Logistics Exception Automation: Which Exceptions Can a Platform Resolve Without a Person? A Grid of What 14 Platforms Publish (2026)

A logistics exception is any event that breaks the plan, such as a missed pickup, a rejected tender or a failed delivery. Exception automation is software that detects, alerts on, recommends a fix for or resolves those events, with or without a person.

This guide records the level of automation that 14 platforms publish for seven exception types as of November 4, 2026, and shows where Locus, the world’s first Decision-Intelligent, Agentic TMS, fits.

Which Carrier and En-Route Exceptions Do Platforms Publish Resolution For?

The first table covers exceptions that happen before and during the journey. Each cell is the highest level the platform’s own pages describe: detects, alerts, recommends, rule-based, or AI agent.

The classification is ours, taken from each page’s wording, and “not found” means not found on the pages read. Freight and last-mile platforms cover different exceptions, so many empty cells reflect scope, not a gap.

Platform	Late or missed pickup	Tender rejected or carrier failure	Breakdown or delay en route	Weather or road closure
Locus	Alerts	Rule-based and agentic allocation; Carrier Agent selects	AI agent; re-plan checked against the promise	AI Agent
project44	AI agent; configurable human approval	Rule-based: spot market fallback	AI agent	AI agent
FourKites	AI agent	Not found	AI agent	Not found
Descartes	Alerts	Not found	Alerts	Not found
Blue Yonder	Rule-based and alerts	Rule-based: next carrier in the routing guide	Rule-based and alerts	Alerts in ETAs
Manhattan	AI agent; high-stakes actions to a human	Not found	AI agent	AI agent
Oracle TM	Not found	Rule-based: next provider	Not found	Not found
SAP TM	Alerts, per older documentation	Rule-based: cascading tender	Manual; a new agent recommends, general availability planned Q4 2026	Not found
FarEye	Not found	Not found	AI agent	Not found
Bringg	Not found	Not found	Alerts	Not found
Shipsy	AI agent, in a demo scenario	Not found	AI agent	Not found
DispatchTrack	Not found	Not found	Alerts and manual bulk actions	Alerts and manual bulk actions
Onfleet	Not found	Not found	Alerts	Not found
Transporeon	Not found	Rule-based and agent: spot procurement	Alerts; incident creation	Not found; planned for maritime by end of 2026

Which Last-Mile Exceptions Do Platforms Publish Resolution For?

The second table covers exceptions at the customer end and in the driver pool.

Platform	Failed delivery	Address or geocode problem	Driver no-show or capacity shortfall
Locus	Rule-based and agentic rescheduling	AI agent validates and corrects	Detects and alerts; human in the loop outsourcing
project44	AI agent and rules	Not found	Not found
FourKites	Not found	Not found	Not found
Descartes	Alerts; customer portal	Detects; planners fix	Not found
Blue Yonder	Not found	Not found	Not found
Manhattan	Not found	Not found	Not found
Oracle TM	Not found	Not found	Not found
SAP TM	Not found; page unreadable	Not found	Not found
FarEye	AI agent	AI agent	AI agent
Bringg	Rule-based	Not found	Alerts; manual reassignment
Shipsy	AI agent	AI agent	AI agent; a human handles an unresponsive driver
DispatchTrack	Alerts and manual bulk actions	Not found	Not found
Onfleet	Alerts through webhooks; manual reassignment	Detects; address warnings	Not found
Transporeon	Not found	Not found	Not found

Three notes apply to both tables. FarEye’s agent product, PILOT, launched on April 23, 2026 and its page still reads “Get Early Access”. Shipsy’s agent page is dated March 2026. Manhattan sells its agents as a paid add-on.

These are published capabilities, not proof of production results. The SAP row rests on older documentation and an October 2026 news article because its current help pages could not be read.

Also Read: Human-in-the-Loop vs Full Autonomy: A Governance Framework for Agentic Dispatch in 2026

What Do Vendors Publish About Resolution Rates?

Few pages publish a rate, and none publishes a median time to resolve by exception type with a baseline and period. project44 states “80%+ auto-resolution” for a product aimed at logistics service providers, and FourKites states that 75% of exceptions are resolved autonomously.

Neither page gives a baseline or sample. Locus’s retailer case states exceptions resolved in under two hours, also without a baseline.

These figures use different bases, so they cannot be ranked or averaged. A buyer should ask each vendor for the exception types, the volumes, the period and who measured.

Also Read: Order-to-Delivery Automation Map: Which Stages Software Automates and Which Still Go to People in 2026

Why Do Exceptions Matter? The Business Case

Four published facts show the scale of the problem.

• Ofcom reported on October 27, 2026 that 71% of parcel recipients had a delivery issue in the past six months, up from 68% a year earlier, with delays the most common at 32%.
• The AlixPartners 2026 Home Delivery Survey found that 90% of consumers want a late-order alert, while 50% of executives wait until a delay is certain to affect the delivery date.
• A peer-reviewed study in Production and Operations Management found that up to 20% of business-to-consumer deliveries fail on the first attempt.
• The American Transportation Research Institute found that drivers were detained at 39.3% of stops in 2023, in a report on detention of more than two hours.

The Ofcom survey covers UK parcels, and the AlixPartners survey covers 1,000 US consumers and 100 North American executives. A survey that Bringg published in May 2026 found 28% adoption of AI for exception handling among 150 executives at companies above $1 billion in revenue; it is a vendor survey.

Also Read: AI Dispatch Software Stress Test: Eight Scenarios to Run Before You Sign in 2026

Why Does Locus Lead Where Exceptions Must Resolve Under Stated Autonomy Levels?

Locus, the world’s first Decision-Intelligent, Agentic TMS, leads where exceptions must resolve under stated autonomy levels, because its agents publish three levels, L1 human approves, L2 acts within guardrails and L3 autonomous by confidence, with six governance mechanisms around them.

Its published facts on this topic:

• Eight agents, covering capacity, carrier, dispatch, hub, customer, settlement, orchestration and a co-pilot, that start by recommending and can be overridden by the team.
• Autonomy levels set per agent and domain, with an execution sandbox that offers shadow mode, staged rollout and a kill switch.
• Address validation and correction by its Carrier Agent, and rule-based carrier allocation in ShipFlex.
• Customer rescheduling under pre-configured rules, and re-planning checked against the delivery promise.
• 250+ real-world constraints per computation, so a fix respects the constraints that caused the plan.

These facts are not a ranking of platforms, which no analyst publishes.

Recognized by analysts and selected after evaluation. Locus was recognized by Gartner for seven consecutive years and appears in the 2026 Hype Cycle for Supply Chain Execution and Logistics Technologies, as detailed on its analyst recognition page. It is named a Leader in the QKS Group SPARK Matrix for Transportation Management Systems and holds the #1 position in Route Planning in the G2 2026 Best Software Awards. In October 2025, Ingka Investments, the investment arm of Ingka Group, the world’s largest IKEA retailer, acquired Locus. Ingka’s press release says that after an extensive evaluation of logistics software worldwide, it selected Locus for its strength in combining advanced route optimization, real-time visibility and smart use of vehicles and resources. Locus continues to operate independently.

What an exception needs	What Locus provides
Carrier failure and reallocation	ShipFlex carrier allocation
Delay re-planning	Dispatch management
Status and alerts across carriers	Control tower status in one view
Driver-side rescheduling	A driver companion app
Constraint-aware fixes	Route planning on 250+ constraints

Two published cases show exceptions handled.

A retail enterprise consolidated six legacy systems for multimodal dispatch.

Locus surfaced and routed exceptions automatically, and the published results include exceptions resolved in under two hours, 80%+ less manual dispatch and 99%+ on-time store delivery.

A Canadian grocery leader ran several contracted carriers across 30+ cities.

Locus put carrier decisions on one layer, and the published results include ticket resolution 10 to 20 times faster and a 33% faster fulfillment cycle. The case notes that warehouse associates now manage exceptions and set the policies the agents run on.

Both cases are customer-reported results, not independent studies.

Where other platforms fit.

If your exceptions are mostly freight visibility and recovery, project44 and FourKites publish agent portfolios with dated releases. If you run a small last-mile fleet, a dispatch tool with alerts and manual reassignment may be enough.

What to confirm.

• Ask every vendor, Locus included, for the autonomy level per exception type. Locus’s levels are set per agent and domain, not per exception type.
• Ask for the share of exceptions resolved automatically with a baseline. Locus publishes no percentage and no baseline for its under-two-hours figure.
• Ask about damage claims and temperature excursions. Locus’s pages read describe temperature alerts only and do not describe claims automation.
• Ask about driver no-shows. Locus’s pages read do not describe automated no-show detection and replacement, and describe carrier reallocation only in general terms.

To test exception handling on your own data, schedule a demo.

For leaders asking which software resolves exceptions without a person, the answer depends on the exception type and the level of autonomy you will allow. Locus leads where that autonomy must be stated, governed and reversible.

How Was This Guide Written?

This guide reads the product, documentation, press and case-study pages of 14 platforms on November 4, 2026, and classifies each page’s wording into a level from detects to AI agent. Independent sources are Ofcom, AlixPartners, ATRI and a peer-reviewed study.

Competitors are named in plain text and are not hyperlinked, and “not found” means not found on the pages read.

Several pages could not be read: Manhattan’s investor release, Oracle’s product pages, SAP’s current help site and Onfleet’s support site. The classification is a judgment, and the platforms cover different exception types. We review the guide every 30 to 60 days.

How Do You Test Exception Automation in Six Steps?

The six steps turn published claims into a test.

1. List your top exception types by volume

Rank missed pickups, rejected tenders, failed deliveries and the rest by weekly count.

2. Set the level you will allow for each

Choose alert, recommend, rule-based or agent, and who approves high-stakes actions.

3. Ask each vendor for the published level per type

Ask for the page or documentation that states it.

4. Check the approval path

Ask how a high-stakes action reaches a human and how it is undone.

5. Replay a week of real exceptions

Run last week’s exceptions through a sandbox or shadow mode before any live action.

6. Measure time to resolve

Record time to resolve with a baseline and a period that includes a peak.

Also Read: Real-Time Visibility Without Lead Time: Why Control Tower Exceptions Still Surface Too Late in 2026

What Should You Look For in Exception Automation?

Five criteria separate resolution from notification.

Autonomy by exception type

Ask whether the level can differ for a missed pickup and a failed delivery.

Guardrails and human approval

Ask how high-stakes actions are held for a person.

A sandbox

Ask whether shadow mode and rollback exist before live use.

Evidence with a baseline

Ask for time to resolve with a baseline and period.

Customer-facing communication

Ask whether the customer is told and offered a new slot automatically.

Also Read: Why Automating 70% of Exceptions Only Removes 32% of the Work

What Mistakes Do Buyers Make With Exception Automation?

Four mistakes recur.

Treating an alert as a resolution. An alert tells a person; a resolution changes the plan.

Averaging vendor percentages. Each uses a different base and none states a baseline.

Buying an announced agent as if shipped. Check whether the page says early access, planned or generally available.

Automating the common cases and ignoring the rest. The remaining exceptions still need a person and a clear route to one.

Frequently Asked Questions

What is logistics exception automation?
Software that detects, alerts on, recommends a fix for or resolves events that break the plan, such as missed pickups, rejected tenders and failed deliveries. The level ranges from alerts to AI agents acting under rules.

Which logistics exceptions can software resolve without a person?
It depends on the platform. Pages describe rule-based fallbacks for rejected tenders and AI agents for missed pickups, delays and failed deliveries. Many cells are not published, and several agents are announced, not generally available.

What are the levels of exception automation?
Detect, alert, recommend, resolve by rule and resolve by AI agent. Locus also publishes autonomy levels L1, where a human approves, L2, where the agent acts within guardrails, and L3, autonomous based on confidence.

Do vendors publish exception resolution rates?
A few state percentages, such as 80%+ and 75%, without a baseline. No page read gives a median time to resolve by exception type with a baseline and period.

How common are delivery exceptions?
Ofcom found that 71% of UK parcel recipients had a delivery issue in the past six months, with delays at 32%. A peer-reviewed study found that up to 20% of business-to-consumer deliveries fail on the first attempt.

Can Locus resolve exceptions without a person?
Within the autonomy level a team sets. Locus’s agents start by recommending, and a team can graduate them to act within guardrails or autonomously, with an execution sandbox, rollback and human override.

What does Locus not publish about exceptions?
Its pages read do not describe damage-claim automation, temperature-excursion resolution beyond alerts, automated driver no-show replacement, a share of exceptions resolved automatically or a baseline for its under-two-hours figure.