When Anthropic first approached me about a community role in Wellington, I wrote back arguing for a bigger one, and the core of the argument was a number: roughly ten thousand public sector roles going in this region. I made the case that AI literacy had just stopped being a conference topic here and become a live civic question with a 12-to-24 month clock. Anthropic agreed, and the Wellington Claude Impact Lab was the first test of what treating it that way looks like. This essay is the argument in full.
Why Wellington, and why now
Wellington is New Zealand's capital, and its civic, policy, and technology communities overlap more tightly than almost anywhere. The people writing digital policy, the people running council operations, and the people building software drink the same coffee within four blocks of each other. That concentration is usually described as a vulnerability, one town, one industry, and when the public sector contracts, the whole city feels it.
I think the concentration is also the opportunity. A capital city where government, civic institutions, and builders already know each other is exactly the place where AI capability can move between sectors fastest, if anyone builds the connective tissue. Thousands of experienced public servants re-entering the job market need modern AI skills to land well. The institutions they are leaving need AI capability to do more with smaller teams. Those are the same problem.
Literacy is built, not taught
The standard institutional response to a re-skilling problem is seminars: awareness sessions, prompt-writing workshops, a lunchtime webinar series. Useful, and nowhere near sufficient. Nobody becomes literate in a technology by hearing about it. Literacy is what you have after you have built something with the tool, watched it fail, and fixed it.
That is what the Impact Lab was designed to test. Fifty people, many of whom had never used these tools in anger, spent one day building working prototypes against Wellington City Council's real emergency management problems, with Council staff embedded in the teams. Ten prototypes shipped by evening. Some participants arrived without so much as an account and left having built and demonstrated software in front of the Mayor. One day did more for fifty people's practical AI literacy than a year of webinars, and I say that as someone who has sat through the webinars.
The demand side needs no argument: around 104 people applied for 50 seats within days, and a public sector employer's staff sat in the teams as problem owners. The appetite is there. What is scarce is structure.
What a city-scale response looks like
My recommendations, for Wellington and for any city in the same position:
- Build-days on real civic problems, on a cadence. Not an annual festival; a repeatable format where an institution brings its actual problems and its own staff, and mixed teams build against them. The institution gets prototypes and, more valuably, staff who have now worked alongside AI-fluent builders on their own problem.
- Let the institution own the problem statements. The Council wrote better problems than we would have written for them, and their ownership is why the outputs have a path to adoption.
- Leave public artefacts. Our event's dataset work, catalogued and documented, is public and reusable. Every event should make the next one, and the next city's, cheaper.
- Count literacy as the output. Prototypes are the visible product, but the durable one is people: public servants who have now built with these tools, and builders who now understand a civic problem. Fund and measure accordingly.
The two-year clock
The window matters. People displaced this year are making career decisions now, and institutions running smaller are setting their operating models now. A civic AI literacy effort that arrives in three years arrives after both have hardened. This is the rare case where the cheap intervention, structured build-days, public artefacts, embedded problem owners, is available exactly when it is most valuable.
Wellington has the concentration, the talent in motion, and now a worked example that the format delivers. If you run a civic institution, bring problems. If you build, bring the building. If you are somewhere between, which is where re-skilling actually happens, come to the next one. The question is being answered either way. Better to answer it deliberately.