The AI consultancy engagement that starts with saying no

Most people call an AI consultant already knowing what they want built: a chatbot, an internal tool, an automation. A meaningful fraction of the time, the most useful thing I can do in the first conversation is explain why that specific project is the wrong place to start.

The tell is usually a mismatch between ambition and foundation: a client wants a sophisticated internal AI assistant but their data lives in six disconnected spreadsheets with no consistent structure, or wants customer-facing automation before their manual process is documented well enough for anyone, human or AI, to execute it consistently.

Saying no to the requested project isn't a sales tactic dressed up as honesty: it's the difference between a six-month engagement that produces a working system and a six-month engagement that produces an expensive demo nobody trusts enough to actually use.

The clients who come back for a second project are almost always the ones I talked out of the first idea they brought me. Trust in this work is built by correctly saying no more than it's built by agreeing to build whatever's asked.