Every agency has been told to do AI. Almost none has been told what that means for the data it already holds.
I spent six years inside government leading WA Health's data function: linkage, analytics, engineering, governance and sovereign AI. I know what a Cabinet-ready case looks like, what an audit finds, and why most AI pilots quietly stall at the data layer. For a year of it I also held the research and innovation brief, and the fund behind it, so I know what research needs from a system and how rarely anyone asks. Now I do that work for agencies, at fixed scope and fixed price, on the outside.
Your data people are the foundation for AI. Most of them haven't been told.
The agencies that get real value from AI in the next three years won't be the ones with the best pilots. They'll be the ones whose data someone is prepared to be believed on. Three things decide that, and none of them is a model.
Fixed scope, fixed price, quoted to fit the way agencies buy. The second opinion is priced on the page and sits under every threshold. The rest are quoted once I've seen the brief. These are the six I'm asked for most, not the whole list; the rest is further down. I take one or two engagements at a time, so you'll hear quickly whether it's one for me.
An AI strategy, a business case, a vendor proposal, a platform roadmap. I read it the way I read them for six years on the approving side, and tell you what I see and what I'd do. In writing.
One day · $2,000 + GSTBring me the brief. I interview the people who actually do the work and tell you what you're really solving, whether AI has anything to do with it, and what to build first.
Two weeksA working thing in weeks, not a discovery phase in months. Built at the keyboard with agents, with your team beside me so they own it when I leave. The shape the AI Investment Fund is asking for.
Four to eight weeksGetting data across agency lines and through the governance that guards it. Linkage, data sharing agreements, the DAT Act, PeopleWA. I led the directorate that runs the state's linkage service, and I've sat on both sides of the request. Permission is the design, not the paperwork.
Four weeksWho owns what, who grants access, what governed means in practice, and how business units keep their autonomy. For a data platform, a data function, or the AI capability you've just been asked to stand up.
Four weeksWhen the brief is "we need it in days". I've led a statewide pandemic build on a platform the state already owned. Calm, fast, and it stays up.
DaysIf you carry research or innovation for a health service, a fund, a bank or a registry, you hold the same three problems as the agencies above, with one difference. The systems your work runs on are bought, built and retired by someone else, on their timeline, and research is usually a request after go-live rather than a customer of the build. The job is getting the research case written down in the program's own language, early, by someone who has sat on both sides.
I've been the data side of that conversation for six years, and the research and funding side of it for one. I know what the program office will accept, what the research office needs, and how far apart the two vocabularies are. Three examples of how it arrives. There are more.
A health service is choosing a system that will hold every patient record for the next twenty years, and nobody has written down what research needs from it: finding the cohort, the consent flag, the extract, the feed to the study database and the linkage service. I write that paper in the program's own language and carry it in, so research is a requirement rather than a retrofit.
Four weeksA governance or ethics system that is years past its life, holds every approval the organisation ever gave, and is being replaced by a national one on somebody else's schedule. What's in it, what the new one will and won't hold, what happens to the history and the applications in flight, and who owns the gap in between. A plan before the migration becomes the incident.
Two to four weeksOrganisations outside the health system that hold something research needs: a tissue bank, a registry, a biobank, a private hospital's data. And research groups who know their science but not the data, the governance or the funding profiles. I translate between them and the system, and I know which door to knock on, because I've sat behind most of them.
A standing arrangement · Scoped after a lookA national program gets endorsed. Then the other party's share doesn't arrive, or arrives two years late. The default is to pause and wait for the next round. By then the case is stale, the people who carried it have moved on, and another state has built the thing and is being asked to present on it.
The work is already scoped and already endorsed. The real questions are what the standalone version looks like, what it's worth on its own, whether your own share will fund it, and how you say all that so it reads as a decision rather than a retreat. I've written that case, and I've sat on the side it had to convince.
The version you build yourself is often the better one, because you're no longer designing for the lowest common denominator across nine jurisdictions. Someone still has to keep it moving while that's being decided. That's usually the part that's missing.
Directors and executive directors of data, digital and ICT. CIOs with an AI mandate and no data strategy underneath it. Program leads with a fund deadline, or a co-funding partner who has just walked. Data linkage units. Whoever carries research and innovation for a health service, a fund, a bank or a registry, and depends on systems chosen by somebody else. Anyone who's been asked "what's our AI plan" and knows the honest answer starts with the data.
Agencies looking for a body-shop or a twelve-month transformation program. I don't do either. If you need thirty people, I'll tell you that in the second opinion and save you the engagement.
Jahan is one person, and that's a design, not a risk. Perth-based, on site when it matters. Fixed price per engagement, scoped in writing before we start. Everything I build stays with your team, documented, with no dependency on me.
Jahan was wound down in 2019 when I joined WA Health, and relaunched in October 2026 when I left. WA Health work is published here only after my last day, and always with the code of conduct in mind. Ask me about anything you don't see.
The six above, and the three for research, are what I'm asked for most. They aren't the boundary. Same person, same way of working, on anything where the data underneath is the real job.
Send it anyway. If it's not one for me I'll say so, and usually say who it is for. Send me the brief.
Not an agency, or not sure which page you belong on? Send it anyway. Or message me on LinkedIn. Perth, Western Australia.