How a Tiny Team Ships Like a Big One · The Product Track
44 use cases we never wrote
A sales team can't pitch what it can't enumerate. We mined public case studies - a customer's own and their competitors' - into a structured use-case library, and ranked 261 accounts against it.
A question came up repeatedly with one of our customers - a company whose technology serves the world's largest banks: what, exactly, can a rep pitch to a specific buyer? Not the product's capabilities - everyone can recite those - but the use cases: the specific, named problems it solves for a fraud team versus a compliance team versus a data platform group. The knowledge existed, scattered across veterans' heads and a decade of collateral. As a list a rep could work from on a Tuesday call? It didn't exist. Nobody had three months to write it.
So nobody wrote it. We mined it.
The library nobody had time to write
The raw material was public the whole time: the customer's own case studies, and - this was the unlock - their competitors'. A competitor's case study is a gift that keeps on giving: it documents a real problem, at a real institution type, solved with your category of technology, in language a buyer already validated. Their marketing department did your discovery for you.
The pipeline reads that collateral and synthesizes it into a structured library: each use case with the problem it addresses, who buys it, what evidence marks an account as needing it. For this customer, that produced 44 graph-analytics use cases spanning fraud detection, anti-money-laundering, entity resolution, KYC. For another customer in a different market, the same pipeline produced 107. The pipeline is the same; the corpus is what changes.
Then the human pass - and this order matters. Domain experts edited the library: merged duplicates, killed a few the machine over-split, sharpened names reps would actually say out loud. Editing a concrete list of 44 is an afternoon of expert time. Authoring from a blank page is the three months nobody had. The machine did the reading; the experts did the judging. Nobody did the typing.
A library is only useful pointed at accounts
The second half closes the loop: the library became the rubric for ranking 261 target companies by fit and intent. Which of these institutions shows evidence - from the public exhaust - of needing use case #17? Each account scored against each use case, with act-now versus monitor signals attached, feeding directly into whose door the reps knock on this quarter and with which story.
That last clause is the point of the whole system. "We ranked your accounts" is a dashboard. "Pitch this use case to this account because of these three pieces of evidence" is a Tuesday morning.
Real numbers
- 44 use cases synthesized from public case studies for one customer; 107 for another - same pipeline, different corpus.
- 261 target companies scored against the library, with act-now vs monitor signals.
- One afternoon of expert editing versus the months of authoring nobody was ever going to do.
- Zero proprietary data required to bootstrap - the corpus was public collateral, theirs and their competitors'.
Where the humans sit
Experts own the library. The machine proposes use cases; people merge, kill, and rename them, and the edited library is the version of record - the ranking runs against what the experts approved, not against raw model output. The same shape as our spec gate: generation is cheap, so human judgment moves to the artifact worth judging.
Steal this
Your competitors' public case studies are the cheapest market map you will ever acquire - a corpus of validated problems, named buyer types, and pre-tested language, free for the reading. Whatever you sell, run the exercise: collect the top competitors' case-study pages, extract problem-buyer-evidence triples, deduplicate, and hand the result to your most senior seller for one editing pass. The deliverable isn't prose - it's an enumerable list where there wasn't one. Enumerable lists are what machines (and Tuesday-morning reps) can act on.
On Thursday the build track gives Claude hands: MCP, from internal tool to product feature. Then back here next week: the money attached to those deals that isn't revenue - the maze of AWS's 80+ funding programs.
This post is part of the product track of How a Tiny Team Ships Like a Big One, a series on how six builders run a production AI company. Building at Aithon - if this is how you want to work, talk to us.