The Freight Intelligence Bet: What Carrier AI Routing Means for Every Logistics Director Locked Into a Contract That Assumed Human Decisions
UPS ORION processes routing decisions for 21+ million packages per day and has avoided 100 million+ miles annually since full deployment. FedEx committed more than $2 billion to its DRIVE AI transformation, redesigning delivery density and dynamic pricing across its entire network. Maersk AI manages vessel routing, predictive port congestion, and dynamic ocean freight pricing across 380+ ports and 750+ vessels. Your enterprise shipper contract — governing rates, data rights, routing accountability, and dispute resolution — was almost certainly written before any of these systems were in production. The governance gap is not a technology problem. It is a contracting problem that your procurement team has not yet surfaced, because the AI transition happened inside the carrier's operations without requiring a contract amendment.
Key Numbers
Background
Enterprise logistics has been algorithmically managed for longer than most sectors acknowledge. UPS began developing ORION in 2008 and completed full US deployment in 2019 — a six-year, $250 million program that replaced human route planning with machine learning across its entire US ground network. By the time most enterprise shipper agreements were last renegotiated, the routing decisions those agreements assumed were made by UPS dispatchers and drivers were already being made by an optimization engine processing 250 data points per stop. The agreements did not change. The routing did.
FedEx's DRIVE transformation accelerated the same structural shift. Announced in 2022 and implemented through 2024–2025, DRIVE collapsed FedEx Express and FedEx Ground into a unified network — a reorganization that is operationally impossible without AI. Human planners cannot dynamically allocate 15 million daily packages across a consolidated multimodal network in real time. DRIVE uses AI to determine which packages move by air versus ground, which routes are consolidated, and how capacity is dynamically priced within contracted rate structures. The "fixed rates" in most shipper master agreements have AI-driven surcharges and capacity allocation decisions layered on top of them that were not part of the rate card when the contract was signed.
Ocean freight went through the same transition more visibly, because spot rate volatility during 2020–2023 made algorithmic pricing legible to shippers who had never thought about how rates were set. Maersk, CMA CGM, MSC, and Hapag-Lloyd all deployed AI pricing engines during this period — systems that set spot rates based on real-time supply-demand signals, predictive port congestion, and shipper-specific booking history. The dynamics that shippers experienced as "extreme market volatility" were in part the output of competing algorithmic pricing systems, each trained on data from thousands of enterprise shippers and optimizing for carrier network economics rather than shipper cost predictability. Long-term contracts negotiated after 2022 include more market-indexed terms specifically because shippers discovered that fixed-rate contracts were being undercut by carrier AI allocating capacity to spot buyers when spot rates exceeded contract rates.
The third-party logistics market amplified the transition. DHL Supply Chain, XPO Logistics, C.H. Robinson, and Coyote (owned by UPS) all manage enterprise shipper volume through AI optimization systems that treat the shipper's freight as an input to a network-wide optimization problem. The 3PL's AI is optimizing for the 3PL's network economics — trailer utilization, lane balance, cost per mile — not for the individual shipper's cost or transit time. When a C.H. Robinson algorithm routes your LTL shipment through a relay point that adds 18 hours of transit time, the routing decision optimizes the carrier's trailer utilization on two lanes simultaneously. Your contract gives you a transit time commitment. It does not give you an explanation of the routing logic, a mechanism to challenge the algorithmic decision, or data rights to the transit performance history the carrier is using to train its next pricing model.
The visibility gap is being filled by a category of AI platforms that sit between shippers and carrier AI systems. Project44 serves 1,300+ enterprise shippers across 200,000+ carriers, providing real-time shipment tracking that does not depend on carrier-reported data. FourKites covers 500M+ shipments annually. Descartes Systems Group's routing intelligence platform aggregates carrier data for enterprise transportation management systems. These platforms exist because the data that carrier AI systems have about enterprise shippers' freight — predictive ETAs, port congestion signals, weather rerouting decisions — is not exposed through standard shipper portals. Enterprise logistics directors are buying a second layer of AI visibility specifically because the first layer, which they are contractually paying for through carrier rates, does not disclose enough about how routing decisions are being made.
The data rights dimension has gone almost entirely unaddressed in enterprise shipper contracting. When UPS, FedEx, or Maersk AI systems process your enterprise's shipping data — origin-destination pairs, volume patterns, seasonal peaks, commodity types, consignee characteristics — that data becomes training input for the carrier's pricing and routing models. The carrier's AI learns the structure of your supply chain and uses that learning to optimize the carrier's network economics, including the pricing of your contract at renewal. Most enterprise shipper agreements include standard data provisions written before machine learning was a material factor in carrier operations. Those provisions do not address whether your shipping data can be used to train the model that prices your next contract negotiation against you.
Decision Required
Your enterprise carrier agreements are due for renewal. Your logistics team is preparing the RFP. The rates, SLAs, and dispute resolution terms were negotiated before your carriers deployed AI routing and dynamic pricing at scale. What contractual terms will you require in the next agreement that you did not require in the last one?
The practical governance questions that most enterprise supply chain and procurement leaders have not answered: Can you get a routing explanation — not just a transit update, but an explanation of why a specific routing decision was made — from your carrier within 48 hours of a service failure? Does your carrier agreement define how your shipping data can be used after the contract terminates? When your carrier's AI-driven capacity allocation changes the effective service level on a contractually committed lane, what is your recourse mechanism and is it defined in writing? If your 3PL's AI optimization platform is acquired, do your data rights transfer with the acquisition?
These are not hypothetical future concerns. They are governance gaps in contracts that are active today, governing freight that moved yesterday, in relationships that your procurement team will renegotiate in the next 12–24 months without having added a single AI-specific provision. The question is not whether to address them — carrier AI is not going back to human routing — but whether to address them in the next contract cycle or discover them through a service failure, a rate dispute, or a data breach incident where your carrier discloses that your logistics data was part of a compromised training dataset.
Options
This is the path most enterprise logistics teams are currently on. Current agreements remain in place; upcoming renewals are negotiated on rate, capacity, and SLA terms that do not address algorithmic routing accountability, data use, or dynamic pricing methodology. The risk is not speculative: rate disputes are already occurring where carriers apply AI-driven surcharges to contracted lanes without disclosing the calculation methodology, and enterprise shippers are discovering at renewal that their multi-year shipping data history has improved the carrier's pricing accuracy against them. Choosing this path is a decision to defer governance negotiation until a specific incident creates enough organizational urgency to address it retroactively — typically at higher legal and commercial cost than proactive renegotiation.
This is the highest-leverage action available in the current carrier contracting cycle. Requiring algorithmic transparency means carriers must disclose, on request within defined SLA windows, the basis for routing decisions that result in service failures — not the underlying model weights, but the operational inputs and decision logic that produced the outcome. Data governance clauses define whether and how your shipping data can be used for carrier AI model training, whether that use requires your consent, and whether data must be purged or de-identified at contract termination. Pricing methodology disclosure defines what inputs drive dynamic capacity surcharges and what notice window you receive before they apply. Most enterprise carriers will negotiate these terms if asked; the market gap is that most shippers do not ask. Carriers that refuse to negotiate any transparency terms are disclosing that their AI optimization is structured in ways that cannot withstand shipper scrutiny — which is itself a material procurement signal.
A carrier-agnostic visibility platform gives you independent shipment data that does not depend on carrier-reported information, a unified performance dataset that spans your carrier mix, and the ability to calculate carrier AI routing performance against your contracted SLAs without relying on the carrier's own measurement. The 1,300+ enterprise shippers on project44 are buying this independence specifically because carrier AI systems surface different data in their shipper portals than they hold internally. The platform cost — typically $200K–$500K annually for mid-market enterprises — is recoverable through improved SLA enforcement and carrier performance management. The strategic benefit is that your shipping data is now governed by your agreement with the visibility platform, not by your agreement with each carrier, and the visibility platform's business model does not depend on pricing freight against you.
A multi-carrier strategy with AI-driven volume allocation limits the data advantage any single carrier builds by seeing your complete origin-destination patterns. When UPS sees 100% of your freight, its AI builds a complete model of your supply chain seasonality, consignee density, and volume predictability — and uses it in renewal pricing. When UPS sees 60% and FedEx sees 40%, neither carrier has the complete picture, and your TMS AI can optimize allocation based on lane performance, capacity availability, and dynamic pricing signals from both. Transportation management system platforms — Oracle TMS, SAP Transportation Management, MercuryGate — now include AI allocation optimization modules specifically for this use case. The operational complexity of multi-carrier management is real but well-established; the strategic benefit is structural: no single carrier's AI can model your complete supply chain from a single-carrier data advantage.
Recommendation
Add algorithmic accountability terms to the next carrier RFP before your procurement team issues it. The specific provisions that matter: a routing decision explanation right (carrier must explain, on request within 5 business days, the operational basis for any routing decision that resulted in a documented SLA miss); a data use definition (your shipping data may be used to operate the current contract but may not be used to train models used in pricing negotiations at renewal without your written consent); a dynamic pricing disclosure standard (AI-driven surcharges applied to contracted lanes must be disclosed with inputs and calculation methodology, not just posted as line items); and a post-contract data governance provision (shipping data associated with your account must be de-identified or deleted within 90 days of contract termination on request). These are negotiating positions, not non-negotiable demands — but they must be in the RFP to be in the contract.
Build carrier-independent logistics data before your next renewal negotiation. Deploy a visibility platform — project44, FourKites, or Descartes — at least 12 months before major carrier renewals. The dataset you build through an independent platform gives you carrier performance analytics that are not filtered by the carrier's own reporting system, and gives your procurement team objective SLA performance data as leverage in the renewal conversation. Carriers with strong performance records will welcome the independent data; carriers whose AI routing has been generating systematic SLA misses will resist it. The resistance is the signal.
Audit your current 3PL agreements for two specific provisions that most contracts are missing: a definition of what constitutes "your" data in the context of the carrier or 3PL's AI systems, and a specification of what happens to that data if the 3PL is acquired. The M&A risk in logistics technology is not theoretical — XPO spun off GXO Logistics and RXO, UPS acquired Coyote and then explored divestitures, CH Robinson faces ongoing activist pressure. When a 3PL with your complete supply chain data is acquired, the acquiring entity inherits the data relationships unless your contract specifies otherwise. Most do not.
Evaluate the multi-carrier strategy as a structural hedge, not just a backup. The data advantage that a single carrier builds from processing 100% of your freight is compounding: each year of single-carrier consolidation improves the carrier's pricing model accuracy against you and increases the switching cost of moving volume. If your current carrier concentration is above 70% of freight spend with one carrier, model what a 60/40 or 50/30/20 multi-carrier split would cost operationally — TMS complexity, rate card management, consolidated billing — against the strategic benefit of competitive tension in renewals and reduced AI pricing model accuracy against you. For enterprises with more than $50M in annual freight spend, the structural hedge is typically worth the operational cost.
Enjoying this brief? The next one ships Tuesday.
One enterprise AI deployment, dissected weekly. Free during beta · No credit card · Unsubscribe anytime
Risks
This is the data rights risk that most enterprise procurement teams have not modeled. When UPS ORION processes four years of your freight — specific origin-destination pairs, volume by day of week and season, consignee concentration by geography, commodity mix, time-sensitivity distribution — it builds a model of your supply chain that is more granular than any analyst your logistics team has ever produced. At contract renewal, the carrier pricing team has access to that model. They know your peak volume windows, your highest-cost lanes, your consignee density in markets where they have pricing power, and the historical elasticity of your freight spend. Your procurement team is negotiating without equivalent insight into the carrier's cost structure. The data asymmetry is a direct product of your single-carrier consolidation and the absence of data use restrictions in your current agreement.
ORION, DRIVE, and their equivalents make routing decisions that human dispatchers would not make — decisions that optimize the carrier's network economics and occasionally produce individual shipment outcomes that are worse than what a human dispatcher would have chosen. When an AI routing decision results in a missed delivery window — a time-sensitive pharmaceutical shipment routed through a consolidation hub that added a day of transit; a just-in-time manufacturing component delayed because the AI allocated the truck to a higher-density lane — the current contractual mechanism for most enterprise shippers is a service failure claim against an SLA that was written for human-routing expectations. The claim process typically results in a rate credit. It does not produce a routing explanation, a commitment to algorithm adjustment, or an accountability mechanism for recurring systematic errors in specific lanes. Most enterprise shipper agreements have no provision for challenging algorithmic routing decisions specifically, because those agreements predate algorithmic routing becoming the default.
The gap between the rate card in a master shipper agreement and the rate actually charged has always included accessorial charges, fuel surcharges, and dimensional weight adjustments. AI adds a new layer: dynamic capacity surcharges that reflect the carrier's real-time AI assessment of supply and demand on a specific lane, applied within the headroom that the master agreement's cap structures allow. Most enterprise shippers discover these surcharges on invoices without receiving advance disclosure of the inputs, calculation methodology, or the AI trigger conditions. The audit burden shifts to the shipper's accounts payable team, which is not positioned to challenge the algorithmic basis of a line item it cannot see. Freight audit and payment firms estimate that AI-driven billing discrepancies that are technically within contracted terms but outside shipper expectations account for 2–4% of enterprise freight spend — money that the shipper is paying and cannot easily recover because the charges are not clearly out of contract.
The logistics technology sector is consolidating. XPO divested GXO and RXO; UPS has evaluated Coyote divestiture; private equity owns significant 3PL positions across mid-market logistics providers. When a 3PL with four years of your complete supply chain data is acquired, the buyer acquires the data relationship as an asset of the business — unless your contract explicitly specifies that data rights do not transfer without your consent. Most do not. The practical consequence: your supply chain data, including the AI-derived models the 3PL built from it, is now owned by an entity you did not select, whose competitive interests may be different from the entity you contracted with, and whose data governance practices you have not evaluated. The risk is not just privacy or confidentiality — it is that a competitor, a carrier, or a company with adverse interests in your supply chain may now hold the most granular model of your logistics operations that exists.
The EU AI Act's provisions on high-risk AI systems and general-purpose AI entered effect in phases through 2025–2026. Logistics AI systems that make consequential decisions affecting enterprise businesses — dynamic pricing, capacity allocation, routing that determines whether goods arrive on time — are attracting regulatory attention as the Act's enforcement mechanisms develop. If your enterprise operates in the EU and your carrier's AI routing or pricing decisions affect your EU operations, the documentation requirements that apply to the AI system may create obligations for both the carrier (as the deployer) and your enterprise (as the affected business). Most enterprise carrier agreements do not include EU AI Act compliance representations. The gap between the Act's documentation requirements and the documentation that carrier AI systems were designed to provide is significant — carrier routing models were not built with regulatory explainability as a design requirement.
Questions Your Team Should Be Answering
These are the questions that distinguish organizations that get this right from those that do not. If your team cannot answer them, that is your first deliverable.
- 1.
If UPS ORION or FedEx DRIVE makes a routing decision on your freight that results in a documented SLA miss, can you get a written explanation of the routing logic within five business days — and is that right defined anywhere in your current carrier agreement?
- 2.
Does your current master shipper agreement with your primary carrier specify whether your origin-destination data, volume history, and consignee information can be used to train the AI model that will price your contract at the next renewal negotiation?
- 3.
When your carrier applies a dynamic capacity surcharge to a contracted lane, what disclosure do you receive about the inputs and calculation methodology — and does your agreement specify a notice window before the surcharge takes effect?
- 4.
Have you modeled the switching cost of your current primary carrier relationship after three or more years of AI platform integration, and does that switching cost match what you calculated when the relationship was established?
- 5.
If your primary 3PL were acquired tomorrow, would your data rights — including the right to receive a copy of your historical shipping data and a commitment to data deletion at contract termination — transfer under the terms of your current agreement, or would the new owner inherit them without restriction?
- 6.
What percentage of your enterprise freight spend is with a single carrier, and have you modeled the pricing model advantage that data concentration gives that carrier at your next renewal versus what a 60/40 split across two carriers would cost you operationally?
If this memo belongs in your next executive meeting or board pack, send it along. One click opens a pre-drafted email — edit or send as-is.
The ATO Bottleneck: What Federal Agencies Discover When AI Procurement Meets the Authorization Process
Federal agencies are deploying AI tools across procurement, benefits processing, and workforce operations — but the ATO process was written for static systems. FedRAMP authorizes cloud infrastructure, not AI behavior. Most frontier AI APIs lack FedRAMP authorization, and most federal ATOs are stale by the time the model updates.
Read memo →The Algorithmic Underwriting Audit: What NAIC AI Requirements Mean for Every Insurer Using AI in Pricing and Claims
State insurance regulators have moved. The NAIC Model Bulletin on AI has been adopted in 38+ states. Colorado mandates external algorithmic audits for life insurance AI. California CDI has challenged AI-generated property risk scores. Most carriers have deployed AI in claims and underwriting without building the governance documentation regulators are now requiring.
Read memo →The SR 11-7 Blind Spot: What Banks Discover When AI Hits Model Risk Management
Banks are deploying AI in credit underwriting, fraud detection, compliance monitoring, and customer service — but SR 11-7, the OCC/Fed model risk framework, was written in 2011 for statistical models. The validation gap for third-party LLM APIs, the model version change management problem, and what bank examiners are beginning to ask.
Read memo →