Why start with the work
A lot of AI adoption is still licences and a training session, and a few months later the tools mostly polish emails. The tools and the people are fine. Nobody looked closely enough at the work.
Discovery studies an organization’s work the way product research studies a customer: what actually happens, which often differs from what the process documents say.
What we look for
Opportunities tend to take a few recognizable forms:
- Repetition. The same task many times a week: reports from the same sources, data copied between systems, routine documents drafted from scratch.
- Handoffs and waiting. Work stalled on approvals, information requests, scheduling or sign-offs.
- Re-keying. Information typed into one system, then again into another that doesn’t talk to it.
- Unanswered questions. Things nobody asks because the answer would take a day. What did customers complain about most this quarter? Which projects are at risk?
- Workarounds. The shared spreadsheet, the personal checklist, the inbox rule, the sticky note. Each marks where the official systems don’t fit the work.
- Expert bottlenecks. Work only one or two people know how to do, queued up behind them.
- Judgment calls that follow a pattern. Quick decisions experienced people make from many past cases, clear enough to describe and support with tools.
How we do it
Discovery is hands-on. Methods usually include:
- Conversations with leaders and with the people who handle the tasks every day. Our AI discovery interviewer reaches more of them than we could in person, then synthesizes across every interview.
- Observation. The tools people switch between, the steps they skip, the shortcuts they’ve invented.
- Following one piece of work end to end: a request, order or report, through every system and person it touches.
- Reviewing the systems. What software is in use, what data it holds, and where it connects or doesn’t.
- Measuring. Rough numbers on how long things take and how often, to estimate where change would matter most.
Discovery is scaled to the organization and the question.
Watch how work arrives
Anyone who claims to know exactly which parts of a job AI will take over is guessing. Watching how the work arrives is how to find out.
Repetitive, rule-based requests are candidates for automation, with a person checking the output. Requests that need judgment call for better tools: a grounded assistant, smarter routing, a good first draft. Discovery also shows where AI would be a poor fit, before money goes into a build.
What people worry about
Someone studying your work and talking about AI raises a fair question about jobs, so we say plainly what we’re doing and why, and the people doing the work help decide what changes.
What discovery produces
- A map of how the work actually moves, with its pain points and workarounds.
- Opportunities sorted into work to automate, work to upgrade and things that weren’t possible before.
- Rough effort and value estimates for each.
- A recommendation for where to start.
- Often, one or two small working tools that prove the value quickly.
We build those tools early on purpose. Once a team has used one, the next conversation gets much more specific.