We want to experiment with AI to increase our team's capacity, but our projects always drift and never get completed. How do we structure these experiments as EOS Rocks?

Question

Tyler Smith · Accepted Answer

AI projects often fail because they are too broad and lack clear boundaries. To prevent this, you must structure every AI experiment as a specific, measurable 90-day Rock.

Start by defining what done looks like for the Rock. Do not write a vague goal like "implement AI in marketing." Instead, make it concrete: "Pilot AI copywriting tool for email campaigns and increase production by twenty percent by end of quarter."

Assign the Rock to a single owner who GWCs the seat. They are fully accountable for the outcome, even if they use external resources to help execute it.

Next, break the Rock down into weekly milestones that can be tracked in your weekly Level 10 Meeting. If the project starts to drift, it will show up on your meeting scorecard or as an issue, allowing the team to IDS it immediately.

Keep the scope of the pilot small. The goal of a 90-day Rock is to prove or disprove a hypothesis quickly. If the pilot succeeds, you can create a new Rock next quarter to scale it. If it fails, you cut your losses and move on.