An 18-person professional services firm, doing about $3.5M a year, selling projects worth $40K to $120K each. Three account executives worked the deals, but the founder still checked almost every proposal before it went out, whether he'd worked on it or not.
There was no single place to keep track of what happened on a discovery call. Notes ended up scattered: some in the CRM, some in a rep's notebook, some in a Slack thread the founder happened to see. A new rep learned the job by digging through old proposals and a pricing sheet nobody had updated in a year, and guessing at the pattern.
Half these deals, I never even sat in on the call, and I was still rewriting the quote.
The original ask was simple: an AI tool that turns a template into a finished proposal. We asked for something slower first, two weeks watching real discovery calls before building anything.
That changed the plan entirely. Most of the rework on a proposal had nothing to do with weak writing. It came from details a client said out loud on the call that never made it into the document. A generator wouldn't have fixed that. What the firm actually needed was a reliable way to get what was said on the call into the hands of whoever wrote the proposal.
At first we optimized for speed, proposals went out within two hours of the call. But win rate on those early drafts actually dropped 4 points. The scope was usually right; what was missing was the founder's feel for pricing something unusual. So we made human review a required step, not optional, but kept it quick: a ten-minute check instead of a full rewrite.
The second surprise came from the catalog itself. Two of eleven service packages hadn't actually sold in over a year, but the matching tool kept suggesting them anyway because nobody had marked them inactive. Now any package untouched for 90 days gets flagged instead of recommended.
Speed wasn't the real win here. Finding two dead service lines still being pitched to new clients was.
The system has been live for four months, running on our monthly retainer. Average turnaround from call to sent proposal is now under six hours, down from about two days. Win rate is up 9 points quarter over quarter, though the founder is quick to point out a stronger pipeline this quarter deserves some of the credit too, and we report that honestly rather than take all the credit ourselves. Both junior reps use the system on every deal; the founder still tweaks wording on the biggest contracts, mostly out of habit now, not necessity.
Every month, the founder gets one slide, not a dashboard, not a spreadsheet full of numbers, just the handful of things that actually matter:
That's the whole review. No hunting through tools, no fifteen-tab report, just enough for the founder to know the system is paying off.
Two weeks. One senior engineer, one analyst, your team, a written diagnosis at the end. Most of our long-term engagements began here.