Twelve-broker shop replaces a spreadsheet with $1/load AI.

The starting point

A twelve-person freight brokerage in the Southeast US, six years old, owner-operated, running ~900 loads per month. Stack at engagement: AscendTMS free tier as system of record, Gmail for the inbox, DAT One for the load board, a shared Google Sheet where coordinators pasted in carrier quotes by hand. No formal carrier vetting tool — MC numbers were spot-checked in FMCSA SAFER whenever a broker felt something looked off.

The pain was specific: every posted load was pulling 30–50 inbound carrier emails, and a broker was spending close to an hour per load reading replies, copying rates into the spreadsheet, eyeballing MC numbers, and deciding who to call back. The coverage rate was a healthy 92%, but volume growth was capped — adding the next 200 loads / month would have required hiring two more brokers at roughly $65K each fully loaded.

Week 1 — connect

• Google OAuth on each broker's Gmail. One label created (Keelway/triaged). Read access only on the carrier inbox.
• FMCSA backfill on the existing carrier book — about 1,400 carriers — to seed trust scores.
• One 45-minute kickoff with the owner and lead coordinator to calibrate ranking weights. Reefer and flatbed lanes weighted differently from dry van.

No floor disruption. Brokers kept Gmail as their working surface. Keelway started ranking inbound replies inside two hours.

Week 4 — calibration

• First measurable change: average time-per-load dropped from 47 minutes to ~14 minutes. The biggest delta was on the long-tail replies — the bottom 30 of the 40 daily replies that brokers used to skim and reject by hand.
• Three fraud-adjacent carriers caught in the first two weeks — email-domain spoofs and one chameleon MC that had reincorporated twice in eighteen months. The owner estimated those three catches alone covered the cost of Keelway for the year.
• Brokers started accepting the top-1 ranked carrier directly from the Gmail side panel ~40% of the time, with the spreadsheet still holding the cumulative log.

Month 3 — the spreadsheet goes away

The shared Google Sheet was retired. Keelway became the per-load carrier-quote log of record, with accepted bookings flowing into AscendTMS via the integration. Time-per-load settled at ~6 minutes, with the broker reading only the top-five ranked replies and spot-checking the trust score on the carrier they accepted.

Coverage rate held at 92%. Margin per load held steady inside a ±2% band, because Keelway's rate extraction surfaced the carriers quoting at or below the broker's target consistently — the high quotes still got declined, but they got declined in seconds, not minutes.

Month 12 — volume up, team flat

Monthly load volume reached ~1,250, up 38% from the engagement baseline. Broker count: still 12. The two coordinator hires the owner had been planning were canceled. Software cost for the year: AscendTMS free + ~$13,500 in Keelway AI fees + DAT One subscription. Same brokerage twelve months earlier was projecting ~$140K in coordinator headcount to cover the same growth.

What this shape buys

Most SMB brokerages of this size deploy the same way. The pattern is:

1. Keep AscendTMS (or whatever lightweight TMS is in place) as system of record.
2. Layer Keelway AI ($1 per load) on the Gmail inbox. No long-term commit, no per-feature add-ons, no implementation fee.
3. Watch time-per-load drop and use the recovered capacity to grow volume instead of adding headcount.

The brokerages in this segment that consolidate further go to Keelway TMS at $997/month flat — one bill, AI bundled, unlimited users. For 12+ brokers that math is usually better than per-seat TMS pricing.