Back to news
Field notesApril 15, 20263 min readBlogPosting

Why GovTech startups shouldn't call themselves "AI Companies"

In this article
3 key takeaways

Calling yourself an "AI Company" raises more concerns than it does any good, and should be avoided if possible.

The AI boom has made it possible to build things in months that used to take years. That's a great thing, but it also means there's a lot of vaporware out there with people chasing cool ideas without the depth to build them properly or safely.

Because of that, governments have been absolutely flooded with half-baked "AI for ___" companies over the last few years, and many of them have been burned by those products.

Now you're not like that, and your product actually works. But the people you're talking to don't have the context to easily tell the difference between what's real and what isn't. Most don't fully understand what AI is, how it works, or how to evaluate it, and more importantly, they've seen it cause tons of problems. Many public workers just know that PII is bad and that AI got a lot of their peers a few cities down in a heap of trouble. Calling yourself an AI Company is cool and sexy, but at the moment doing so creates an air of uneasiness around your product that it does not deserve.

Sell the output, not the model

So your job isn't to talk about how cool your LLM is or explain how your system works. It's to prove that it does work, that it's reliable, that you understand exactly where it fits into their operations without introducing new risk, and that you're comfortable talking about that risk.

Mention "AI" only once or twice if possible, be willing to go to great lengths to defend it if needed, and only go into great depth if they ask. Be willing to teach if they're interested, but don't lead with it.

Your product is the output. Your product is the value created by that output. Talk about that.

Continue reading
8.76.44.1
Field notes4 min read

Why the city you're pitching hasn't already fixed the problem

Most municipal stakeholders already know about the problem you're solving. Something has stopped them from fixing it. Your job is to find out what.

Read article
Field notes4 min read

Three ways to build credibility before you have your first government case study

If a public case study is what unlocks the first sale, how do you make the first sale without one?

Read article
Cookies & Privacy

We use one essential cookie to remember your privacy choice. Optional analytics from Google Analytics and Vercel Web Analytics stay off until you opt in. If you allow them, Sentry may also record masked session replays of opted-in visits to help us reproduce interaction bugs. Sentry error and performance monitoring run by default as a legitimate-interest signal to keep the site stable and secure.

Advertising and cross-site data stay disabled. Review privacy controls.