An ops platform where agents do the chasing and humans do the judging
Castor Freight
Open the live demo →of quotes produced with zero human touch
70%
median quote turnaround, down from 3.5 hours
8 min
quote volume handled by the same ops team
3.2×
The problem
§ 01Castor's operations team spent most of every day on work a careful reader could automate: pulling rates from carrier portals and inboxes, assembling quotes, chasing confirmations. Quoting speed was the business — and theirs was capped by headcount.
Leadership wanted automation but had been burned before by 'RPA' that shattered every time a carrier changed an email template. They needed autonomy that failed safely.
The build
§ 02We ran an architecture sprint first and agreed the success metric in writing: percentage of quotes produced end-to-end without human touch, at equal-or-better accuracy. Build commitment came after, not before.
The platform runs a small fleet of agents — rate extraction, quote assembly, confirmation chasing — on durable workflows that resume after failures instead of dying silently. Anything below a confidence threshold escalates to a human with every document and decision attached.
The ops team got an observability console, not a black box: every agent action is traced, costed and replayable. Trust was a feature we built, and rollout went team by team as the numbers earned it.
“The escalation design is the whole product. My team stopped being data-entry clerks and started being underwriters.”
Want a result like this one?
Four short questions, then a call with the people who'd actually build it.
We reply within one UK business hour.