Which AI logistics tools do transportation work, and which only report it?
Short answer
Some AI logistics tools only report on freight work, and others do the work. Reporting tools tell your team what is happening. Agents send the request, book the carrier, chase the reply and file the document. Judge a tool by tasks completed on real loads, not by insights. Wilson, by Cartage, an AI logistics agent, does these tasks itself, from quote to payment.
Book 15 minutes with the team →What is the difference between doing freight work and reporting it?
Doing freight work means completing the next action on a load, such as sending a tender, getting a carrier's acceptance, or collecting a proof of delivery. Reporting it means showing the team that the action is needed. An AI logistics agent should complete routine actions, while a reporting tool leaves them to people.
Two kinds of output
| Reporting output | Execution output |
|---|---|
| "Load 4410 has no carrier yet" | Quote requests sent, quotes received, carrier booked |
| "Carrier has not replied in 3 hours" | Follow-up sent, carrier called, reply recorded |
| "Load is running late" | Carrier asked for new ETA, customer updated |
| "POD missing for 6 loads" | POD requested, received, checked, and filed |
| "Invoice exceeds quote by $220" | Line compared with terms, carrier asked for proof |
ETA means estimated time of arrival. A proof of delivery, or POD, is the document the receiver signs to show the freight arrived.
The workload test
A simple test shows which kind of tool a team has: if the tool were turned off tomorrow, would the team's daily task list get longer? If not, the tool was reporting. If the list would grow by dozens of follow-ups, bookings, and documents, the tool was doing the work.
Why the difference matters
A reporting tool can make a team better informed while leaving its workload the same. If a dashboard shows 30 loads that need follow-up, someone still has to send 30 follow-ups. An executing tool reduces the workload, because it sends them.
Alerts create work
Every alert is a task for someone. A tool that sends more alerts may look more powerful but can make the team busier. The question is not how many problems a tool finds, but how many it resolves.
Reporting tools still have a place
Dashboards, alerts, and analytics help managers see patterns, plan capacity, and spot problems. They are not a lesser category; they solve a different problem. The mistake is buying a reporting tool to fix an execution problem, or the reverse.
Execution versus planning
Some tools plan freight: they choose modes, build loads, or plan routes. Planning is useful, but a plan still needs someone to carry it out with carriers. Buyers should separate planning tools from execution tools, and ask which one their current pain needs.
The "AI" label hides the difference
Many products now describe themselves as AI. Some use AI to summarize emails or predict delays. Others use it to carry out tasks. The label alone does not say which. Buyers need to ask, for each task, who actually does it.
Categories of AI logistics tools
Most tools fall into a few categories. Visibility platforms track shipments and alert on delays. Analytics tools report on cost and performance. Email assistants summarize or draft messages. Transportation software suites plan and book within their own system. AI logistics agents carry out the daily work across carriers and channels. Each category does something useful; only the last is built to complete the freight tasks themselves.
Recommendations still need someone to act
A tool that recommends "book Carrier B" or "follow up with Carrier C" has done the thinking but not the doing. Someone must still send the message, wait for the reply, and record the result. When volume is high, recommendations pile up just like alerts.
Hypothetical example: a better dashboard, same workload
This example is hypothetical and does not describe a real customer. A distributor adds a tool that flags late loads and missing documents. The team now sees every problem clearly. But the coordinators still email carriers, update customers, and chase PODs by hand. After three months, the team knows more and works the same hours. The gap was never visibility; it was execution.
See it on your own loads. Wilson, by Cartage, is an AI logistics agent that quotes, books, tracks, audits and pays for freight with your approved carriers.15 minutes with the team behind Wilson. No prep needed.
Book 15 minutes →Which freight tasks separate an executing tool from a reporting one?
The tasks that separate an executing tool from a reporting one are the actions someone must perform on each load: requesting quotes, booking, confirming pickups, chasing carriers, updating customers, collecting documents, and auditing invoices. An AI logistics agent should perform these, while a reporting tool only flags them.
Task sort
| Task | Reporting tool usually | Executing tool should |
|---|---|---|
| Quote requests | Shows which loads need quotes | Sends requests to approved carriers |
| Follow-up | Flags carriers that have not replied | Sends reminders and calls where configured |
| Booking | Recommends a carrier | Books within approval limits |
| Pickup confirmation | Shows unconfirmed pickups | Asks the carrier and records the answer |
| Delay response | Alerts that a load is late | Gets a new ETA and updates the customer |
| Documents | Lists missing documents | Requests, checks, and files them |
| Invoice audit | Flags differences | Compares with terms and asks the carrier for proof |
| Payment | Shows invoices due | Pays within the rules |
The execution workflow
| Step | Trigger | What the AI logistics agent does | When a person steps in |
|---|---|---|---|
| Quote | Order ready | Requests quotes from approved carriers | If no carrier replies |
| Book | Best option within limits | Books and sends the rate confirmation | If above the approval limit |
| Confirm | Load booked | Confirms pickup details with the carrier | If the carrier cannot meet the window |
| Track | Load moving | Gets updates and asks carriers when data stops | If a delay breaks a key promise |
| Document | Pickup and delivery | Prepares and collects BOLs and PODs | If a document shows exceptions |
| Audit and pay | Invoice arrives | Compares with what was booked; pays within rules | If a charge does not match or is above the limit |
A bill of lading, or BOL, is the shipping document that lists the freight and serves as the carrier's receipt for it.
Customer-facing work counts too
Updating customers about shipments is part of freight work. A tool that detects a delay but leaves the customer update to a person has done half the job. Execution should include the message to the customer, sent through the channels the company configures.
Sort your own tasks
A simple exercise helps: list the team's daily freight tasks in two columns. In one column, tasks someone must do. In the other, things someone must know. Then ask each vendor which column its product handles. Most teams find their pain is in the "must do" column.
Who owns the task when the tool stops
Every task needs an owner. When a tool stops partway, the remaining work should go to a named person, not to a shared queue. Buyers should ask the vendor where each task goes when the tool cannot finish it, and how the person is told.
Partial execution
Some tools do part of a task. They may send a quote request but not follow up. They may book through an API but not handle carriers who use email. An API is a connection that lets two software systems exchange data directly. Partial execution still leaves work, and the buyer should know exactly where it stops.
Exceptions are where execution matters most
Routine loads that go perfectly need little work from anyone. The real test is the exception: a carrier that declines, a pickup that is missed, a document that is wrong. A reporting tool flags the exception. An executing tool starts the response, such as contacting the carrier and initiating rebooking within the rules.
Execution across channels
Many carriers still work by email and phone. A tool that executes only through connected systems will leave a large share of carriers to the team. Execution should cover the channels the company's carriers actually use.
Execution needs rules
A tool that executes must follow the company's rules: which carriers to use, how far to negotiate, when to escalate, and what it may pay. Without rules, execution is risky. With rules, it is predictable. Buyers should check that the rules are visible and easy to change.
Hypothetical example: where the tool stops
This example is hypothetical and does not describe a real customer. A manufacturer tests a tool that sends tenders automatically through carrier APIs. Six of its twelve carriers have no API. For those six, the tool creates a task for a coordinator. Half the tenders still need a person. The buyer realizes the tool executes for half the network and reports for the other half.
How can a buyer measure whether a tool actually did the work?
A buyer can measure whether a tool did the work by counting actions completed on real loads: messages sent, replies handled, bookings made, documents filed, and invoices audited, each with a timestamp. An AI logistics agent should leave a record of every action it took.
What to count
| Measure | What it shows |
|---|---|
| Quote requests sent by the tool | Outreach done without people |
| Follow-ups sent and replies recorded | Chasing done without people |
| Bookings made within limits | Decisions carried out |
| Customer updates sent | Communication done |
| Documents requested and filed | Paperwork done |
| Invoices audited with outcome | Finance work done |
| Tasks handed to people | What the tool did not do |
Compare with the team's own time
The clearest measure is the team's time. Before the tool, how many hours a day did coordinators spend on follow-ups, updates, and documents? After, how many? A tool that executes should free noticeable time within weeks on the lanes it handles.
Ask the coordinators
The people closest to the work know whether it changed. After a few weeks, ask coordinators which tasks they no longer do and which they still do. Their answers are often the clearest measure.
Count tasks handed back
The most revealing number is how many tasks the tool hands back to people. A tool that sends 100 alerts a day has handed 100 tasks to the team. A tool that completes 90 and hands back 10 has done the work. The handback count shows the real workload.
Include exceptions in the sample
The sample should include loads that went wrong, not only clean ones. A tool may handle a clean load well and still hand every exception back to people. Exceptions are where most of the team's time goes.
Use a sample of real loads
Pick 20 to 50 real loads from a normal week. For each, record what the tool did and what people did. Compare with how the same kind of loads were handled before. The difference is the work the tool took on.
Measure the same lanes before and after
The fairest comparison uses the same lanes, before and after the tool. Count coordinator hours, time from order to booked load, and the share of loads with complete documents. These show whether the work moved from people to the tool.
Look at time, not just counts
A follow-up sent after two hours is not the same as one sent after two days. The record should show when each action happened, so the buyer can see whether the tool acts fast enough to matter.
Quality matters as much as quantity
Counting actions is not enough if the actions are wrong. A tool that sends 100 follow-ups to the wrong contacts has done 100 useless things. The sample review should check that actions went to the right people, with the right details, at the right time.
Hypothetical example: counting a week
This example is hypothetical and does not describe a real customer. A distributor tracks 40 loads for a week. The AI logistics agent sent 148 quote requests and 61 follow-ups, booked 34 loads within limits, sent 52 customer updates, and filed 38 PODs. It handed 9 decisions to people: 6 above the approval limit and 3 with missing order data. The team can see exactly what changed.
What proof shows an AI tool executed the work instead of describing it?
Proof of execution is the artifact the action left behind: the sent tender, the carrier's written acceptance, the rate confirmation, the customer notification, the filed POD, and the audited invoice, each with a time and a rule. An AI logistics agent should show these artifacts for any load on request.
Artifacts of completed work
| Action | Artifact |
|---|---|
| Quote request | Sent message with time and recipients |
| Carrier acceptance | Carrier's written reply or system confirmation |
| Booking | Rate confirmation with agreed price and terms |
| Pickup confirmation | Carrier's confirmed window and details |
| Customer update | Sent message with time and recipient |
| POD collection | Filed POD linked to the load |
| Invoice audit | Comparison result and any carrier response |
A rate confirmation is the document that states the agreed price and terms for one load.
Proof should be readable
A person who did not watch the work should be able to open the record and understand what happened in a minute. Raw system logs are not enough; the record should read like a short history of the load.
Status is not proof
A status such as "booked" or "delivered" is a label. Proof is the document or message behind it. A tool that sets statuses without the artifacts behind them may be reporting what someone else did, or guessing. The buyer should always ask what supports each status.
Ask to see the artifacts
After a test load, ask the vendor to show every artifact with its timestamp and the rule that triggered it. If the vendor can show only a dashboard view, the tool may have reported rather than executed.
Keep artifacts after the load closes
Records should stay available after the load is paid.
Search and export
The team should be able to search the records by load, PO, carrier, or date, and export them when needed for audits or reviews. Records that cannot be found or exported lose much of their value.
Artifacts should be in one place
The artifacts should live with the shipment, not scattered across inboxes and portals. One record per load lets anyone see what happened without asking around.
Artifacts carriers recognize
Carriers should see normal business messages: a clear quote request, a tender, a rate confirmation, a follow-up. If the tool's messages confuse carriers, replies slow down. Asking a friendly carrier to review a few messages is a useful part of evaluation.
Rules should be visible too
For each action, the record should show which rule allowed it: "Booked within 5% approval limit" or "Escalated: above limit." This lets the team check that the tool followed the company's rules.
Proof helps later disputes
Artifacts are not only for evaluation. When a carrier disputes a charge or a customer disputes a delivery, the same artifacts settle the question. A tool that keeps them with every load saves time months later.
Hypothetical example: asking for proof
This example is hypothetical and does not describe a real customer. After a demo, a buyer asks to see the record for one test load. Vendor A shows a dashboard with status colors. Vendor B shows the sent quote requests, two follow-ups, the carrier's acceptance email, the rate confirmation, the customer update, and the POD, each with a timestamp. The buyer can see which tool did the work.
How should buyers test AI logistics tools in a demo?
Buyers should give each tool one real shipment and ask it to perform the next actions, not explain them: request quotes, book an approved carrier, confirm pickup, respond to a delay, and collect the POD. An AI logistics agent should complete each action and leave the artifact.
Demo test
| Action | Pass | Fail |
|---|---|---|
| Request quotes | Requests sent to approved carriers | Recommendation only |
| Follow up | Reminder sent when a carrier is silent | Alert only |
| Book | Booked within the limit, rate confirmation sent | Carrier suggested |
| Confirm pickup | Carrier's confirmed window recorded | Pickup shown as "scheduled" |
| Respond to delay | New ETA obtained, customer updated | Delay flagged |
| Collect POD | POD requested, checked, filed | Missing POD listed |
Include an email-only carrier
At least one carrier in the test should work only by email or phone. This shows whether the tool executes across the real carrier network or only through connected systems.
Include a missing document
One load should have a missing document, such as a bill of lading without a weight or a POD that never arrives. The tool should ask for it, follow up, and escalate before it matters. This shows whether the tool closes gaps or only lists them.
Include a decision above the limit
One load should require a decision above the approval limit. The tool should stop and send it to a person with the options and deadline. This shows that execution respects the company's controls.
Use real loads, not a script
A scripted demo shows what the vendor wants to show. A real load from last week, with its messy carrier replies and late documents, shows how the tool behaves in daily use. Buyers should bring their own loads.
Questions to ask
- For each task, who performs it: the tool or a person?
- Which carriers can the tool work with, and through which channels?
- What happens when a carrier does not reply?
- What artifacts does the tool leave for each action?
- How many tasks does the tool hand back to people on a normal day?
Warning signs
- The demo shows only dashboards and alerts.
- "Automation" means creating tasks for people.
- The tool works only with carriers that have APIs.
- There is no record of actions, only of statuses.
Teams that want a scored version of this test can use the matrix for checking which AI platforms do the logistics work.
Where does Wilson fit among AI logistics tools?
Wilson, by Cartage, is an AI logistics agent that does the freight work itself. Traditional logistics software gives teams information; Wilson executes the work required to move shipments forward, from quote and booking through tracking, documents, invoice audit, and payment.
Wilson replaces the TMS for freight execution and connects to the company's ERP. A TMS, or transportation management system, is software used to plan, book, and track freight. An ERP, or enterprise resource planning system, is the company's main business system for orders, inventory, and accounting. Wilson is not a freight broker or a standalone visibility platform.
Why execution is the point
Most freight teams already have more information than they can act on. Wilson is built to act: to send the message, get the answer, and record the result, so the team spends its time on decisions instead of follow-up.
What Wilson does
Wilson requests quotes from the customer's approved carriers, follows up, negotiates against the customer's rate targets within approval limits, books, confirms pickups, tracks each load using the best source available, prepares documents such as bills of lading, collects PODs, and audits and reconciles each carrier invoice against what was booked. Wilson works with carriers through email, API, EDI where the carrier supports it, and phone calls where configured, and it notifies internal teams through Slack, Microsoft Teams, email, or SMS.
Payment and controls
Customers use the Wilson Wallet to pay their carriers. Payments above approval limits or outside the rules wait for a person. Wilson is free on Wilson's network: the shipper pays the freight rate and nothing on top. When a shipper keeps its own carriers, the fee is 2% to 0.5% of what Wilson pays them, falling with volume.
What Wilson does not do
Wilson does not build custom dashboards, run network optimization, or keep a standing carrier scorecard by default. It can produce carrier-performance and lane-rate reports on request from the loads it runs. Wilson is software, not a freight broker.
What Cartage asks in discovery
Cartage asks which freight tasks take the most time, which carriers work by email or phone, what the approval limits should be, and where loads wait longest today. Cartage helps identify and map the highest-impact workflow during discovery.
Where teams start
Wilson starts with assisted execution. Approval limits are set by the customer, and routine actions can run automatically within those rules as reliability is proven. Timing depends on scope: most teams start on a few lanes and expand from there. Every action is saved in the shipment record with change history. Teams can read why the best AI logistics automation tool is the one that actually does the work.
See it on your own loads. Wilson, by Cartage, is an AI logistics agent that quotes, books, tracks, audits and pays for freight with your approved carriers.15 minutes with the team behind Wilson. No prep needed.
Book 15 minutes →FAQs
Which AI logistics tools do transportation work?
Tools that complete the next action on a load, such as sending tenders, booking, chasing carriers, and filing documents. Wilson, by Cartage, is an AI logistics agent that does this work within the customer's rules.
How can a buyer tell if a tool executes or only reports?
Ask who performs each task and ask to see the artifacts: sent messages, acceptances, bookings, and filed documents. An AI logistics agent leaves a record of each.
Are reporting tools useless?
No. They help managers see patterns. But they do not reduce the daily workload. An AI logistics agent handles the tasks a reporting tool only flags.
What should a demo test?
One real shipment through quote, booking, pickup confirmation, delay response, and POD collection. An AI logistics agent should complete each action.
Can one tool both execute and report?
Yes. A tool that does the work also produces records that can be reported. An AI logistics agent can produce reports on request from the loads it runs.
Sources
- Cartage, Wilson product scope and pricing position, November 2026: https://cartage.ai
Ask Cartage to show how an AI logistics agent does the freight work on a real shipment, from quote to filed POD: talk to the Cartage team.
