Decision-Ready: A Competitive-Intelligence Cockpit That Runs on Live Data
A leadership team in a specialty-pharma franchise already paid for a good market-intelligence subscription — the kind that aggregates broker research, transcripts, filings, expert calls, and clinical and regulatory news into one searchable place. The data wasn't the problem. The problem was that a search box is not a decision. Every morning someone still had to go dig, read, connect the dots, and figure out what actually moved overnight and whether it mattered.
Sit on top of what they already own
The instinct in this situation is to buy another platform. I did the opposite. Instead of replacing the subscription, I built a cockpit that sits on top of it and does the one thing the raw firehose can't: turn it into a curated, franchise-specific view of what changed, who's a threat, and what's coming. The pitch was deliberately blunt — you already pay for the data; we make it decision-ready. A daily brief instead of a search. A threat board ranking the competitive set instead of a document list. A radar of dated catalysts instead of a calendar someone maintains by hand.
Live where it can be, honest where it can't
A demo that shows fake data is a brochure. So wherever a credible public source existed, the cockpit pulled it live. The trials board makes a real call to the public clinical-trials registry and renders the actual studies — real identifiers, real status. But a live demo that dies when the venue's Wi-Fi hiccups is worse than no demo, so every live source falls back to a validated snapshot the moment the network misbehaves. The room never sees a broken screen, and nobody has to pray to the demo gods beforehand.
It keeps itself current
The part I'm proudest of isn't a screen — it's that the cockpit stays current on its own. A server-side poller sweeps public sources on their real publication cadence: regulatory filings hourly, clinical-trial changes every few hours, drug-label updates daily. It diffs each sweep against saved state using stable per-item keys and content hashes, so it detects genuine changes rather than re-alerting on noise. The first sweep of any source records a silent baseline — no alert flood on cold start — and only real deltas after that raise an event. Material changes score as high severity and dispatch to the team's chat as an alert; routine ones stay quiet. The service scales to zero when idle and a scheduler wakes it on a cron, so it runs around the clock without running up a bill.
One switch from demo to live
Under all of it is a single data-source interface. The demo runs on a curated, offline adapter with simulated latency, so it's fast, self-contained, and needs no credentials. Behind the same interface sits a live adapter that speaks to the real enterprise API, mapped into the identical domain types — dormant until the client provisions access. Going live is one environment variable; the UI never changes. Persistence is pluggable the same way: a zero-infra local store for the demo, a managed cloud database for production, selected by config. The demo and the real product are the same code wearing different adapters.
Honesty as the credibility play
The temptation with an AI-driven brief is to make it look authoritative. I went the other direction. Every generated brief can reveal the exact API call that produced it — the real query, not a hand-wave. Every estimate that rests on a single source carries a confidence marker and renders a flag in the interface. In a room full of people whose job is to be skeptical of vendors, showing your work builds more trust than polish does.
The real lesson
The easy version of this project was a slide deck full of mockups. The version that actually persuades is one where the data is real, the app monitors itself, and the tool is candid about the edges of what it knows. The demo isn't a picture of the product — it is the product, running, on live data, in the room. That's the whole argument: when the thing works in front of you, you don't need to be sold on it. And the same discipline that makes a convincing demo — real sources, honest limits, one clean seam between what's shown and what's swapped in later — is exactly what makes the production version trustworthy too.