Workflow · Journey and blueprint
Where AI acts and where a person takes over
Real work Rebuilt from a Claude plugin I built for a content team
The research says merchants find the rules only when something goes wrong. This workflow maps that moment as a journey and a service blueprint, then designs the future with artificial intelligence (AI) agents and people. A person approves the rules for when the agent hands over, before anything is prototyped.
Try it
- Select Map the journey.
- Switch between today and the future.
- Approve the hand-over rules, and the workflow builds a prototype conversation.
Not run yet
What this workflow cannot tell you
- Every customer's path. A journey shows a typical route, so it is checked against real cases.
- How people feel, exactly. Feeling scores come from the research quotes and are a rough guide.
- Whether merchants trust the agent. The prototype needs testing with merchants before anything is built.
Safeguards
What keeps the workflow safe to run on a real knowledge base.
- The agent never decides an outcomeDecisions about money and disputes stay with a person. The agent informs, suggests and drafts.
- A person approves the hand-over rulesNothing is prototyped until a person has agreed when the agent must hand over.
- The agent says it is an AIEvery message from the agent is labelled, and the merchant can ask for a person at any time.
- Nobody repeats themselvesWhen the agent hands over, the person gets the case and the whole conversation.
- It only uses what it is givenThe agent reads the case, approved articles, and records the merchant chooses to share. Nothing else.
- The rules are testedTest conversations try to make the agent break each rule before any customer sees it.
Take it to the next level
How to get results that match the best teams, and what to build next.
- Test with five merchantsPut the prototype in front of five merchants, and watch where they hesitate or ask for a person.
- Try to break itRun a set of hostile and tricky test conversations on every change to the agent.
- Measure the hand-overTrack how often people repeat information after a hand-over. It should be close to zero.
- Map the team's journey tooDraw the risk team's journey beside the merchant's, so the agent helps both.
- Keep it currentRefresh the journey each quarter from the insight library and the latest support themes.
- Design for the merchant's own AIMake the case page clear enough for a merchant's own AI assistant to read and explain.
Rebuilt from a Claude plugin I built for a real content team. Keystone is an invented payments product, and every name, figure and conversation here is made up.