Note
Agent Experience: designing for the agent, not just the user
Aug 2026
A new audience
For years we designed products around the user experience. Today a second audience shows up: the AI agents that carry out tasks inside the operation. If an agent can’t find the business rules, if the data lives in an Excel file with three versions, or if nobody defined what it can and can’t do, the agent fails just like a new hire with no onboarding.
What it includes
Building systems like Vale AI and BrandOS left me with a short list of what an agent needs to work well:
- Explicit rules. Bonuses, campaigns, brand tone: written, versioned and accessible — not living in someone’s head.
- Data with an owner. One source of truth per data point, with someone accountable for keeping it current.
- Boundaries of action. What it decides alone, what it proposes, and what it escalates to a person.
- Traceability. Every action logged so it can be audited and improved.
Why the board should care
Agent Experience isn’t a technical topic: it’s operational governance. Companies that solve it first turn AI into installed capability; the ones that don’t pile up pilots that work in the demo and break in production.