When we kick off an enterprise AI engagement, the CEO usually has the big vision. Autonomous agents. Customer-facing personalisation. AI-generated proposals. Big, ambitious, on-the-keynote-slide stuff.
We almost always start with something else. Boring. Internal. A document search tool. A meeting summarizer. A ticket classifier. Something that 50 employees will use every day and nobody outside the company will ever see.
The reason is not that the big vision is wrong. It might be right. But executing the big vision first is the highest-risk path. It requires all the data to be clean, all the integrations to work, all the guardrails to be in place, and the organizational change management to be perfect. Any one of those failing blows up the rollout.
Starting with a boring internal use case lets you build the foundations under lower stakes. The data pipeline gets built. The access control gets figured out. The evaluation framework gets set up. The team learns how to operate an AI system in production. When those pieces are in place, the next use case is half the work, because the foundation is already there.
The second reason to start boring is credibility. If the first AI thing the organization ships is a customer-facing agent that hallucinates, everyone will be skeptical of the next thing for the next two years. If the first AI thing is a search tool that saves everyone 15 minutes a day, everyone becomes an advocate, and the next thing is easier to ship.
We have had multiple clients where the first "boring" use case ended up generating enough internal value that it paid for the next three projects. Nobody was going to put a ticket classifier on the corporate keynote, but it made the support team 30 percent more efficient, and the business case for the bigger investments wrote itself.
Boring first, ambitious second. That order almost always wins.