Give AI more to work with

Once you've built the habit of asking better questions, the next limit is what the AI knows. A great question about a pasted fragment only gets you so far. Two things move you forward: giving it more to work with, and turning what works into tools you reuse.

Climb the data ladder

• Paste it in. Copy text or a table into the chat. This is where everyone starts.
• Hand it a file. Upload a PDF, a spreadsheet, or a contract.
• Give it a collection. A folder of twelve months of bills, or a project loaded with your reference documents. Now it can find patterns across documents you'd never compare by hand.
• Connect live sources. Shared drives, calendar, and email, so it works from what's current rather than what you remembered to paste.
• Let it act. It drafts the reply, books the meeting, or files the document, with your approval.

Each step adds context, and the answers become specific to you. Each step also adds access, so the care has to rise with it: start read-only, keep approval on any action, and keep company data inside tools your company has approved.

It's the same principle that makes a customer data platform worth building. Unified data creates value when it's put to work, and governance has to keep pace.

Then build your own clubs

• Save your best prompts as templates. Reuse beats rewrite.
• Give it standing context: a Claude Project, a custom GPT, or a Gemini Gem loaded with your background, standards, and reference documents.
• Turn a task you repeat into a tool. Same input in, same quality out, every time.
• Build something real. Tools like Claude Code, Cursor, and Replit can take you from a plain-language description to a working prototype in hours.

Same rules as any product

When you build, the rules don't change. Pick a problem you actually have, not a clever idea. Start with good data, because the output is only as good as what you give it. And know who the user really is: where they work, what device they use, and whether the first "reader" is a person at all.

I learned the last one the direct way. I built a guest check-in app for a club in about ten hours with an AI coding tool. It worked. It never went live, because the people at the gate don't have a computer. AI made building fast. It didn't tell me I was solving the wrong problem. That part is still the product manager's job.

Next in the series: From the Range to the Course →

Want your customer data working this hard for your business? See CDP & Data Product Strategy, or book a 20-minute call.