AI Does Not Have Its Own Intent
AI does not decide what matters on its own. Human intent remains central to the systems we build.
5 min read
Current work / 2025–present
Keeping human intent in the work AI builds.
AI agents can write software quickly. The harder question is whether they are building what a team actually meant. At Nera, that is the problem we are working on.
The work
We are building a shared record of intent for enterprise software teams: the requirements, architectural decisions, standards and context that should guide people and AI agents as they work. The aim is to make that intent explicit, testable and useful throughout development.
This is a continuation of a question I have explored for years: how do we give increasingly capable systems direction, and know whether they are doing what we intended? My earlier writing on narrow AI, governance and human agency connects directly to what we are building today.
My contribution
- Building the company and its product direction with a team of long-time collaborators.
- Connecting requirements and verification to the day-to-day work of software teams.
- Making human judgment, context and accountability part of how AI-assisted work gets done.
Connected thinking
From the writing.
4 Steps to Good Narrow AI
Four practical steps for making intent explicit and evaluating whether an AI system achieves it.
9 min read
Designing for the Humans in the Loop - Part III
How AI design can preserve the judgment and agency of people in the decision or process loop. The final essay in a three-part series.
13 min read
Keep the conversation going
Good questions
lead to useful work.
For conversations about building with AI, responsible adoption, speaking and media.
Get in touch