Agree what the numbers mean before AI reads them
Ask three teams how many active customers you have and you may get three answers. Why that matters more once AI reads your data, and a short way to fix it.
Blog
Data and AI, written plainly.
Founder notes on data, AI tools and agents, and when not to build. Every figure has a source.
-
-
AI speeds up the work, not the decision
AI can draft the report in an hour, yet the decision still waits for the same meeting. Why gains stall at the slowest step, and what to change before you buy.
-
Every report needs an owner, a decision and a review date
Reports pile up faster than anyone retires them. Why that operating debt costs more once AI reads your data, and a simple rule for clearing it.
-
Name the approver before you build the agent
Human approval of AI output is easy to promise and hard to design. Which actions need it, who signs off, and how to stop review becoming a rubber stamp.
-
Start with the decision, not the dashboard
Most reporting and AI projects start with a request for visibility. Name the decision, its owner and its deadline first, and the build gets much smaller.
-
The model is the easy part
Why can't Claude or an agent you already have do this? Often it can do part of it. The hard part is the system around the model. A technical checklist.
-
Will it still run when the builder leaves?
Most data and AI projects are scoped as a build. The real test comes when the person who built it moves on. Five checks to run before you call it done.
-
Your AI assistant is not the hard part
Claude or ChatGPT can often do part of the job. Why company AI pilots stall anyway, what DIY really costs, and what to ask before buying another AI tool.