Commercialization with AI

AI has changed the kinds of commercialization problems I think are practical to work on.

Many of the underlying questions are old: Who matters? What is changing? Where might demand be forming? Where do we appear to fit? What happened after we acted? What can we learn from the outcome?

What feels different now is the amount of information that can be observed, compared and worked with continuously.

A practical constraint has changed

For a long time, many commercial ideas were possible in principle but expensive to operate. AI appears to change some of those constraints. How much it changes them — and under which conditions — is one of the things I am currently working on.

Some things we currently think we have observed

• More signals do not automatically create better opportunities.
• Several independent signals pointing in the same direction appear more useful than raw event volume.
• Buyer need and supplier fit need to be evaluated separately.
• More capable AI models are not always more useful for routine tasks.
• Commercial learning requires clearer outcome definitions than we initially expected.
• Learning across customers creates governance and evidence problems quickly.
• Operational architecture matters much more than it appears to in a prototype.

These are current findings from our work, not universal laws.

What we do not know yet

How much evidence is enough before a system should change? Which signal patterns generalize? Where should customer-specific learning stop? Which tasks genuinely benefit from stronger reasoning models? How much organizational learning can be captured without creating more work for people?

Current view

The interesting work has increasingly moved away from “What can this AI model generate?” toward “What commercial problem are we trying to understand? What evidence do we have? What changed? And what did we learn?”