AI transformation
Moving from AI experiments to business outcomes
I am not an AI researcher. I am a product and technology executive who builds with these tools and has to decide where they belong in a real software business. Most of the hard questions are about product, economics and how people work, and only some of them are about models.
What I think about
- Where AI-native product opportunities exist, and where a copilot bolted onto the old product is not enough.
- Agentic workflows, and which steps are safe to hand over.
- AI-assisted software engineering, and what it does to team shape and review habits.
- Data and platform readiness, which usually decides the timeline.
- Human oversight and governance that people will actually follow.
- Product economics, because inference costs change margins.
- Organisational redesign, so the work changes and not only the tooling.
- Moving beyond isolated copilots to changes in how the company runs.
How I approach it
I begin with the work, not the technology. Pick the product or operating workflow where a change would matter to a customer or to cost, build a small version quickly and measure it against what happens today. Much of what I know here comes from building with AI through ByteJam, so I can tell when a demo is hiding the hard part.
Speed has become cheap, and that changes what is scarce. I have written about this in a few places, including why speed is not the moat and why customers never bought software to use software. The durable parts are judgement, distribution, trust and a clear view of the customer outcome.
I also keep the people in view. AI changes jobs. The leaders who handle that openly, with a plan for the people affected, get better adoption and keep the trust they will need for the next change.
Where I have done this
What I have written about this
Related
Let’s talk
Tell me what isn’t moving.
No pitch deck required. A little context about the business, the constraint and what you have already tried is a good place to start.