We pick carriers on service and trust, not just price. Can an AI agent learn which carriers we trust?

Freight pickups guideNov 15, 202617 min read

Can an AI agent learn which carriers your team trusts?

Consider a hypothetical distributor of building products with two warehouses and about 70 outbound loads a week. Its logistics lead has run freight there for eleven years. She doesn't book on price alone, and everyone knows it. One flatbed carrier gets the long lumber loads because its drivers tarp properly without being asked. A regional LTL carrier gets anything going to a particular contractor's yard, because it's the only one whose drivers call ahead and wait for the forklift. One truckload carrier with excellent rates never gets loads into the city on Fridays, after two missed appointments a year ago. None of this is in any system. It lives in her head and in the habits of the two coordinators she trained.

Then she takes three weeks of leave. The coordinators cover most of it from memory. One load to the contractor's yard goes to a cheaper carrier whose driver arrives unannounced, finds no forklift and leaves. The redelivery costs more than the rate difference and an angry phone call.

An AI agent can learn which carriers a team trusts when that trust is expressed as specific preferences and rules by lane, customer or situation, kept on a shared record the team can read and edit. The agent applies them on every load: it gives first look to the preferred carrier, keeps a carrier off a lane where the team has blocked it, and breaks ties in favor of the carrier the team prefers. It learns new preferences from what the team tells it and from the decisions the team makes on loads it handles, and the team can correct any of them.

What an agent shouldn't do is invent trust. It shouldn't decide on its own that a carrier is unreliable and stop using it, or quietly rank carriers by a score nobody agreed to. Trust is a company decision. The agent's job is to apply it consistently, and to make it visible.

What "learning" means here

There are two ways an agent picks up a preference:

  1. Someone tells it. "Always quote the flatbed carrier first on lumber loads from the north yard." The preference is stated once and applies from the next load.
  2. It notices a pattern in the team's decisions and asks. If the team overrides a booking to the same carrier on the same customer three times, the agent should ask whether that should become a rule, not assume it.

Both end the same way: a written preference, with a scope, that the team can see and change.

Let's review the carrier preferences your team keeps in its head. 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 →

How do you get carrier preferences out of people's heads?

Walk through recent loads with the people who booked them and ask why each one went to the carrier it did. Answers about real loads are specific and usable. Answers about carriers in general tend to be vague.

A one-hour preference session

This works well with the person who books most loads, plus one other coordinator:

  1. Pull 30 recent loads. Mix lanes, modes and customers. Include a few where a more expensive carrier was chosen.
  2. For each load, ask one question: "Why this carrier?" Write down the answer in their words.
  3. Sort the answers. Many will be "best price." The rest are the preferences worth capturing.
  4. Turn each preference into a rule with a scope. "They're good with the contractor" becomes "Deliveries to the contractor's yard: only carriers whose drivers call ahead; first look to the regional LTL carrier."
  5. Ask about the carriers that never come up. "Is there anyone you avoid?" usually surfaces the blocks: the carrier that damaged a load, the one that missed Friday appointments.
  6. Ask what would change their mind. "If the flatbed carrier's rate went 15% above the next one, would you still use them?" This sets the price limit on a preference.

What the session usually reveals

A hypothetical result for the building-products distributor:

What the lead said What it becomes
"They tarp right without me asking." First look on flatbed lumber loads from the north yard
"The contractor's yard needs someone who calls ahead." Customer rule: call-ahead required; first look to the regional LTL carrier
"Not into the city on Fridays after last year." Block: that truckload carrier, city deliveries, Friday
"If it's close on price, give it to the one who answers the phone." Tie-breaker: within 5% on total cost, prefer the carrier with the fastest confirmations
"The new reefer carrier is fine but I watch them." New carrier: non-critical loads only until reviewed
"Their driver shortage should be over by spring." Temporary note on one carrier, with a review date

Six sentences from one person became six rules anyone can read. That's the core of the work. The agent can only follow what's been said, and saying it out loud is often the first time the team has heard the reasoning.

Who else knows something

The logistics lead isn't the only source. Ask:

  • Warehouse leads: which carriers arrive on time, which drivers refuse to load certain freight, which ones block the dock.
  • Customer service: which customers complain about which carriers.
  • Accounts payable: which carriers send invoices that never match the rate.
  • Sales: which customers have named a carrier in their terms.

Each group sees a different side of a carrier, and a preference that only one group knows about is the easiest one to lose.

What kinds of carrier preferences can be written as rules?

Almost every carrier preference fits one of six types: blocks, requirements, first look, tie-breakers, price limits on a preference, and temporary notes. Naming the type tells the agent how strictly to apply it.

The six types

Type What it does Hypothetical example How strict
Block Keeps a carrier off a lane, customer, day or freight type No Friday city deliveries with one truckload carrier Absolute unless a person lifts it
Requirement Limits a load to carriers that meet a condition Contractor's yard: call-ahead carriers only Absolute for that customer
First look Gives one carrier the chance to quote or accept before others Flatbed lumber from the north yard: preferred carrier first Others are asked if it declines or doesn't reply in time
Tie-breaker Picks between close quotes Within 5% on total cost, prefer the fastest confirmer Applies only when prices are close
Price limit on a preference Says how much more the team will pay for a preferred carrier Preferred flatbed up to 10% above the next quote Above the limit, a person decides
Temporary note Applies for a period, then gets reviewed Carrier short on drivers; avoid hot loads until March 1 Expires or comes up for review on the date

First look in practice

First look is the most common preference and the most misunderstood. It doesn't mean "always book this carrier." It means "ask this carrier first, and give it a set time to accept." If the preferred carrier accepts at a rate inside the limits, the load is theirs. If it declines, doesn't answer in time, or quotes far above the market, the agent moves to the next approved carriers. The time window matters: a first look that waits all day for a reply turns a preference into a delay.

Scope is what makes a preference safe

A preference without a scope spreads where it shouldn't. "Use the regional LTL carrier" might be right for the contractor and wrong for a shipment two states away that the carrier doesn't serve. Every preference should say where it applies:

  • Lane (origin and destination, or region to region)
  • Customer or ship-to
  • Mode and equipment
  • Freight type (long lumber, fragile, temperature-controlled)
  • Day or time (Fridays, after 2 p.m., weekends)
  • Date range (for temporary notes)

What one written preference should contain

A preference is only as useful as the detail behind it. A complete entry on the shared record answers six questions, shown here for the hypothetical flatbed rule:

  • What: first look on flatbed loads of long lumber.
  • Where: from the north yard to any destination within the carrier's service area.
  • Why: drivers tarp and secure loads correctly without instructions; no damage claims on these loads in the past year, according to the lead.
  • Limit: the preferred carrier keeps the load if its total cost is no more than 10% above the next approved quote.
  • Who added it and when: the logistics lead, during the preference session, with the date.
  • When to review: at the next quarterly review, or sooner if a claim or a missed pickup appears on the record.

Written this way, the preference answers the questions a new coordinator or an auditor would ask. It also gives the agent everything it needs to apply the rule without guessing: the scope tells it which loads, the limit tells it when to stop, and the review date tells the team when to look again. A preference missing its reason is the one most likely to be deleted by mistake later, because nobody can tell whether it still matters.

Preferences and approval are different things

A preference only chooses among carriers the company has approved. It can't approve a carrier. If a coordinator says "my old contact at a carrier we've never used would be great on this lane," that's a request to approve a new carrier, which goes through the company's approval process first. The article on keeping approved carriers approved when booking is automated covers that boundary.

Which mistakes make carrier preference rules go wrong?

Most problems come from preferences that are too broad, never reviewed, in conflict with each other, or blind to price. Each one has a simple fix.

Mistake 1: a block that outlives its reason

A carrier gets blocked after one bad week. Two years later nobody remembers why, and the company is paying more on that lane because the best-priced carrier is still off the list. Fix: write the reason and the date with every block, and review blocks at least twice a year.

Mistake 2: a preference with no price limit

"Always first look to the preferred carrier" works until its rate drifts 25% above the market and nobody notices because it always accepts. Fix: give each preference a price limit relative to the other quotes, so the agent brings it to a person when the gap grows.

Mistake 3: conflicting rules

The customer rule says call-ahead carriers only. The lane rule gives first look to a carrier that doesn't call ahead. Which wins? Fix: decide the order once. A common order: blocks and customer requirements first, then lane first look, then tie-breakers.

Conflict Which rule should win
Customer requirement vs. lane first look Customer requirement
Block vs. first look Block
Temporary note vs. first look Temporary note, until its review date
Tie-breaker vs. lowest total cost outside the tie range Lowest total cost
Preference price limit exceeded vs. first look A person decides

Mistake 4: preferences hidden in one person's notes

A coordinator keeps a personal notes file of "who to use for what." The agent can't see it, and neither can the person covering their desk. Fix: move every preference to the shared record, even the informal ones. "Prefer them if it's close" is still worth writing down.

Mistake 5: treating every override as a new preference

When a coordinator picks a different carrier once because the customer called, that's not a rule. Fix: a person reviews repeated overrides on the record and decides whether they should become a rule. One-off decisions stay one-off.

Mistake 6: confusing trust with habit

Sometimes "we trust them" means "we've always used them." That's worth testing. Ask the carrier for a quote alongside others on the lane for a month and compare. If the trusted carrier is also the best total-cost option, the preference costs nothing. If it isn't, the team can decide whether the service is worth the difference. For a structured way to weigh that, see carrier selection software: how to weigh service against cost.

How do you check that trust still matches performance?

Review preferred carriers against what the load record shows, on a set schedule, and adjust the preferences where the record and the reputation disagree. The review should be a team decision based on evidence, not an automatic score.

What the record can show

For each preferred or blocked carrier, the load record holds the evidence:

  • Acceptance: how often the carrier accepted first-look loads, and how fast.
  • Pickup: whether pickups happened on the scheduled day.
  • Delivery: whether deliveries met the required date or appointment.
  • Communication: how often the carrier had to be chased for status or documents.
  • Invoices: how often invoices matched the rate confirmation, and how often they included charges nobody approved.
  • Exceptions: missed pickups, cancellations, damage claims.

None of these needs a permanent scorecard. They need a report pulled when the team sits down to review, from the loads that actually moved.

A quarterly preference review checklist

  1. List every active preference and block, with its reason and date.
  2. For each preferred carrier, pull acceptance, on-time pickup and delivery, and invoice match for the quarter.
  3. Compare the preferred carrier's total cost on its lanes with the other quotes received for the same loads.
  4. Mark preferences where performance slipped or the price gap grew beyond its limit.
  5. Review every block older than six months: is the reason still true?
  6. Review temporary notes past their date; remove or renew them.
  7. Ask coordinators and warehouse leads for anything new they've noticed.
  8. Update the record and note who changed what.

When the evidence and the reputation disagree

A carrier the team loves may show late deliveries on one lane. That doesn't always mean the trust is wrong; it may mean the lane is wrong for that carrier. Narrow the preference's scope before removing it. Equally, a blocked carrier may have improved. Lifting a block to non-critical loads for a few weeks is a lower-risk test than lifting it everywhere. For a closer look at measuring by lane, see comparing a carrier's on-time performance by lane.

How does Wilson, by Cartage, learn which carriers your team trusts?

Wilson, by Cartage, is an AI logistics agent that follows the company's carrier preferences and rules on every load, keeps them as memories the team can read and edit, and books only carriers the company has approved. Traditional logistics software gives teams information. Wilson executes the work required to move shipments forward.

Wilson works as a free logistics department for manufacturers and distributors: what big shippers build over years, Wilson does from day one. Here is how each part handles carrier trust.

The labor

The labor is applying the preferences on real loads. Wilson does the work start to finish, from the first quote to the final POD: quotes, bookings, paperwork, tracking and carrier follow-ups. It works from the company's ranked list of carriers for each lane (a waterfall) plus its decision rules by lane and mode: it gives first look to the carrier a lane rule names, moves to the next approved carriers if that carrier declines or doesn't reply in time, and keeps blocked carriers off the loads where the team blocked them. It compares quotes on total cost and negotiates toward the company's rate targets within approval limits. When a preferred carrier's quote falls outside the limits, the decision goes to the logistics team. The team doesn't chase carriers.

The data

The data is where trust is written down. Wilson keeps every order, load, quote, rate, document and contact on record as it works, connected to the company's ERP. It also keeps memories and rules, such as "a regional carrier gets first look on deliveries to one customer's yard" or "one receiving dock takes appointments only in the morning." The team adds them and can edit them at any time. When the team changes the carrier or time on one load, Wilson treats it as a change to that load, not a permanent rule. If a repeated change looks like it should become a rule, Wilson asks the team whether to make it one; it never adopts it silently. The company owns its carriers, rules, approval limits and contacts, and records are searchable and exportable with change history, so anyone can see when a preference changed and who changed it. 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 widens the choice without overriding it. 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. Cartage can suggest additional carrier capacity, but the company approves each carrier before Wilson uses it, and the company's preferences still decide which approved carrier gets first look. Shippers can keep their own carriers alongside the network.

The payments

The payments add evidence to the trust question. Wilson audits every freight invoice against the rate confirmation and POD before money moves, then pays carriers from the Wilson Wallet. Payments above the approval limits or outside the rules wait for a person, and when an invoice is over the rate, Wilson asks the carrier why before paying. A carrier whose invoices keep needing questions shows up on the record.

Scores only when the team asks

Wilson doesn't vet or score carriers by default, and it doesn't keep a permanent scorecard. When the team wants to review its preferences, Wilson can produce carrier-performance and lane-rate reports on request from the loads it runs. The team reads the report and decides; Wilson applies whatever the team changes.

Situation What Wilson does Who decides
Lane has a first-look carrier Asks that carrier first, within the set time The rule, set by the team
Preferred carrier declines or is slow Moves to the next approved carriers The rule
Preferred carrier's quote is outside the limits Sends the decision to the logistics team A person
Team overrides to the same carrier repeatedly Treats each change as a change to that load and asks the team whether it should become a rule The team; Wilson never adopts a rule silently
Team adds or edits a preference Applies it from the next load; change kept in history The team
Team wants to review carrier trust Produces a performance report on request The team

Control starts with assisted execution. The company sets approval limits, and routine bookings within those rules can run automatically as each lane proves out. Wilson handles 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.

FAQs

Can Wilson learn which carriers we trust?

Wilson, by Cartage, an AI logistics agent, follows the carrier preferences and rules the team sets by lane, customer or situation, such as first look, blocks and tie-breakers. It keeps them as memories the team can read and edit. When the team keeps making the same change, Wilson asks whether it should become a rule and never adopts it silently. It books only carriers the company has approved.

Will Wilson stop using a carrier on its own if it performs badly?

No. Wilson doesn't vet or score carriers by default, and blocking a carrier is a company decision. The team can ask for a carrier-performance report from the loads Wilson runs and change the rules based on it.

What is a first-look carrier?

A first-look carrier is asked to quote or accept a load before others on a lane, within a set time. If it declines, doesn't answer in time or quotes outside the limits, the load goes to the next approved carriers.

How do we capture carrier preferences that only one person knows?

Walk through about 30 recent loads with that person and ask why each went to its carrier. Turn every answer that isn't "best price" into a rule with a scope, a reason and a date.

Should every carrier preference have a price limit?

Yes. A preference without a limit can cost more and more without anyone noticing. Set how far above the next quote the team will pay, and send loads above that gap to a person.

What happens when two carrier rules conflict?

Decide the order once. Blocks and customer requirements usually come first, then lane first look, then tie-breakers. When a preferred carrier's quote exceeds its price limit, a person decides.

Can a preference make Wilson use a carrier we haven't approved?

No. Preferences choose among approved carriers only. Adding a new carrier goes through the company's approval process first.

How often should we review carrier preferences?

Quarterly works for most teams, with blocks older than six months and expired temporary notes checked every time. Update a preference sooner when something on the record clearly contradicts it.

Can we keep our own trusted carriers if we use the network?

Yes. Shippers can keep their own carriers alongside Wilson's network. The fee on loads paid to the shipper's own carriers is 2% to 0.5% of what Wilson pays them, falling with volume.

Sources

Let's review the carrier preferences your team keeps in its head. You don't need to prepare anything: Cartage maps the workflow with you on the call.

Add Cartage as a preferred source on Google