AI-Driven Dispatch for 3PLs: When Two Clients Want the Last Vehicle

AI-driven dispatch allocates orders to vehicles against live constraints and learns from what happened. For a shipper running its own freight, the objective is singular: serve the network at the lowest cost consistent with the service promise. For a third-party logistics provider it is not, because the loads competing for a vehicle belong to different clients with different contracts, different penalty regimes and different relationship value.

Most of the time this is invisible. Capacity is adequate, everything moves, and the allocation engine looks like an optimizer. The problem appears only under scarcity, when two clients’ loads want the same vehicle in the same window and one of them will not be served. At that moment the dispatch decision stops being operational and becomes commercial, and an engine minimizing cost will confidently pick the wrong client.

Why 3PL dispatch is an arbitration problem

Capacity scarcity is not hypothetical in current conditions. ATRI’s operational cost analysis put the industry-average cost to operate a truck at $2.336 per mile in 2025, up 3.4%, with truckload and refrigerated operating margins below 1.0%, carriers cutting truck counts by 2.4% and leaving 10% of trucks unseated. A market where fleets are shrinking and a tenth of the equipment has no driver is a market where contention is routine rather than exceptional.

The client base is also fragmenting, which multiplies the number of contracts a dispatcher is arbitrating between. AlixPartners’ annual home delivery survey, covering 100 North American transportation and supply chain executives at companies with sales above $100 million, found more than 90% now running a mix of last-mile carriers and 32% using four or more. From the provider’s side of that relationship, every shipper running a mix is a shipper whose volume can move, which raises the cost of serving them badly.

The cost of getting it wrong is poorly measured. McKinsey surveyed 35 senior leaders at 28 North American consumer packaged goods companies and found only 17% believe they recover more than 75% of the true cost to serve. For a 3PL the equivalent problem is sharper still, because the cost of a service failure to one client is a penalty on that client’s account rather than a line in the operation’s cost base.

And the efficiency logic that normally guides allocation pulls the wrong way here. McKinsey’s work on out-of-home delivery finds that raising drops per stop from one to five cuts labor and vehicle cost by more than 50%. Density rewards co-routing clients together, which is correct and which is also what creates the contention: the same consolidation that lowers cost puts two clients on one asset.

Also Read: Top 8 AI-Driven Dispatch Companies for Enterprise Logistics

How AI-driven dispatch should arbitrate between clients

1. Work the simplest case, because it inverts

One vehicle, two loads, one will not move today.

Load	Revenue	SLA penalty if missed
Client A	$400	$200
Client B	$300	$900

A revenue-ranking engine serves A, the higher-value load. The outcome is $400 earned and B’s $900 penalty incurred, netting minus $500. Serving B instead earns $300 and incurs A’s $200 penalty, netting plus $100.

The higher-revenue choice is $600 worse. Nothing about the operation differs between the two plans. The only difference is which contract absorbed the failure, and revenue ranking is blind to that entirely.

2. Rank on value per unit of capacity, not on revenue

The general rule follows directly. The value of serving a load is the revenue it earns plus the penalty it avoids, and the relevant denominator is the capacity it consumes.

Load	Revenue	Penalty avoided	Capacity used	Value	Value per unit of capacity
A	$400	$200	1.0	$600	600
B	$300	$900	1.0	$1,200	1,200
C	$800	$150	2.0	$950	475
D	$250	$600	0.5	$850	1,700

Priority by value per unit of capacity runs D, B, A, C. Priority by revenue alone runs C, A, B, D. The two orders are almost exact reverses of each other, and only the first minimizes total loss. A dispatcher ranking by revenue will serve the largest load first and finish the day having paid the most penalties.

3. Accept that penalty is not the only term

Penalty is the part that can be computed. Several terms cannot, and pretending otherwise produces a confident engine making commercially naive decisions.

Volume commitments mean a client whose contracted minimum is at risk carries value beyond today’s load. A client in renewal negotiations carries value that appears in no contract clause. A client with an exclusivity arrangement may be more expensive to disappoint than the penalty implies. These require a client priority tier set commercially and held as a constraint, not inferred from load attributes.

4. Size the exposure before deciding how much machinery it deserves

Contention frequency multiplied by the cost of a wrong choice gives the number that justifies the work.

Contention frequency	$300 wrong-choice cost	$700	$1,500
5% of days	$3,750 per depot per year	$8,750	$18,750
10% of days	$7,500	$17,500	$37,500
20% of days	$15,000	$35,000	$75,000
30% of days	$22,500	$52,500	$112,500

At 20% of days and $700 a time, a twenty-depot network is carrying around $700,000 a year in avoidable penalty. The two inputs are both measurable: count the days a load was deferred for capacity, and price the difference between the choice made and the best available alternative.

Also Read: How AI Dispatch Allocation Works for Freight Carriers

5. Close the attribution gap, or none of this gets fixed

This is why the problem survives. The dispatcher defers a load on Tuesday. The penalty is calculated at month end, appears on a client account statement, and is reviewed by an account manager who was not involved in the decision and cannot see which load was served instead. The operation therefore never learns which arbitration calls were wrong.

The fix is a record rather than an algorithm: log the deferral, the alternative served, and the penalty subsequently incurred, against the same reference. That produces a dataset the arbitration rule can actually be tuned on, and without it any AI-driven dispatch engine is learning from decisions whose outcomes it never observes.

6. Distinguish the scarce resource from the scarce moment

Contention is usually framed as a shortage of vehicles. More often it is a shortage of vehicles inside a window: two clients both need a morning delivery and the same asset cannot do both. That reframing opens options an asset-count view misses, including offering one client a later slot, splitting the load, or sourcing a single movement to a carrier rather than re-planning the day.

There is a paradox worth naming here, because it changes how the trade-off should be read. Density rewards co-routing clients onto shared assets, and co-routing is what puts two clients in contention for one vehicle in the first place. The efficiency gained from consolidation is therefore partly repaid in arbitration cost, and a provider that never co-routes never faces this problem while also never earning the density margin. The right conclusion is not to separate clients but to price the arbitration properly, because the consolidation gain is almost always larger than the penalty exposure it creates. It only stops being larger when the arbitration is decided badly, which is the case this piece is about.

7. Let the engine act on the routine cases and escalate the material ones

Most contention is small. A deferral worth $50 of penalty differential should be resolved automatically against the stated priority. A deferral that risks a client’s contracted volume commitment or lands during a renewal should reach a person who owns that relationship. Setting that threshold in currency rather than in principle is what keeps the escalation volume workable.

Where 3PL dispatch differs from shipper dispatch

Dimension	Shipper dispatch	3PL dispatch
Objective	Single, cost against one service promise	Multiple, one per client contract
What a deferral costs	Internal service failure	Penalty on a named client account
Who absorbs the failure	The same P&L that made the choice	A different client relationship
Correct ranking	Cost or service urgency	Revenue plus penalty avoided, per unit of capacity
Non-computable terms	Few	Volume commitments, renewals, exclusivity
Failure visibility	Same team, same week	Different team, following month
Required input	Constraint set	Constraint set plus contract hierarchy

The last row is the buildable conclusion. A 3PL running an allocation engine on the shipper configuration is running an engine with a missing input, and the missing input is the one that decides who gets served when the vehicle runs out.

Also Read: Top 10 Dispatch Platforms for Scheduling and Tracking

Five criteria for evaluating AI-driven dispatch for a 3PL

1. Can it hold penalty terms per client contract as a dispatch input? Not as a note in the account file. As a value the allocation logic reads when it decides which load moves.

2. Does it rank on value per unit of capacity? Ask specifically whether the priority calculation includes penalty avoided and divides by capacity consumed, because revenue ranking is the default and it is wrong.

3. Can a commercial client priority tier override the computed ranking? Volume commitments and renewals are real and uncomputable. The engine needs to accept an instruction it cannot derive.

4. Does it log the deferral alongside the alternative served? Without that pairing the operation cannot learn which arbitration decisions were wrong, and the engine cannot be tuned.

5. Is tenant separation genuine at the decision layer? Clients should be co-routed for density and reported separately for cost and service, which means one plan and many books rather than many plans.

Also Read: 10 Best AI Dispatch Software Platforms (2026 Comparison)

What this looks like in enterprise deployments

A Fortune 50 parcel and freight enterprise moving more than a million freight shipments a year across a 120-country network centralized dispatch across 51 sites and a 4,500-strong pool split between captive and third-party capacity, lifting weekly execution rate from 75% to 92% and surfacing more than $14 million in unused capacity. The finding that matters here is the $14 million: much of what looks like scarcity is capacity that exists somewhere the deciding plan cannot see. Before arbitrating between two clients, it is worth confirming the vehicle is genuinely unavailable rather than invisible.

A leading paint manufacturer processing more than 1,500 carrier invoices a month across 160 depots used Locus Settlement, Carrier and Orchestrator agents to automate freight reconciliation, catching 5% to 6% variance above contracted rates and compressing payment cycles from 30 to 45 days down to 7 to 10. That is the same matching discipline this problem needs, pointed at carrier invoices. Applied to client penalties it becomes the attribution record that makes arbitration decisions reviewable.

Four mistakes in 3PL dispatch under scarcity

Ranking competing loads by revenue. It is the intuitive choice and it can be several hundred dollars worse per event, because it ignores the penalty attached to the load not served.

Treating penalty as the whole commercial picture. Volume commitments, renewals and exclusivity are real value that no load attribute reveals, and they need a priority tier set by the people who own the contracts.

Leaving the deferral unrecorded. If the operation does not log which load was deferred and what was served instead, it cannot know which calls were wrong and no engine can learn from them.

Assuming contention means the fleet is full. It usually means a window is full. Later slots, split loads and single-movement carrier sourcing are options that an asset-count view never surfaces.

Also Read: What is Dispatch Management? Key Components, Best Practices, Challenges

How Locus approaches multi-client dispatch arbitration

Locus, the world’s first Decision-Intelligent, Agentic TMS, is built for operations where one plan serves many commercial relationships, which is the structural condition of a 3PL. Dispatch planning runs across more than 250 real-world operating constraints, and Transporter Management holds contract lifecycle, rule-based order allocation on cost, speed, zones, contract and performance, and invoice reconciliation. That last capability is the one that carries client-specific terms into the allocation decision rather than leaving them in an account file.

Because the route planning system produces dispatch-ready plans in roughly two minutes and re-optimizes continuously, a contention event can be answered by re-planning rather than by choosing between two fixed options. That frequently dissolves the arbitration entirely: the second load moves later on the same asset instead of being deferred to the next day.

Allocation also runs across owned fleet, contracted transporters and a network of more than 1,000 carriers, decided per shipment against live cost and serviceability. For a 3PL facing genuine scarcity, sourcing the marginal movement is usually cheaper than paying a penalty, and the arbitration only becomes necessary when neither owned nor sourced capacity is available.

The governance layer sets which decisions a person sees. Autonomy Levels run per agent and per domain, so a low-value deferral can be resolved autonomously against the stated priority while one touching a volume commitment reaches an account owner. Explainability and Traceability record the trigger, context, reasoning, action and outcome for each decision, which is precisely the attribution record the arbitration problem lacks: a penalty arriving at month end can be traced to the decision that caused it, the alternative that was served, and the rule that ranked them.

One boundary is worth stating. Locus does not hold your commercial client hierarchy by default, and it should not invent one. Penalty terms, volume commitments and relationship priority come from the contracts and from the commercial team, and mapping them into dispatch priority is a named implementation requirement rather than an assumption. An arbitration engine is only as good as the contract data behind it.

Locus supports more than 360 enterprise customers across 30-plus countries, with over 1.5 billion deliveries optimized, more than $320 million in documented client logistics savings and 99.99% uptime. It has been recognized by Gartner for seven consecutive years, featured in the 2026 Hype Cycle for Supply Chain Execution and Logistics Technologies, named a Leader in TMS by QKS Group (SPARK Matrix), and ranked #1 in Route Planning on G2’s 2026 Best Software Awards.

In October 2025, Ingka Investments, the investment arm of Ingka Group, the world’s largest IKEA retailer, acquired Locus. Locus continues to operate independently.

So how should AI-driven dispatch decide between two clients wanting the same vehicle? On value per unit of capacity, where value is revenue plus the penalty avoided, and with a commercial priority tier layered on top for the terms no load attribute reveals. The arithmetic is unforgiving: in a two-load case, serving the higher-revenue client nets minus $500 while serving the lower-revenue one nets plus $100, and across a four-load queue the correct priority order is close to the exact reverse of a revenue ranking. At contention on 20% of days and $700 a wrong choice, a twenty-depot network carries around $700,000 a year of avoidable penalty that nobody attributes, because the penalty arrives a month later in a different report. Locus carries client contract terms into allocation through Transporter Management, re-optimizes in roughly two minutes so most contention dissolves into a later slot rather than a deferral, sources the marginal movement across a 1,000-plus carrier network before arbitration is needed, and records the decision and its alternative so the ranking can be reviewed against what the penalty actually cost. Request a Locus dispatch assessment to price your own contention events.

FAQs

Why is 3PL dispatch different from shipper dispatch? Because the objective is plural. A shipper optimizes one network against one service promise. A 3PL allocates one asset base across clients with different contracts, penalty regimes and relationship value, so under scarcity the choice of which load to serve is commercial rather than operational, and no single cost objective captures it.

Should competing loads be ranked by revenue? No, and this is the common default. In a worked two-load case, serving the $400 load with a $200 penalty exposes the operation to the other client’s $900 penalty and nets minus $500, while serving the $300 load nets plus $100. Ranking by revenue alone ignores the cost of the failure it creates.

What is the correct priority rule? Revenue plus penalty avoided, divided by the capacity the load consumes. On a four-load example that ordering is close to the exact reverse of a revenue ranking, because a large load consuming double capacity with a small penalty is worth less per vehicle-slot than a small load with a large penalty.

How do you handle client value that is not in the contract? With a commercial priority tier held as a constraint. Volume commitments at risk, renewal timing and exclusivity arrangements are real value that no load attribute reveals, so the engine needs to accept a priority instruction from the commercial team rather than deriving one.

How much is this worth fixing? Multiply contention frequency by the cost of a wrong choice. Contention on 20% of days at $700 a time is roughly $35,000 per depot per year, or $700,000 across twenty depots. Both inputs are measurable: count capacity-driven deferrals and price the gap between the choice made and the best alternative.

Why does the problem persist in most operations? Attribution. The deferral happens on a Tuesday and the penalty appears at month end on a client account reviewed by someone who was not part of the decision and cannot see what was served instead. Without logging the deferral alongside the alternative, the operation never learns which arbitration calls were wrong, and an AI-driven engine cannot be tuned on outcomes it never observes.