An agent-run freight brokerage

What a brokerage sells is not trucks

A freight brokerage owns no equipment. It sells two things a shipper cannot buy directly: coverage, meaning a truck that actually arrives on the booked day, and one counterparty to invoice instead of dozens of carriers. The regulatory footprint around that is thin. Under 49 CFR 387.307 a property broker must keep a $75,000 surety bond or trust fund in force so shippers and carriers have something to claim against when it fails to perform, and a trust fund has to sit in assets that can be liquidated to cash within seven calendar days. Under 49 CFR 371.3 the brokerage keeps a record of every transaction for three years, naming the consignor, the carrier, the compensation it received and the date it paid the carrier, and every party to that transaction has the right to demand a look at the record. None of that requires a licensed individual. The duties attach to the registered entity, which is why an agent-run brokerage is legally unremarkable in a way an agent-run contractor is not.

The load is the unit, and the unit repeats

Every load runs the same arc. A shipper asks for a rate on a lane. The brokerage prices it, wins it, posts it on DAT One or Truckstop, fields the carrier calls and emails that follow, negotiates to a number that leaves margin, and awards the load to one truck. A rate confirmation goes out fixing that number, the pickup and delivery appointments, the free time before detention starts to accrue, and which accessorials the brokerage agrees to pay. The truck is tracked to delivery, the paperwork comes back, and two invoices settle in opposite directions on different terms.

That arc runs thousands of times a year with an identical shape and different numbers. It is the single fact that makes this business agent-shaped, and the vendors found it before the analysts did: HappyRobot, Fleetworks and Vooma all sell agents that take carrier calls and carrier email against a brokerage’s own load list, and Parade sells capacity matching against past carrier behaviour. The interesting claim is not that any one of those products works. It is that the entire distance between a shipper asking and a carrier being paid is machine-legible.

The stack costs less than the seat

DAT publishes its broker prices, which is rare enough in this market to be worth quoting. On dat.com’s load board pricing page, DAT One Express is $159 a month, DAT One Select Broker $319, and DAT One Office Broker $449. Tracking, carrier vetting and a broker TMS sit on top of that. Software has never been the binding constraint in this business and is not the binding constraint now. The object at risk is a full trailer of somebody else’s freight, moving on a document that carries the brokerage’s name.

The fraud desk is the ceiling

Carrier onboarding looks like a checklist and behaves like a security perimeter. The packet is close to standard: evidence of operating authority, a certificate of insurance, a W-9, a signed broker-carrier agreement, and a notice of assignment where the carrier factors its invoices. The federal floor for a for-hire carrier of non-hazardous property rated at 10,001 pounds or more is $750,000 of liability cover under 49 CFR 387.9, and the broker-carrier agreements brokerages publish routinely ask for a million of auto liability with a hundred thousand of cargo cover on top. Every item on that list can be checked by machine against FMCSA records, Carrier411, RMIS or Highway, and an agent will run the checks more consistently than a night-shift coordinator ever did.

The checks are also known to the people attacking them. Double brokering and carrier identity theft work by presenting a real but stolen motor carrier number with paperwork that matches it, so the automated tests pass and the freight leaves on a truck nobody can trace afterwards. What catches that is not another field to validate. It is the pattern nobody wrote down: an authority reactivated last week, a dispatch number registered yesterday, a dispatcher who answers email and will not answer a phone. An agent is better at the checklist and no better at the inference, which is why the fraud desk is the last seat in a brokerage to be automated rather than the first.

Money leaves before money arrives

A brokerage pays carriers on short terms and bills shippers on long ones, so it is a lender whether it thinks of itself that way or not. That gap is why factoring appears on both sides of the transaction, and it creates a settlement trap an agent has to be built around. Under section 9-406 of the Uniform Commercial Code, once a carrier has assigned its receivable and the account debtor has notice of the assignment, paying the carrier no longer discharges the debt. Releasing funds to the wrong party means paying for the same load twice. Assignment status is a field to read before payment, not a note to discover afterwards.

Claims run on their own clock. The obligation is the carrier’s, not the broker’s, but the brokerage is usually the party driving the file. Under 49 CFR 370.5 a carrier must acknowledge a cargo loss or damage claim in writing within 30 days of receiving it. Under 49 CFR 370.9 it must pay, decline, or make a firm compromise settlement offer in writing within 120 days, and then report status every 60 days until the claim is resolved. A brokerage that lets a damage report sit unread in a general queue for a fortnight has spent close to half of that first thirty-day window before anyone opened the file.

What survives the extension

Three seats. Somebody decides whether a new shipper gets open credit, because that decision spends the brokerage’s own cash. Somebody owns the carrier that passes every check and still looks wrong. Somebody makes the call after a loss, when the shipper is deciding about the next hundred loads rather than the one it lost. Everything between those three is a document, a lookup, or a message, and all of it can run unattended.