👋 Hey {{first_name|there}},
The board asked for AI, so there's a chatbot penciled in for the corner of the product by Q4. Before you burn a quarter on it, here's how to tell whether it'll move anything that matters.
📉 Why this matters
The last three issues were about AI inside your delivery process: how it shoves your bottleneck downstream (Lesson #56), the instability tax you pay for pretending it didn't (Lesson #57), and the three numbers that tell you whether any of it is real (Lesson #58). This one turns the camera around. Not AI in how you build. AI in what you ship.
The failure mode here is different, and it costs more. Most teams choose their first AI feature based on what will look best in a demo. The trouble is that the feature easiest to show off is rarely the one that moves retention, or expansion, or any number that gets reported upward. I have watched a team ship a polished assistant nobody asked for while the useful idea sat a row below it in the same backlog, unbuilt, because it didn't photograph as well.
Lead with what the model can do instead of what the customer needs, and you win the meeting. Then you spend the year proving the win didn't matter.
🧭 The shift
From: "What AI feature could we build?"
To: "Which number are we trying to move, and is an AI feature the cheapest way to move it?"
That second question kills most of the ideas in the room. Good, that's what it's for. You're not after the most impressive feature, you're after the one where the upside is real, and the build is honest about what it'll take. There are fewer of those than people expect.