What logistics work can an AI agent actually do today, and what still needs a person?
Short answer
An AI agent can do most of the daily freight work today. That means quotes, booking, pickups, tracking, carrier follow-up and documents. It can also check invoices and pay carriers within your rules. People still set the rules, approve anything outside them and deal with carriers and customers. Wilson, by Cartage, an AI logistics agent, does that daily work and sends the out-of-rule decisions to your team.
Book 15 minutes with the team →What logistics work can an AI agent do today?
It is Wednesday afternoon. The logistics manager at a building products manufacturer is reading a list of open tasks. Twenty loads need quotes for next week. Six carriers have not confirmed pickup. Three PODs are missing, so finance cannot invoice those customers. Two carrier invoices are higher than the rate confirmation. And one customer wants to know why yesterday's load is late. The manager's question is simple: which of these could an AI agent take off the list today, and which ones still need someone on the team?
An AI agent can do most of the daily freight work today: request quotes, compare them, negotiate within limits, book carriers, schedule pickups, track loads, follow up with carriers, handle routine exceptions, create documents, request PODs, audit carrier invoices and pay carriers within the company's rules. People still set those rules, approve anything outside them and own the relationships with carriers and customers.
That answer has a condition built in. The agent works inside rules. It books only carriers the company approved. It negotiates up to a limit and stops. It pays an invoice that matches the rate confirmation and the POD, and holds one that does not. Those rules are what make the work safe to hand over.
What "doing the work" means
Many tools that call themselves AI show information. They flag a late load or suggest a carrier. A person still sends the email, makes the call and updates the record. An AI agent is different when it takes the action itself: it sends the quote request, calls the carrier, books the load, saves the confirmation and tells the team what happened. When you evaluate one, ask to see the action, not a dashboard.
Where the work starts
The work starts when an order arrives. Today, orders can reach an AI logistics agent by email, from an ERP connection or as spreadsheet files, depending on the product and setup. The agent reads the order, checks that it has what it needs to quote and book, and asks for anything missing. It should never guess a missing address, date or lot number.
Want to sort your own list?
Let's review which tasks on your team's list an AI agent can take first. You don't need to prepare anything: Cartage maps the workflow with you on the call.
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 tasks can an AI agent run, step by step?
Follow one load from order to payment. At each step there is work an AI agent can do and a decision that may need a person. The table shows the split for a typical shipper-run load.
| Step | Work an AI agent can do | When a person decides |
|---|---|---|
| Order intake | Read the order, check required details, ask for missing ones | When the customer cannot give the detail, or the date is not possible |
| Quoting | Request quotes from approved carriers, compare on total cost | When no approved carrier quotes, or all quotes are above the limit |
| Negotiation | Counter toward the rate target, within approval limits | When the best rate is still above the limit |
| Booking | Tender, get written acceptance and a rate confirmation | When the rule says a load needs approval first |
| Pickup scheduling | Confirm the pickup date and time with the carrier | When the carrier can only pick up later than agreed |
| Documents | Create the bill of lading, packing slips and pallet labels | When a document needs a detail only the team has |
| Tracking | Get status from the best source per load, follow up when quiet | Rarely, unless the carrier cannot be reached at all |
| Exceptions | Contact the carrier, tell the team and customer, start rebooking | When rebooking falls outside the rules |
| POD | Request, chase and file the signed POD | When the POD shows damage or a shortage |
| Invoice audit | Check the invoice against the rate confirmation and POD | When the invoice is over the rate or has new charges |
| Payment | Pay the carrier within the rules | When the amount is above the approval limit or outside the rules |
Order intake
An agent reads the order and checks for what it needs: addresses, dates, weight, pallet count, freight class, special services and any customer rules. If something is missing, such as a lot number, or a date is impossible, a good agent holds the order and asks. A person decides only when the answer changes the deal, such as a new delivery date.
Quoting and negotiation
An agent sends the same request to each approved carrier on the lane, collects the answers and compares them on total cost, including fuel and expected accessorials. It can counter-offer toward the company's target and stop at the approval limit. Above the limit, a person decides. See AI freight negotiation within approval limits for how those limits work.
Booking and scheduling
An agent tenders the load, gets written acceptance and a rate confirmation, and confirms the pickup. If the carrier says it cannot pick up until a later day, the agent should not accept that quietly. It should show the team the option and let a person decide.
Tracking and follow-up
An agent gets status from the best source on each load: a carrier connection, GPS or ELD data when shared, the carrier's website, or an email or phone call when nothing else is current. When a carrier goes quiet, it follows up by phone and email until it gets an answer. This is the work that takes most of a coordinator's day, and it is one of the best places to start.
Exceptions
When a pickup is missed or a load is late, an agent contacts the carrier, tells the team and tells the customer if that is set up. It can start rebooking with another approved carrier within the rules, or after approval. A person decides when the fix costs more than the rules allow or changes what the customer was promised.
Documents, POD and payment
An agent creates shipping documents from the booked load, requests the POD after delivery and files it on the shipment. It audits each carrier invoice against the rate confirmation and the POD. Invoices that match can be paid within the rules. Invoices over the rate, with new charges or above the payment limit, wait for a person.
Updates to the team and customers
An agent can also keep people informed without being asked. It can tell the team about a delay in Slack, Microsoft Teams or email, and tell a customer or consignee by email or text where that is set up. It can follow up with people on the company's own team who have not answered a question, in the same thread, and then raise it to the team. This removes a lot of the small interruptions that fill a coordinator's day, because the answer reaches people before they ask for it.
Documents for cross-border loads
On cross-border freight, an agent can prepare documents such as the commercial invoice and packing list and send the information to the company's customs broker. The broker still clears the goods. That split matters, because it keeps a licensed party responsible for customs.
What still needs a person, and why?
People should keep the decisions that set the rules, that fall outside the rules, or that change a promise to a customer or a carrier. An AI agent does the work. People decide what the work should aim for and handle the cases the rules did not cover.
Decisions people should keep
- Approving carriers. Which carriers may haul the company's freight is a business decision. An agent should book only from the approved list.
- Setting targets and limits. Rate targets per lane, approval limits, and payment limits come from the team and finance.
- Out-of-rule cases. A rate above the limit, an invoice that does not match, a payment above the limit.
- Date changes the customer cares about. If a carrier can only pick up a day late, someone who knows the customer should decide.
- Possible fraud signals. A change in a carrier's bank or remit-to details should go to a person, never be acted on by the agent alone.
- New rules. If the team keeps changing the same thing on loads, a person should decide whether that becomes a rule.
- Relationships. Annual rate talks, a hard conversation with a failing carrier, a call with an upset customer.
Work that is outside an AI logistics agent's job
Some work belongs to other roles entirely:
- Customs clearance belongs to a licensed customs broker. An agent can prepare documents and send information to the broker, but it does not clear customs.
- Freight brokerage means arranging transportation as an intermediary. An agent that works for the shipper is not a broker. See AI logistics agent vs freight broker for the difference.
- Route and network planning, such as where to put a warehouse or how to design a multi-stop network, is a planning job, not daily execution.
Common mistakes when deciding what to hand over
Mistake 1: Handing over decisions before rules exist. If the team has no written approval limits, the agent has nothing to follow. Write the rules first.
Mistake 2: Keeping every action under review forever. Approving each routine step is useful at the start. After the work is proven, routine actions inside the rules can run on their own. Otherwise the team just has a new queue to check.
Mistake 3: Believing claims without proof. Ask any vendor to run a real load and show the emails, calls and saved documents. A slide that says "AI does it" is not proof.
Mistake 4: Starting with the hardest lane. Start with lanes and customers where the work is routine. Add harder ones after the routine work runs well.
Mistake 5: Assuming the agent knows what the team knows. Preferences like "this customer only takes appointments in the morning" need to be added or learned. A good agent asks before turning a one-time change into a rule.
How do you check that an AI agent did the work correctly?
Check the record, not the claim. Every action an AI agent takes should leave proof on the shipment: the message it sent, the answer it got, the document it saved and the change it made, with the time and who did it. If you cannot see that proof, you cannot trust the work, and you cannot hand over more of it.
What proof looks like for each task
- Quote: the request sent to each carrier, each answer with its total, and the comparison.
- Booking: the tender, the carrier's written acceptance and the rate confirmation.
- Pickup: the confirmed date and time, and who at the carrier confirmed it.
- Tracking: each status update with its source, such as a carrier connection, the carrier's website or a call.
- Follow-up: each email or call to the carrier, and the answer.
- Exception: what went wrong, who was told and what was done next.
- POD: the signed document, filed on the load.
- Invoice audit: the invoice next to the rate confirmation and the POD, with any difference marked.
- Payment: the amount, the date and the rule that allowed it, or the person who approved it.
How to review in the first weeks
In the first weeks, set aside a short review each week. Pick a handful of loads at random and read their history from start to end. Ask three questions:
- Did the agent do the right thing at each step?
- Did it ask for help when it should have?
- Did it leave enough on the record that anyone on the team could follow what happened?
Also read every case where the agent asked a person for a decision. These show where the rules are unclear. If the same question comes up again and again, the fix is usually a new or clearer rule, written by the team.
Signs the work is going well
- The team gets fewer "where is my load?" questions, because status is on the record.
- Exceptions reach the team early, with the carrier already contacted.
- PODs are filed before finance asks for them.
- Invoices that do not match are caught before payment, not after.
- People stop opening carrier websites and side spreadsheets for loads in scope.
Signs something needs fixing
- The agent asks for help on routine loads. The rules may be missing a case.
- People still track loads by hand "just to be sure." Find out what they do not trust and look at those records.
- Customers hear about delays before the team does. Check how exceptions are routed.
- The same manual change appears on many loads. It may need to become a rule, decided by the team.
A hypothetical first month
Consider a hypothetical distributor that starts an AI agent on three outbound lanes. In week one, a person approves every booking. The weekly review shows that most quotes and bookings were fine, but two orders were held because the lot number was missing. The team adds the lot number to the order template. In week two, bookings inside the approval limit run on their own, and the review focuses on exceptions. By week four, the team adds two more lanes. The numbers here are made up. The pattern is the point: review, fix the rules, then widen the scope.
Which tasks should you hand to an AI agent first?
Start with the tasks that happen on every load, follow clear rules and take the most time. For most teams, that means tracking and carrier follow-up, quote requests, POD collection and invoice checks. Use this checklist to sort your own list.
| Task | Happens on most loads? | Clear rule exists? | Good first candidate? |
|---|---|---|---|
| Carrier follow-up for status | Yes | Yes: ask until answered, tell the team if late | Yes |
| Quote requests to approved carriers | Yes | Yes: same lane, approved list | Yes |
| POD requests and filing | Yes | Yes: request after delivery, file on the load | Yes |
| Invoice check against rate confirmation and POD | Yes | Yes: match or hold | Yes |
| Booking within approval limits | Yes | Needs written limits | Yes, once limits are set |
| Pickup scheduling | Yes | Needs dock hours and rules | Yes, once rules are set |
| Shipping documents | Yes | Needs templates and customer rules | Yes |
| Rebooking after a dropped load | Sometimes | Needs a ranked carrier list | After the basics run well |
| Carrier payment | Yes | Needs payment limits | After the audit runs well |
| Approving a new carrier | Rarely | Business decision | No, keep with a person |
| Annual rate talks | Once a year | Business decision | No, a person leads; the agent can help prepare |
How to roll it out
- Pick a few lanes or one or two customers.
- Write the rules: approved carriers, targets, approval limits, payment limits, customer preferences.
- Start in assisted mode. The agent does the work and a person approves key steps.
- Review the change history weekly. Look for cases where the agent asked for help and why.
- Let routine actions run on their own once they are proven.
- Add lanes, then customers.
What to measure
Measure what changed for the team: time spent on follow-up, how many status questions came in, how fast PODs were filed, how many invoice errors were caught before payment. Use your own counts from before and after. For a fuller framework, see how to evaluate an AI logistics agent.
How does Wilson, by Cartage, split logistics work between the agent and the team?
Wilson, by Cartage, is an AI logistics agent that runs freight work from the first quote to carrier payment for manufacturers and distributors, and sends decisions outside the company's rules to the logistics team. Traditional logistics software gives teams information. Wilson executes the work required to move shipments forward.
Wilson works as the company's logistics department. What it does today falls into four parts:
- The labor. Wilson requests quotes from the company's approved carriers and compares them on total cost. It negotiates toward the company's rate targets within approval limits; rates outside the limits go to the logistics team. It tenders and books with written acceptance and a rate confirmation, schedules pickups and checks carrier availability before booking. It tracks each load with the best source available: carrier API or EDI, GPS or ELD data where the carrier shares it, the carrier's website by PRO number, and email or phone calls when nothing else is current. It follows up with carriers by phone and email until it gets an answer. When a pickup is missed, a load is cancelled or delayed, Wilson contacts the carrier, tells the team, tells the customer if configured and starts rebooking from the approved carrier list automatically within the rules or after approval. It generates rate confirmations, bills of lading, packing slips and pallet labels, and requests and files PODs.
- The data. Wilson keeps every order, load, quote, rate, document and contact on record as it works. Every message, confirmation and change is kept in the shipment's change history. Wilson replaces the TMS for freight execution and connects to the company's ERP.
- The network. The shipper's volume counts with every other Wilson customer's to get better rates on each lane. The network includes carriers and brokers. The company still approves each carrier before Wilson uses it.
- The payments. Wilson audits each carrier invoice against the rate confirmation and the POD, then pays carriers from the Wilson Wallet. Payments above the approval limits or outside the rules wait for a person. A change in a carrier's remit-to or bank details is flagged to the team, and Wilson does not act on it alone.
What stays with the team
The company approves carriers, sets rate targets and approval limits, and decides on anything outside the rules. When a carrier offers a later pickup, Wilson shows the option to the team and leaves the agreed date unchanged until a person decides. When the team changes something on one load, Wilson treats it as a change to that load. If it looks like a new rule, Wilson asks first.
Wilson does not clear customs, is not a freight broker and does not do route or network planning. It covers LTL, FTL, ocean, drayage, parcel and cross-border freight.
Wilson works with SAP, NetSuite, Microsoft Dynamics 365, Oracle, Infor, Epicor, Sage, Acumatica and other ERP environments, with the exact connection depending on the customer's data flow and scope.
Control starts with assisted execution. The company sets approval limits, and routine actions can run automatically within those rules as the workflow is proven. Timing depends on scope: most teams start on a few lanes and expand from there.
Pricing
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. There are no seat fees, no setup cost and no contract.
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
What can Wilson do today, and what still needs our team?
Wilson, by Cartage, an AI logistics agent, quotes, negotiates within limits, books, schedules pickups, tracks, follows up with carriers, handles exceptions, creates documents, files PODs, audits invoices and pays carriers within the company's rules. The team approves carriers, sets targets and limits, and decides anything outside the rules. Wilson does not clear customs, broker freight or plan networks.
Can an AI agent really make phone calls to carriers?
Some can. Wilson calls carriers where configured, and phone calls are supported in English and Spanish today. It can also call or text a driver when it has the driver's number.
Will an AI agent book a carrier we have not approved?
It should not. Wilson books only carriers the company has approved. Cartage can suggest additional capacity, but the company approves each carrier before Wilson uses it.
What happens when the agent is not sure?
A good agent asks instead of guessing. If an order is missing a detail, Wilson holds it and asks for the detail. If a quote, invoice or payment falls outside the rules, it goes to a person.
Does an AI agent replace the logistics team?
It takes over the repeat work, so the team spends its time on decisions, carriers and customers. The team still sets the rules and handles the cases the rules do not cover.
Can an AI agent create carrier scorecards?
Wilson does not keep a permanent scorecard by default. It can produce carrier-performance and lane-rate reports on request from the loads it runs.
Can an AI agent upload documents to customer portals?
Wilson does not upload documents to customer portals. It can request, chase, file and attach documents configured in the workflow.
Does the AI agent learn our preferences?
Wilson keeps preferences and rules that the team adds or that it learns from its own work, and the team can edit them. It asks before turning a one-time change into a permanent rule.
Is our data used to train the AI?
Wilson does not use customer data for training. Each customer's data is encrypted and kept inside that customer's organization.
Which modes can an AI logistics agent handle?
It depends on the product. Wilson covers LTL, FTL, ocean, drayage, parcel and cross-border freight. On cross-border loads, it prepares documents and sends information to the customs broker, but it does not clear customs. Ask any vendor to show a real load in each mode you ship before you rely on it.
Sources
- Cartage, Wilson product FAQ and product scope, October 2026: https://cartage.ai
Let's review which tasks on your team's list an AI agent can take first. You don't need to prepare anything: Cartage maps the workflow with you on the call.
