Gilead had built the software they needed over a decade — piece by piece, one micro-app at a time. By the time we arrived, there were more than thirty of them. Nobody was using most of them.
Gilead is a biopharmaceutical R&D company with roughly 15,000 employees. Internal tooling was a layered accumulation of thirty-plus micro-apps on SharePoint, each solving a narrow problem someone had urgently needed to solve three years ago, or seven, or ten. Apps were slow. Security posture uneven. API surface missing. Employees had quietly gone back to email, pen and paper, and the shared drive.
The mandate: a single cross-platform internal app, iOS and Android, mobile and tablet, that people would actually open. Five weeks of discovery. Two on-site in California, three in Bucharest with engineering. Sign-off on wireframes at the end.
I led discovery. In practice: build rapport, reprioritize the feature set against what users actually did, produce the wireframes and user stories that unblocked Phase 2.
1. Reframe from “build a better app” to “find the right thing to build first.” The client walked in with a feature list. My first intervention was to slow that down. A Lightning Decision Jam with Senior Directors and VPs — structured so senior voices weren’t the only ones heard — followed by three days of user interviews across HR, Legal, Medical Affairs, Security, Business Strategy, and Library Services. The top-of-list feature was not the one driving daily friction. One headline was de-scoped from Phase 1; another moved up.
Search — can’t find anything. It’s not a Google. ‘Did you mean…’— user interview, Gilead Sciences, 2019
2. Design for the org that needed to see itself. The real cost of the fragmented tooling wasn’t user time — it was leadership blindness. Senior Directors had no way to see what teams were doing day-to-day without requesting a spreadsheet. The new IA did two jobs at once: surface the actions a field user needed, and produce — as a byproduct — the signal leadership had been missing. That reframing turned an app refresh into an organizational tool.
3. Trade depth for discoverability in the first release. Phase 1 collapsed around a small cluster of high-use workflows done well — search, directory, news — and explicitly planned for the rest in later phases. The architecture was built to extend. The first release was built to be used.
When you post announcements on Gnet about events and notice, you have visibility for only 4 of them. If yours goes below the fold, nobody will ever see it.— user interview, Gilead Sciences, 2019
That quote landed the news-feed decision. Four announcements surfaced; a fifth invisible. The replacement had to handle the long tail without relying on someone posting at the right moment.
Two weeks in California, three in Bucharest. On-site meant workshops and user interviews could happen at stakeholder pace, not at video-call pace. Bucharest was where wireframes got pressure-tested against engineering reality.
Stakeholder priorities are never locked until research shows them something they didn’t know. The LDJ was the moment two senior executives reprioritized a feature set they’d arrived defending — not as a concession, as a discovery. That’s the move I carry into every enterprise discovery since.
And narrower: internal tools make the org visible to itself. A well-designed operator surface is also an intelligence surface.