Why the first project should be boring

When an organisation first talks about AI, the conversation turns to the visible things. A chatbot on the website. A voice that answers the phone. Something that looks like the future. Our advice is nearly always the same: make the first project boring.

Customer-facing AI makes its mistakes in front of customers. The chatbot says something wrong, someone takes a screenshot, and a reputation built over years takes the hit, while the team has no experience yet of looking after an AI system. Boring internal projects fail in private. The tool that reads invoices misses an odd one, someone in accounts catches it, and you adjust. Nobody outside notices. You learn quickly and cheaply.

Visible projects also need better information than most organisations have. A customer-facing assistant has to know your products, your policies, your records and what is happening today. Most of that is scattered, half written down, and was never meant to be read by a machine. Internal projects work on documents your team already knows well: invoices, contracts, meeting notes, reports.

And visible projects are hard to measure. When a website chatbot works, what does that mean? Happier customers, fewer emails, more sales? Few organisations ever manage to say. When an internal task gets lighter, the measure is obvious. The weekly report takes one hour now, not four. Anyone who signs off the budget understands that without a slide deck.

Boring looks like meeting notes turned into a list of actions with names and dates. Monthly reports drafted from raw figures, then edited by a person in half an hour. Invoice details read from PDFs into the accounting system. An inbox sorted into reply today, reply this week, and no reply needed. First drafts of proposals from a template and a few notes. Each can be tried within weeks. Each gives back real hours. Each teaches the organisation something it will need later.

The test is one question: if this project fails quietly, what do we lose? If the answer is some time and a little money, you have chosen well. If the answer is our customers’ trust, choose something else first. Boring first, visible later, if at all. That order is often the difference between an organisation that grows into AI and one that spends a year apologising for it.