SOPs and ADRs: the context your AI agents need is already written

You spent weeks designing the measurement plan. Event taxonomy, naming conventions, triggers, dataLayer push. No loose ends. Then, one week before launch, everything you’ve done… POOF!

Your manager texts you saying a HiPPO has a better idea, or your teammates have another idea that will improve the measurement. Your work gets replaced — and nobody writes down why. If you’ve been in this role for a while, you know exactly what I’m talking about.

Do you want to save some headaches? The engineers have the solution: SOPs and ADRs.

• SOPs (Standard Operating Procedures) capture how we do things.
• ADRs (Architecture Decision Records) capture why we made a specific decision — the business context, the constraints, the alternatives we rejected.

Engineers write these down. Analysts rarely do. Now I say to myself: how could I have been so blind and not done this sooner? After that, everything seemed flawless. No more shrugs. No more Slack threads nobody could find. No more decisions being reversed because the original reasoning existed only in someone’s head.

Don’t take it personally — your measurement plan didn’t fail. It was never prepared to adapt, to evolve, or to adjust to the company’s needs.

And now the AI part: why all this? Well, those same documents also help AI accomplish tasks better and faster. When I started working with MCPs, CLIs and custom agents, I realised how critical SOPs and ADRs are for getting the best reply out of a prompt. The agent knows not just what to do, but why we do it this way. Fewer corrections. Less back-and-forth. Tasks done right the first time.

Document how the project ended up. Document why a decision was made instead of another. I know it sounds like double the effort — but trust me, it will save you a lot of time on the next implementation. And on the one after that.

Have you ever had to throw out everything you planned one week before launch? Did you also think documenting was a waste of time?

#analytics #webanalytics #dataanalytics #gtm #googletagmanager #ai #documentation #datadriven