Before an AI roadmap, ask who is responsible

AI conversations often open with a capability: summarise this, predict that, automate the next step. I enjoy exploring what these systems can do. But the question I want in the room early is less glamorous: who is responsible for the outcome?

A buyer may be thinking about speed or cost. An employee may be wondering how their role changes. A customer may not know that a model will influence a decision about them. If we skip those perspectives, we can build a very efficient version of a poorly understood process.

My interest in this is not new. Years ago I wrote about algorithmic bias, the assumptions of people building AI and the danger of equating technical efficiency with social progress. The technology has advanced considerably since then. That makes the conversation more concrete, not less necessary. I do not want to sell a feature on the strength of a demonstration if the customer has not decided how it will be judged in everyday use.

I start with the business decision: what are people doing today, and what would improve if AI helped? Then the human decision: who sees the output, who can correct it, and whose work or experience changes? Then the operating decision: what data and process support it, how will failure be noticed, and who owns a response when the output is wrong?

Those are not an argument for waiting until every uncertainty disappears. They help us choose a useful first use case with clear limits. A responsible pilot, product feature or workflow should tell people what has been tested, what remains uncertain and where their judgment still matters. The team should be able to describe the value without saying ‘because it uses AI.’

Adoption is not the same as deployment. A customer may buy a capability and still struggle to get people to rely on it—or know when not to. That is why a good AI roadmap needs product, commercial and human accountability in the same conversation.

The roadmap matters. So does the person who has to live with it.

Read related: Beyond Business Outcomes