Choose a moment that happens again.
A weekly plan, a practice session, a familiar order: recurring tasks offer a clear place to explore continuity. The next visit gives you a way to see whether previous context actually helped.
Pick a moment where users already have a goal. An agent is easier to evaluate when the desired outcome is specific.
Walk through the second visit.
Write the first conversation, then leave it behind. Start a new one as the same person. Which details should carry forward? Which should be checked again? What would make the result more useful this time?
The exercise exposes unnecessary questions and stale assumptions before you add more capabilities. It also shows where your existing product needs to receive the result.
Expand from what people use.
Look for the parts of the experience people choose to repeat. Did they finish the task? Did they need to correct the agent? Could they find the result afterwards?
Use those observations to choose the next moment. A personal agent grows more useful when it fits real habits, one interaction at a time.
Thoughts on building personal agents into the products people already use.
More about why we’re building shardr