<aside> 💡

Duplicate this page into your own Notion workspace so you can check items off as you go. Use the "Duplicate" option from the page menu.

</aside>

🤖 AI Service Blueprint Review Checklist

What this is: A practical review tool for finding AI-driven behaviour that may be hidden inside generic "system", automation or backstage steps in an existing service blueprint.

Who it's for: Service designers, product and service owners, researchers, delivery teams and anyone reviewing a service where AI already influences an experience or may do so soon.

<aside> 🎯

Use it when: an AI agent recommends, interprets, prioritises, decides, routes, generates or acts anywhere in the service.

</aside>

<aside> 👥

Best used with: the touchpoint owner and the person who would receive the problem if the AI-supported step failed.

</aside>

How to use the checklist

Work through a real blueprint touchpoint by touchpoint. Check an item only when the team can point to something concrete that satisfies it. An unchecked item is not automatically a failure; it is a decision, ownership question or recovery path that may not yet have been designed deliberately.

Use the six sections below to capture where AI is acting, how visible it is, what it can decide, where control changes hands, who owns the outcome and what happens when it cannot continue safely.

1. 👤 Actor — who or what is really acting?

• [ ] We can name whether this step is performed by a person, fixed automation, or an AI agent selecting or generating an action.
• [ ] If AI is involved, we can describe the goal it is working toward here, not only the technical task it performs.
• [ ] We are not using "system" or "automated" as a catch-all label for behaviour that is actually adaptive.
• [ ] We know whether the same input could produce different actions because the agent interprets context differently.

<aside> 🔎

Look for: boxes labelled "system", "automation", "platform" or "back end" that contain more judgement than the blueprint currently shows.

</aside>

2. 👀 Visibility — would people know AI shaped the outcome?

• [ ] A customer can reasonably understand when AI materially shapes what happens to them, where that matters to the experience.
• [ ] Frontline staff can tell when an output or recommendation was AI-generated or AI-influenced where they need that context.
• [ ] If AI involvement is invisible, that is a deliberate and defensible design choice rather than an oversight.
• [ ] The service does not imply that a human made a judgement that an AI agent actually made.