier reliability that no broker used to capture systematically.

The Global Logistics Network is where that lands. Descartes describes the GLN as "the world's largest connected community" of over 30,000 connected customers managing billions of transactions a year, and says plainly that it uses "trusted, clean, structured, real-time data to fuel AI-driven innovation." This is not speculation about intent. It is the stated product strategy, and the Tai release is the acquisition rationale in the same words.

Tai's Terms Already Authorized This. Here Is the Clause.

You do not have a change-of-control termination right, because Tai's published terms grant the assignment in advance and grant it one way only. The terms and conditions state that Tai "may assign or transfer it without your consent to (a) an affiliate, (b) a company through a sale of assets by Tai or (c) a successor by merger." Customers, by contrast, cannot assign without written approval. Both routes an acquirer might take are pre-authorized; you get neither notice nor a veto.

The data permission is in the same document, and it is broader than most brokers realize. The clause reads: "You give Tai permission to combine non-identifiable information you enter or upload for the Services with that of other users of the Services and/or other Tai services." Note the tail — other Tai services. After an acquisition, the perimeter of "Tai services" is defined by the entity that now owns Tai, and nobody negotiated where that line sits.

Then the commercial mechanics, which are the part you can actually act on. The agreement is a yearly subscription that will "auto-renew in perpetuity upon initial expiration date." You may cancel only by notifying Tai "at least sixty (60) days prior to the beginning of each Renewal Term." And "pricing adjustments of up to 5% will be applied to all agreements unless otherwise specified," annually. Tai markets itself as 12-month terms without multi-year lock-in, which is true and is also exactly why the 60-day window matters: it is the only recurring moment when you hold anything.

Two Licenses, Two Very Different Scopes

Tai's terms grant two separate licenses over your data, and only one of them is broad. This distinction is the whole negotiation.

For your identifiable content, the grant is narrow and purpose-limited: "You grant Tai a worldwide, royalty-free, non-exclusive license to host and use any Content provided through your use of the Services for the sole purpose of providing the Service." Ownership stays with you — the FAQ says "you own your data 100%" and that you can export at any time.

For non-identifiable, aggregated data, the grant has no purpose limitation at all. It permits combination with other users' data and with other Tai services, full stop.

So the honest reading is this: pouring your identifiable load records into a shared network product is not obviously covered by the license you signed. Pooling de-identified aggregates across Tai's own users is covered, plainly; whether "other Tai services" now reaches Descartes products is the part nobody has answered. Which means the fight is not over ownership — everybody concedes ownership — but over the definition of "non-identifiable." A lane, an origin-destination pair, a commodity and a date can re-identify a shipper with very few contributors on that lane. If your MSA does not define de-identification with a minimum aggregation threshold, you have granted something you have not measured.

The Network Already Pools Broker Data. MyCarrierPortal Is the Proof.

This is not a hypothetical risk, because Descartes already ships a product built on exactly this mechanism and markets it as a feature. Descartes MyCarrierPortal, its carrier onboarding and fraud-prevention product, runs a community incident database: when one customer submits a report on a carrier, "it becomes part of an accessible knowledge base" available to others, and users "can also see how many others have 'blocked' carriers." Descartes reports that between February 2024 and January 2025, over 50% of subscribing companies submitted a total of 4,468 incident reports, against a base of over 350,000 carriers with certificates of insurance on file.

It goes further than reports. In April 2025 MyCarrierPortal added tracking history so a broker can see, before onboarding, the total number of loads a carrier tracked over the past six months and the percentage successfully tracked via MacroPoint. That signal only exists because other brokers' loads generated it.

And the AI layer is already built on pooled execution data. Descartes' Fleet Data Intelligence platform, launched April 14 2026, applies machine learning to "the trusted execution data flowing through the GLN" — real-world delivery durations, route conditions, vehicle type, geography — and Descartes claims early deployments improved route density by up to 30%. That is a vendor claim, not an audited result. But the architecture is unambiguous: execution data from many customers, one model, better product.

None of that is sinister. It is the correct engineering answer, and brokers benefit from it. The question is whether the click-through you signed with a California software company is the contract you want governing a network operator with 30,000 connected customers.

The Steel Man: Your TMS Has Been Feeding a Rate Index for Years

Before you escalate this, know that the freight industry settled this argument a decade ago and settled it in favour of pooling. DAT's rate benchmarks — the numbers your team quotes against every day — are built from contributed transaction data, and DAT is explicit about the pipe: "Rate data is uploaded automatically, every day, directly from the contributors' transportation management systems." Hundreds of brokers, carriers and shippers contribute. DAT masks contributor identity, and each published lane rate "includes a minimum of three contributors, and no individual company dominates a lane rate."

That last sentence is the model to argue for. DAT did not win trust by refusing to pool; it won trust by publishing the aggregation floor and the anonymization method. Descartes has published neither for the GLN.

There is also a genuine argument that Descartes is the better custodian. Its FY2026 revenue was $729.0 million with $329.5 million in adjusted EBITDA and $356.5 million in cash at year end. A broker choosing between a venture-funded point solution and a profitable public company with a formal DPA program is not obviously choosing wrong by taking the public company.

Descartes Now Owns Two Freight Broker TMS Products

The roadmap question is real, and the historical answer is reassuring but not thrilling. Descartes has owned Descartes Aljex, a freight broker TMS with lane pricing, capacity sourcing, shipment visibility and accounting, since it paid $32.4 million for Aljex in February 2018, when Aljex had 400 customers and roughly three million freight moves a year. Aljex is still actively marketed and now advertises "500+ freight brokerages."

So the good news: Descartes does not buy TMS products in order to kill them. Eight years on, Aljex has its own brand, its own site and its own release cadence. The less good news: 400 to 500-plus customers in eight years is maintenance, not investment, and two overlapping broker TMS products plus 3GTMS, bought for about $112.7 million in March 2025, is three roadmaps competing for one engineering budget. FreightWaves counts 34 Descartes deals since 2017, including Drivin for $30 million last month and Idelic for $28 million in April. This is a serial acquirer with a portfolio, and portfolio companies get triaged.

What Descartes' Own Terms Say — and Don't

Here is the part that should change your negotiating posture: Descartes' published contract language on data is narrower than Tai's, and the gap is where you should aim. The GLN Service Terms grant Descartes and its affiliates "a worldwide, limited-term license to host, copy, transmit, modify, encrypt, and display Customer Data, but only to the extent that it is reasonably necessary for Descartes to provide the Services." The Master Terms state that "Descartes does not claim any title or ownership rights over any Customer Data." Neither published document contains an aggregation-and-benchmarking clause of the kind Tai's terms carry.

But the same Master Terms introduce a second category — Application Data — and say "as between Customer and Descartes, Descartes retains all rights to Application Data," while prohibiting customers from scraping, caching or creating derivative works from it. The definitions sit in a section the public page truncates. Get the definition of Application Data in writing, in your order form, before you sign anything on Descartes paper. The boundary between "Customer Data you own" and "Application Data we own" is the entire commercial question in a network business, and it is currently invisible to you.

One more thing the terms do not give you: an exit. Descartes' legacy Master Terms bind the agreement to "their respective successors and permitted assigns" with no change-of-control trigger, and the current Master Terms let Descartes terminate on 120 days' notice if it discontinues a service for all customers. The convenience termination rights run one direction here too.

What to Do

This Week:

1. Pull your executed Tai order form and any addendum, not the website terms. Website terms are the default; a negotiated MSA may say something different, and you need to know which document governs before you argue about it.
2. Find your renewal date and count back 60 days. Put that date in the calendar as a hard decision point, not a reminder. Under the published terms it is the only cancellation mechanism you have, and it repeats annually.
3. Write down which fields you would object to seeing in a pooled benchmark: buy rate, sell rate, margin by lane, named shipper, named carrier. You cannot negotiate a data clause you have not enumerated.

Before Your Renewal Date:

4. Send Descartes a written request for its aggregation floor and de-identification method — the DAT standard of a minimum contributor count per lane is the ask. If the answer is a marketing sentence rather than a number, that is your answer.
5. Ask, in writing, whether "other Tai services" in the aggregation clause now includes Descartes products, and get the scope narrowed to the Tai product line if it does. This costs nothing to ask before renewal and is unwinnable after migration.
6. Request the definitions of Customer Data and Application Data from the Master Terms and have counsel map your fields onto both. Do this before any migration onto Descartes paper, because migration is when the old contract quietly stops governing.
7. Export a full copy of your shipment history, carrier list and lane pricing while the export tooling is still Tai's. The FAQ promises export "at any time in multiple formats"; verify that against a real export, not a screenshot.

This Quarter:

8. Price a fallback. You do not have to switch, but a broker with no second quote has no position at the 5% annual uplift, let alone at a post-acquisition repricing.
9. Add a mutual change-of-control clause to your template for every operational SaaS renewal this year — a termination right on 60 days' notice, exercisable for 90 days after announcement. Tai's terms are not unusual; they are standard, which is the problem.

The Bottom Line

Every generation of logistics software has ended the same way: the execution layer gets bought by the network layer, because the network is worth more than the software and the software is how you get the data. Descartes said so out loud, in the second paragraph of its own press release, and it was not being sneaky — it was telling shareholders what it bought. The failure here is not the acquirer's. It is that brokers signed a yearly click-through that pre-authorized assignment to a successor by merger and pre-authorized aggregation with other users, then assumed a deal announcement would hand them a seat at the table.

The deal closed on Monday. Your seat at the table is 60 days before your renewal date, and it is the only one you get.

Continue Reading

• Stripe Bought OpenRouter. A Toggle Is Not a Contract.
• DoiT Bought Attribute. It Also Sends Your Cloud Bill.
• Visa Bought BioCatch. Get Neutrality in Writing.
• Your AI Vendor Joined OpenAI in March. You Heard in August.
• Nvidia Hired 109 Poolside Engineers. No Clause Fired.
• Dynatrace Bought Arize. Fix Your Spans Before It Closes.
• Oakley Bought Graphwise. GraphDB Isn't Open Source.
• Bending Spoons Bought Airtable. Reprice Before It Closes.