Should an AI agent show you several freight quotes or make one recommendation you can override?
Should an AI agent show several freight quotes or make one recommendation?
Here is a hypothetical Monday. A manufacturer of HVAC parts has eight pallets ready in Memphis for a distributor in Denver, delivery needed by Thursday. Four LTL carriers and one volume LTL option reply to the quote request within two hours. The lowest line rate is $1,412. The highest is $1,890. The logistics manager has three other loads to cover before noon and a call with a plant manager at 11.
If the agent sends her all five quotes in a table and asks which one to book, she has to do the comparison herself: check which quotes include the liftgate the Denver receiver needs, which carrier can actually pick up Tuesday, which transit time makes Thursday, and whether the cheapest carrier is one the team had trouble with in the spring. That's the work she hoped the agent would do.
If the agent books the $1,412 quote without showing anything, she finds out on Thursday that the carrier didn't include the liftgate, transits four days instead of three, and the receiver refused the late delivery.
An AI agent should make one recommendation based on total cost and your rules, show the quotes it compared and why it chose, and let a person override the pick before or after booking. Within the limits your company sets, the recommendation can turn straight into a booking. Outside those limits, the agent brings the recommendation and the options to the person who approves. Either way, every quote stays on the load, so anyone can see later what was available and why the load went where it did.
What a good recommendation looks like
A recommendation that a busy manager can act on in under a minute has five parts:
- The pick: carrier, total cost and transit.
- The reason in one line: "lowest total cost with liftgate, delivers Wednesday, approved for this lane."
- The runner-up: what it would cost to choose the next option and what you'd get for it.
- What was ruled out and why: "Carrier C cheapest line rate but excludes liftgate and delivers Friday."
- What happens next: "booking at 10:30 unless you change it" or "waiting for your approval because it's above $1,500."
That's a recommendation you can trust because you can check it. It's also one you can ignore when you're busy and the pick is inside your rules.
Let's review how your team picks between quotes today. 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.
Book a demo →What does a fair comparison of freight quotes need?
A fair comparison puts every quote on the same basis: the same shipment description, the same services, the full cost including fuel and accessorials, the transit against the required date, and whether the carrier can make the pickup. Without that, the agent's recommendation is only as good as the cheapest-looking number.
The Memphis to Denver quotes, normalized
Continuing the hypothetical, here's what the five replies look like once they're on the same basis:
| Carrier | Line rate | Fuel | Liftgate at delivery | Total cost | Transit | Makes Thursday? | Pickup Tuesday? |
|---|---|---|---|---|---|---|---|
| Carrier A | $1,520 | Included | $95 | $1,615 | 3 days | Yes | Yes |
| Carrier B | $1,480 | $210 | $85 | $1,775 | 2 days | Yes | Yes |
| Carrier C | $1,412 | $195 | Not quoted | $1,507 + liftgate if billed | 4 days | No | Yes |
| Carrier D | $1,890 | Included | Included | $1,890 | 2 days | Yes | Wednesday only |
| Volume LTL option | $1,650 | Included | $110 | $1,760 | 3 days | Yes | Yes |
The cheapest line rate (Carrier C) misses the delivery date and leaves the liftgate off, so its real cost is unknown and the service fails. Carrier D is fast but can't pick up until Wednesday. Carrier A is the lowest total cost that meets the date and the pickup. That's the recommendation, with Carrier B as runner-up if the team wants the extra day of margin for $160 more.
The fields that change the ranking
- Fuel. Some carriers quote all-in; others add a fuel surcharge that changes weekly. Compare after fuel.
- Accessorials. Liftgate, appointment, inside delivery, residential or limited access. A quote that leaves out a service the shipment needs isn't cheaper; it's incomplete.
- Transit against the required date. A quote that misses the date isn't a valid option unless the date can move.
- Pickup availability. A carrier that can't come on the ready date adds a day the quote doesn't show.
- Quote validity. A quote that expires before booking is a number, not an offer.
- Approval status. A carrier the company hasn't approved for the lane isn't an option, whatever its price.
The article on comparing freight quotes from approved carriers goes deeper on normalizing quotes. The point for this question is that the recommendation is only as good as the comparison underneath it, and both should be visible.
Why showing raw quotes isn't neutral
An agent that "just shows the options" might seem safer, but raw quotes sorted by line rate quietly recommend the cheapest-looking one. A busy person scanning a list will often pick the top row. Normalizing and recommending is more honest than an unsorted list, because it puts the real comparison in front of the person.
When should the agent decide, and when should it bring options to a person?
Let the agent book its recommendation when the load is inside your approval limits, on a lane with approved carriers, and the pick meets every rule; bring the options to a person when any of those conditions fails. The line between the two is a company decision, written as rules, not a feature setting.
A decision table
| Condition | Agent books its pick | Agent recommends and waits for a person |
|---|---|---|
| Total cost vs. approval limit | Under the limit | Over the limit |
| Lane | Lane with approved carriers and history | New lane, new customer or new mode |
| Carrier | Approved for this lane | Approved generally but untested on this lane |
| Price gap | Pick within normal range for the lane | Pick far above the usual rate, or every quote is |
| Service | All rules met (date, equipment, accessorials) | A rule can't be met by any quote |
| Customer requirement | No special requirement | Customer-specified carrier or service in play |
| Timing | Enough time before the pickup cutoff | Cutoff close; trade-off between cost and certainty |
Setting the approval limit
Most teams set the limit in dollars per load, sometimes per mode. A hypothetical example:
- LTL under $2,000: the agent books its recommendation if all other conditions pass.
- Truckload under $4,500 on a lane with three or more approved carriers: the agent books.
- Anything over the limit, any expedited move and any new lane: the agent recommends, and the logistics manager approves.
- Anything over $10,000: the agent recommends, and finance co-approves.
Numbers vary by company and by spend. The structure matters more than the values: limits by amount, extra conditions for new or unusual loads, and a named approver for each tier. For how finance teams set up their part, see how finance should approve freight quotes before booking.
A truckload example near the cutoff
Consider a second hypothetical load: a dry van from a plant in Ohio to a retailer's DC in Georgia, picking up tomorrow at 7 a.m. Three approved carriers quote by 2 p.m. The lowest is $2,950 and comes from a carrier that has hauled the lane twice. The next is $3,120 from a carrier that has hauled it forty times. The approval limit is $3,000. Two of the quotes are valid until 4 p.m.
Here the agent shouldn't pick quietly and shouldn't dump three numbers on the manager either. A useful message says: "Recommend Carrier B at $3,120, above the $3,000 limit by $120. Carrier A is $2,950 and inside the limit, but has two loads on this lane. Both quotes hold until 4 p.m. Approve B, book A, or ask me to negotiate B toward $3,000." The manager can answer in one line. If she doesn't answer by 3:30, the rule should say what happens next, such as booking the in-limit quote or escalating to a backup approver.
Notice what makes this a recommendation rather than a list. The agent named one option, explained the trade-off in plain words, priced the alternative and gave a deadline. That's the same work an experienced coordinator does before walking into the manager's office.
What the person sees for a load the agent booked alone
For loads inside the rules, the person doesn't need a message at all. They need to be able to open the load later and find the quotes, the pick, the total cost and the carrier's acceptance. A weekly summary of loads booked inside the limits, with any pick that came close to a limit flagged, is usually enough for a manager to stay in touch with the decisions without approving each one.
Why "always show me" doesn't scale
Some teams start by asking to see every comparison before anything is booked. That's a reasonable first week: it shows the team how the agent ranks quotes and builds confidence. As a permanent setup, it brings back the bottleneck the agent was meant to remove. A team moving 60 loads a week that approves every quote is spending its morning on decisions where the answer is obvious. Widen the limits as the recommendations prove themselves on each lane.
Why "never show me" doesn't work either
At the other end, letting the agent book anything without visible options leaves the team blind when a customer asks why the freight went with a slower carrier, or when finance asks why the cost on a lane jumped. Even for loads the agent books alone, the quotes should be on the record, so the decision can be checked after the fact.
How do you override an AI agent's quote recommendation?
There are three overrides: choose a different quote before booking, change the carrier after booking within the carrier's terms, or change the rule that produced the pick. Each should be easy and each should record a reason.
Override 1: before booking
When the recommendation is waiting for approval, or there's a window before the agent books, a person picks a different quote. Common reasons:
- The customer called and asked for a specific carrier.
- The team knows the recommended carrier's terminal is short on drivers this week.
- The runner-up is worth the extra cost for a high-value customer.
- A cheaper option exists outside the request, such as consolidating with another order going to the same region.
The person states the reason in a sentence, and the agent books the chosen quote. The original recommendation and the reason stay on the load.
Override 2: after booking
Sometimes the reason to switch appears after the tender is accepted: the customer moves the delivery date, or the carrier's pickup window slips. Switching then is a cancellation and a new booking, and the carrier's terms decide whether a cancellation fee applies. The agent should show the cost of switching (cancellation terms, the new quote, whether the remaining quotes are still valid) before anyone confirms it.
Override 3: change the rule
If the team overrides the same kind of pick three times on one lane, the right fix is the rule. Examples:
| Repeated override | Rule change |
|---|---|
| Team keeps choosing the runner-up for one customer | Add a customer rule: this customer's loads go to carriers with appointment-setting history |
| Team rejects a carrier on Friday pickups | Add a rule: no Friday pickups with that carrier on this lane |
| Team approves picks above the limit every time on one lane | Raise that lane's limit, or set a rate target for it |
| Team overrides to a faster carrier when the delivery date is under three days away | Add a transit buffer rule for short lead times |
Rules like these turn individual judgment into a standard the agent applies to every load. The team can edit them when circumstances change.
Record the reason
The reason for an override is the most useful piece of data in the whole process. "Customer asked" means the customer's preference belongs on the customer record. "Terminal short on drivers" is temporary and shouldn't become a permanent rule. "Runner-up is more reliable" means the comparison is missing something, and the team should say what. For a closer look at keeping that reasoning, see how freight software can record why a carrier was selected.
How do you tell whether the agent's recommendations are good?
Measure how often the team overrides the pick, why, and whether the booked loads delivered at the quoted cost on time. A good recommendation is one the team rarely changes and the invoice confirms.
Five measures worth tracking
- Override rate. Share of recommendations a person changed. A high rate in the first weeks is normal; a high rate that doesn't fall means the rules or the comparison need work.
- Override reasons. Grouped by type: customer request, carrier concern, timing, cost. Each group points to a different fix.
- Quote-to-invoice difference. For booked loads, how far the paid invoice sits from the quoted total. A gap means the comparison missed charges.
- On-time delivery of recommended picks. Whether the carrier the agent picked delivered by the required date.
- Time from quotes received to booking. How long loads wait for decisions. A long wait on loads inside the limits means the team is still approving work that the rules allow the agent to book.
A hypothetical first month
| Week | Recommendations | Overrides | Main override reason | Rule change made |
|---|---|---|---|---|
| 1 | 48 | 14 | Customer-specified carriers not on record | Customer carrier rules added for 3 accounts |
| 2 | 52 | 8 | Delivery appointment carriers preferred for one DC | Appointment rule added for that receiver |
| 3 | 55 | 5 | Timing near cutoffs | Transit buffer rule for orders under three days |
| 4 | 50 | 3 | Mixed | None; limits widened on two lanes |
The numbers are illustrative. The pattern is what to look for: overrides fall as rules capture what the team knows, and the remaining overrides are genuine judgment calls.
Spot-check the loads nobody looked at
Once the agent books routine loads alone, pick five a week at random and read the comparison. Ask whether a person would have chosen the same quote. If not, find out what the person knows that the rules don't.
How does Wilson, by Cartage, choose between freight quotes?
Wilson, by Cartage, is an AI logistics agent that requests quotes from the company's approved carriers and the network, compares them on total cost against the shipment's requirements, books within the company's approval limits and keeps every quote on the load record. Traditional logistics software gives teams information. Wilson executes the work required to move shipments forward. Instead of handing the team another screen of prices to sort, Wilson does the comparison and the booking, and shows its work.
Wilson works as a free logistics department for manufacturers and distributors: what big shippers build over years, Wilson does from day one.
The labor
The labor is requesting, comparing and booking. Wilson does the work start to finish, from the first quote to the final POD: quotes, bookings, paperwork, tracking and carrier follow-ups. It sends the same shipment details to every carrier so the quotes compare, follows up on carriers that haven't replied, and negotiates toward the company's rate targets within approval limits. When the recommendation is inside the rules, it books. When it isn't, it brings the recommendation and the quotes to the right person by email, Microsoft Teams, Slack, WhatsApp, text or phone, and books once someone approves or picks a different quote. The team doesn't chase carriers.
The data
The data is why the team can see the options after the fact. Wilson keeps every order, load, quote, rate, document and contact on record as it works, connected to the company's ERP. The quotes it compared stay with the load, along with the pick, and every message, confirmation and change, including an override, is kept in the shipment's change history. Rules and preferences, such as which carriers get first look on a lane or which receivers need appointments, are learned from Wilson's work or added by the team, and the team can edit them. Records are searchable and exportable with change history. Wilson replaces the TMS for freight execution and connects to the company's ERP. 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.
The network
The network adds more quotes to compare. Wilson's network includes carriers and brokers, and the shipper's volume counts with every other Wilson customer's, which can bring better freight rates on its lanes. The company still approves who hauls its freight, so a network quote enters the comparison only from carriers the company has approved.
The payments
The payments check the recommendation against reality. Wilson audits every freight invoice against the load before money moves, then pays carriers from the Wilson Wallet. When an invoice is over the rate, Wilson asks the carrier why before paying, and invoices over the rate or outside the rules wait for a person. That closes the loop on the comparison: if a quote left out a charge, it shows up here.
How Wilson's quote choices work, step by step
| Step | What Wilson does | Where a person can step in |
|---|---|---|
| Order is complete | Requests quotes from approved carriers and the network | Before quotes go out, if the lane needs review |
| Quotes return | Normalizes them on total cost, transit and service | Not needed |
| Pick is inside the rules | Books it, with written acceptance and a rate confirmation, and keeps the quotes on the load | Before booking, if the team wants a different quote |
| Pick is outside the rules | Sends the recommendation and options to the approver | The approver confirms or chooses another quote |
| A person overrides | Books the chosen quote; the change is kept in the change history | The team can later turn repeated overrides into rules |
| Invoice arrives | Audits it against the booked quote and documents | When the invoice doesn't match |
Teams begin with assisted execution, approving Wilson's picks while they see how it ranks quotes. Approval limits are set by the company, and routine bookings can run automatically inside them as each lane proves out. Wilson covers LTL, FTL, ocean, drayage, parcel and cross-border freight. It is not a freight broker or a customs broker.
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. Timing depends on scope: most teams start on a few lanes and expand from there.
For the rules side of this decision in more depth, see configurable approvals for an AI logistics agent.
FAQs
Will Wilson show us the quotes or just pick one?
Wilson, by Cartage, an AI logistics agent, does both. It compares every quote on total cost against the shipment's requirements and recommends one. Inside your approval limits it books the pick; outside them it sends the recommendation and the options to the approver. Every quote stays on the load record either way, so the team can see what was available and why.
Is the cheapest freight quote always the best choice?
No. The cheapest line rate can leave out fuel or a needed accessorial, miss the delivery date, or come from a carrier that can't pick up on the ready date. Compare total cost and service, and choose the lowest total that meets every requirement.
How many freight quotes should we compare per load?
Enough to see the market on the lane, usually three to five from approved carriers. More quotes help on volatile or unfamiliar lanes. On a stable lane with a carrier that holds its rate, fewer can be enough if the rate is checked against recent loads.
What should happen when no quote meets our rules?
The agent should stop and bring the options to a person, saying which rule each quote fails. The person decides whether to relax the rule for this load, move the date, or look for more carriers.
Should overrides change the rules automatically?
No. Overrides should be recorded with a reason, and the team should decide which ones become rules. A one-time customer request shouldn't change how every load on the lane is booked.
Can Wilson book a quote that is above our approval limit?
No. Rates outside the limits go to the logistics team. Wilson sends the recommendation and the other quotes to the approver and books only after someone says yes.
Does Wilson score carriers when it recommends a quote?
Not by default. It compares quotes on total cost against the shipment's requirements and the company's rules. If the team wants a view of carrier performance, Wilson can produce carrier-performance and lane-rate reports on request from the loads it runs.
Can we require approval on every load at first?
Yes. Many teams approve every pick for the first weeks to see how quotes are ranked, then widen the limits lane by lane as the recommendations hold up.
Sources
- Cartage, Wilson product FAQ and product scope, October 2026: https://cartage.ai
Let's review how your team picks between quotes today. You don't need to prepare anything: Cartage maps the workflow with you on the call.
