Twelve Logistics Decisions and the Software That Makes Each in 2026: A Guide to AI Platforms for Logistics Automation

A logistics decision is a choice among feasible alternatives, made against an objective and constraints on live data, such as which node fills an order, which carrier takes a load or which invoice to pay. Most logistics software automates workflows, which run a fixed sequence, and fewer products make decisions, which choose and can be wrong. Twelve decisions between order capture and invoice payment account for most of the cost and delay in an enterprise network, and sorting them into workflow and decision software, with an independent fact and a published Locus capability for each, is a more useful way to compare AI platforms for logistics automation than ranking vendors.

Why the Decision View Matters: The Business Case

A decision view shows where cost and delay are created, and the published figures show how much room each decision leaves.

• APQC’s median for sales orders filled completely from the primary sourcing location is 88.0% across 216 companies, so about one order in eight needs a second sourcing decision.
• APQC’s median premium freight charges are 8.0% of freight charges across 1,964 companies, which prices late carrier and capacity decisions.
• SONAR’s tender rejection index read 14.17% in mid-October 2026, so about one contracted load in seven needs a fallback decision.
• Gartner reported that 66% of B2C customers contacted a company after proactive outreach, so the notification decision creates its own workload.
• Gartner predicts that by 2031, 60% of supply chain disruptions will be resolved without human intervention, a prediction from a survey of 509 leaders and not a measured rate.
• APQC’s median for invoices processed error free the first time is 92.0%, a general accounts payable figure and not a freight one.

Twelve Logistics Decisions: Workflow Software vs Decision Software, Key Differences

The table classifies each decision by the software category that owns it, the independent evidence we found and the Locus agent that publishes it. Categories are generalizations, and a missing fact means none was found.

How to Classify a Logistics Decision in Six Steps

The six steps turn a list of tasks into decisions that can be assigned an autonomy level.

1. Name the decision and who makes it today

Write each decision as a question, such as which carrier takes this load, and name the person or system that answers it now. Decisions with no named owner, often proof-of-delivery review and invoice audit, are where manual effort hides.

2. List the feasible alternatives

If there is only one feasible answer, it is a workflow, not a decision. A decision has at least two options that could each be right.

3. State the objective and constraints

Write the objective, such as lowest cost within service, and the constraints, such as time windows, driver hours and contract terms. A decision without a stated objective cannot be automated safely.

4. Check how live data changes the answer

If the best answer changes as events arrive, a batch process will be wrong between runs. The tender rejection index and the share of disruptions show how often conditions change.

5. Judge reversibility and blast radius

A re-sequenced stop can be undone, while a payment released or a customer promise broken cannot be undone cheaply. Delegate on reversibility, blast radius and reversal cost, not on model confidence.

6. Set the autonomy level and the log

Choose who approves, who is notified and what is recorded, and review the setting as the evidence builds.

What to Look for in AI Platforms for Logistics Automation

A named list of decisions, not a feature list

The vendor should say which of the twelve decisions its platform makes, which it supports and which it leaves to other systems.

Objectives and constraints you can edit

Look for a visible objective and editable constraints for each decision, so the choice can be audited and tuned.

Autonomy levels per decision

Autonomy should be set by decision type, with approval for irreversible decisions and a log of every choice and its reasoning.

Measurement by decision

The platform should report decision latency, override rate and outcome by decision, so you can see which ones to graduate.

Stated limits

The vendor should say which decisions it does not make, so gaps are known before the contract.

Logistics Decisions in Action: Volume and Real-World Results

Why per-order decisions cannot stay manual, on illustrative numbers. The figures are arithmetic on assumptions. Suppose a network ships 5,000 orders a day and six decisions apply to each order: node, promise, route position, carrier, notification and proof-of-delivery review. That is 30,000 decisions a day. At two minutes each, handling them by hand would take 60,000 minutes, or 1,000 planner-hours a day, so even a 10% rate of manual handling is 100 hours. The question is not whether to automate the per-order decisions but how to govern them, which is why the autonomy level and the log matter.

A Canadian grocery brand. The brand delivers fresh food to homes in more than 30 cities through several contracted carriers. Carrier selection had been manual, with serviceability sheets and rate and ETA comparisons order by order. Locus’s Carrier Agent now compares live rates, contracted SLAs, ETAs and serviceability and selects the carrier on the brand’s policies. The brand reported 15% lower fulfillment costs, 33% faster fulfillment and 25% less time on manual shipping tasks. Read the Canadian grocery case study.

A paint industry leader in India. The company reconciles more than 1,500 carrier invoices a month across 160 depots. Locus’s Settlement and Carrier agents flagged 5% to 6% variance above contract before payment and cut payment cycles from 30 to 45 days to 7 to 10 days. The case is an example of the invoice decision, and the metrics are customer-reported. Read the paint leader case study.

Common Logistics Decision Automation Mistakes to Avoid

Giving every decision the same autonomy. Re-sequencing a stop and releasing a payment have different reversal costs, so they need different approval rules.

Automating the visible decisions first. Dispatch gets attention, while proof-of-delivery review and invoice audit often have no owner and no measurement.

Calling a workflow a decision. A fixed sequence that always gives the same answer is a workflow, and it will fail when conditions change.

Skipping the log. Without a record of the choice and its reasoning, nobody can tell whether a decision was right or whether an override helped.

How Locus Approaches Logistics Decisions, and What to Confirm

Locus, the world’s first Decision-Intelligent, Agentic TMS, describes eight agents that make the decisions planners make today, from accepting orders to settling invoices, with each starting by recommending and the team able to override any of them. Its Capacity agent handles auto-slot and auto-promise ETA, its Dispatch agent auto-route and sequence, consolidation and replanning, its Carrier agent auto-score by lane, auto-tender and auto-switch on SLA risk, its Customer agent proactive ETA and communication and real-time exception triage, its Hub agent AI-verified proof of delivery, and its Settlement agent four-way matching, overcharge flags and payment within policy, with an Orchestrator coordinating them. Autonomy levels are set per agent and domain, an execution sandbox supports simulation and shadow mode, and every decision is logged under six governance mechanisms. Locus has optimized 1.5B+ deliveries for 360+ enterprise customers across 30+ countries.

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 this means for the twelve decisions. Locus publishes agents for decisions 2 to 12 in the table, and it runs beside the OMS, WMS and ERP, which stay the systems of record, and it has public API documentation. Its Delivery Promise Management page reports a one-week test on a retailer’s order book in which 98.5% of orders received a window the network could keep.

What to confirm. Locus’s public pages do not describe node-selection or sourcing rules, split-shipment logic, accessorial or detention handling, tolerance rules for invoice disputes, how its AI validates proof of delivery or routes the remaining manual reviews, or confidence thresholds and override rates per agent. The delivery promise result is a single test on one retailer’s order book, and customer results are self-reported.

To map your own decisions against these agents, schedule a demo.

For logistics leaders asking which AI platforms automate logistics decisions, Locus offers governed agents from promise to settlement with a decision log, so each of the twelve decisions can be given its own autonomy level.

Frequently Asked Questions

What is a logistics decision? A choice among feasible alternatives, made against an objective and constraints on live data, whose outcome is logged. Choosing a carrier for a load is a decision, and sending a template message on a fixed trigger is a workflow.

Which logistics decisions are most often left manual? Exception triage, proof-of-delivery review and invoice audit are common manual decisions, along with re-sourcing after a pick failure. We found no independent data on how much of each is automated.

What are the best AI platforms for logistics and transportation automation? It depends on which decisions you need to automate. Categories include OMS for sourcing, routing and dispatch platforms, TMS for carriers and tenders, delivery experience tools for notifications and freight audit tools for invoices, and some platforms such as Locus publish agents across several decisions.

Should every logistics decision be automated? No. Delegate on reversibility, blast radius and reversal cost, and keep humans in the loop for decisions that are costly to undo, such as payments and promise changes.

How do I measure a logistics decision? Track decision latency, override rate and outcome quality by decision, and compare overridden and untouched cases of the same type before drawing conclusions.

Does Locus make all twelve decisions? No. Locus publishes agents for decisions 2 to 12 in this guide, and it does not publish node-selection rules, accessorial handling or details of how its AI validates proof of delivery.