What should a one-month freight automation pilot prove on our own lanes?
Key Takeaways
A one-month pilot should show how much freight work gets completed, how much work your team still does and which loads remain open. Compare the same type of freight before and during the test. A good result needs records, clear definitions and an agreed decision at the end.
- Choose a small group of repeat lanes.
- Measure the current process before the pilot starts.
- Track booking time, manual touches, missed pickups and document collection.
- Show unfinished work beside completed work.
- Agree when to expand, extend or stop.
What should the pilot prove?
Freight management software should help your team finish the recurring work around a shipment. That may include collecting freight quotes, obtaining carrier acceptance, checking a delay, collecting a signed document and reviewing an invoice. Start by naming the work you want the pilot to own.
A pilot becomes hard to judge when its scope keeps changing. If week one covers two LTL lanes and week four adds parcel shipments from another plant, the combined average may hide what happened. Keep the original group visible. Report new work separately.
Write a short test statement. For example: “On these two lanes, we want fewer manual follow-ups without more missed pickups or weaker approval control.” This is a proposed test goal, not a promised outcome.
An AI logistics coordinator should be evaluated on the work assigned to it. If finance still reviews every bill by policy, keep those reviews in the workload count. Explain why they remain. Do not report the whole shipment as hands-free because the quote request was automatic.
The existing pilot setup guide explains how to organize a trial. This article focuses on the measurement sheet and the decision it supports. Confirm the actual pilot scope and commercial terms with Cartage; a one-month measurement window does not establish a standard contract term.
Which measures belong on the pilot sheet?
Use a short set of measures with a source for each number. The team should be able to open a shipment and understand why it counted as complete, late or still waiting.
| Measure | Simple definition | Evidence |
|---|---|---|
| Manual touches per load | Team actions divided by eligible loads | Emails, calls, reviews and corrections |
| Time to carrier acceptance | Complete request to explicit acceptance | Request and acceptance timestamps |
| Missed pickup rate | Eligible scheduled pickups missed / eligible scheduled pickups | Pickup window and event evidence |
| Exception response time | Known problem to first useful action | Alert, carrier contact and response |
| POD collection time | Confirmed delivery to signed POD filed | Delivery evidence and document receipt |
| Open work | Loads or tasks not finished at cutoff | Current state, owner and next action |
Define “manual touch” before the first week. A touch might be a call, an email, a correction or an approval. Count the same actions in the baseline and pilot. Otherwise a new counting method can look like an improvement.
Keep the starting event clear. An ERP order created at midnight may not be ready for freight until dimensions arrive at 10 a.m. Measure total waiting time if it matters, but also show the wait for missing order information separately from carrier response.
Use the middle completed time, often called the median, beside the slowest cases. A few long waits can move an average sharply. Neither number replaces the list of open requests. A request with no carrier acceptance has no completed booking time yet.
How do you build a fair baseline?
Select a recent period with comparable lanes, freight modes and service needs. Record the number of loads and the period covered. Note unusual events such as a plant shutdown or a large customer rush. Those conditions help explain differences; they are not automatic reasons to remove a bad result.
Collect the baseline before changing the process when possible. If historical timestamps are incomplete, mark the gaps. Do not create precise old booking times from memory. A short observed baseline can be more useful than a long period with weak records.
Compare effort per load as well as total effort. In a fictional example, a team spends 240 minutes on 24 loads before the pilot and 150 minutes on 25 loads during it. That is 10 minutes per load before and 6 minutes during. The example shows the calculation only. It is not a Cartage customer result.
Separate recurring operating work from setup work. Mapping fields, sharing the approved carrier list and checking the first files are real costs. Show them as setup effort. Do not hide them, and do not assume every setup hour will repeat each week.
Supply chain management software can touch many processes outside freight. Keep this pilot limited to the assigned transportation work. Purchasing, production planning and warehouse labor are separate measurement areas unless they are expressly included.
What should happen each week?
Before the first live load, agree the scope, owners, evidence and success criteria. Check that the order data arrives, the approved carriers are correct and approval requests reach a person who can answer. Start the measurement clock when the defined live workflow begins.
In the first week, inspect a few complete shipment histories. Look for missing events and duplicate actions. If a team member calls a carrier after the AI logistics employee already called, record both actions and find out why. Unclear ownership is part of the result.
In the second week, review the slowest and unfinished cases. Ask whether each delay came from missing shipment facts, carrier silence, an internal approval or a technical problem. Assign one next step. A useful review changes an unresolved case into an owned action.
In the third week, include a realistic exception if one has not occurred naturally. Use a safe demonstration or agreed test, such as an unanswered quote or an above-limit offer. Label simulated cases separately. A simulated missed pickup must not enter the live missed-pickup rate.
In the final week, stop adding new scope and prepare the comparison. Take a dated snapshot of open loads. Let the team see which outcomes are complete and which need more time. An invoice that has not arrived cannot count as an invoice successfully audited.
How should cost and approval control be checked?
Keep freight spend, coordination effort and software charges in separate columns. A cheaper rate does not prove fewer coordinator hours. Fewer emails do not prove lower freight cost. Measure each result using the evidence that fits it.
For freight cost, compare similar shipments and included services. A new liftgate requirement or rush delivery can explain a higher bill. Keep approved changes visible. A shipping cost comparison based only on the first quoted price can miss later charges.
For approvals, count requests, decisions and waiting time. Review whether the right person received the request and whether the final action matched the approved amount and shipment version. Fast booking is not a good result if it skips a required decision.
Wilson has no subscription or platform fee. Pricing is a percentage of the freight he coordinates, with pricing scaling down as volume grows. Enterprise pricing is custom. Use the actual proposal when calculating pilot cost. Do not insert an assumed percentage or promise a saving.
Wilson starts with assisted execution and works under the customer’s rules. A person handles actions that need approval or fall outside those rules. A pilot can test whether the boundaries work; it should not treat fewer approvals as a goal by itself.
How do you decide whether to expand, extend or stop?
Expand when the agreed work is being completed, required controls hold and the evidence supports the next lanes. State what changes with expansion: more volume, another mode or a different level of authority. Do not increase all three just because one month ended.
Extend when the original question remains unanswered for a clear reason. Perhaps too few eligible loads moved or invoices have not arrived. Name the missing evidence and the next review date. An extension without a question can become an endless trial.
Stop or narrow the scope when a required workflow fails and there is no accepted correction. Preserve open shipments, carrier commitments, documents and invoice responsibilities. The end of a software test does not end the obligation to finish the freight work.
When you want to automate logistics with Wilson, bring Cartage the actual lane data and one difficult case. Wilson can coordinate quoting, booking, tracking, documents, invoice audit and payment under configured rules. Ask to see the history and reporting available for the proposed workflow rather than assuming every metric has a ready-made dashboard.
Use the ERP workflow guide to define the intake boundary. A reliable order feed makes the pilot easier to measure, but order import alone is not proof that a carrier accepted the load.
FAQs
What should we measure in a one-month freight software pilot?
Measure manual touches, coordinator time, carrier acceptance time, missed pickups, exception actions, POD collection and open work. Use clear starting and finishing events, and compare similar loads.
How many loads does a pilot need?
There is no universal number. The sample should cover the lanes and decisions you want to evaluate. Report the count and any missing scenarios. Extend the measurement if the evidence is too limited.
Should unfinished loads be removed from the results?
No. Show them separately with their current state, age, owner and next action. Removing them can make completed booking or document times look better than the full workflow.
Does one month prove the software will work everywhere?
No. It supports a decision about the tested work. New modes, plants, carriers or authority levels need their own scope review and evidence.
Sources
- Cartage, Wilson product FAQ and product scope, November 2026: https://cartage.ai
- Cartage: pilot measurement guide