Use

A brief, then the four core activities of service design

The approach follows the four core activities described in This is Service Design Doing (Stickdorn, Hormess, Lawrence and Schneider, 2018), and adds to each what is missing to decide where AI goes.

The approach

Brief1.Framing the briefWhat result, for whom, within what limits?
Research2.Demand analysisWhat share of demand do we create ourselves?
Research3.Current journey mapWhere is the friction for the person?
Research4.Service blueprint and flow layerWhere does work wait, and why?
Research5.Synthesis: the key insightWhich step limits everything? Is it a rule?
Ideation6."How might we…?"What can change without AI? Where does AI help?
Ideation7.Decision matrixWhich ideas pass the five criteria?
Prototyping8.Rehearsal, Wizard of OzDoes the idea really relieve the bottleneck?
Implementation9.Rollout and iteration reviewDid throughput move? Where will the bottleneck go?

Three moments make the difference

Demand analysis, up front

For two weeks, the team classifies every incoming request: value or failure. Failure demand is the demand we create ourselves: the "where is my file?" follow-up, the wrong form, the call never returned. In the public services he studied, John Seddon often found 40 to 80%. It is eliminated at the source. It is not automated.

Ideation without AI first

The synthesis produces a key insight: the bottleneck. Ideation starts there: "How might we free up this step?" Look first without AI: remove from the bottleneck anything that does not need it, send it only complete files, change the rule if the constraint is one. AI costs money and attention: it comes after.

A prototype before the tool

Before buying or building anything, test the idea Wizard of Oz style: a person plays the role of the AI for a few weeks, without any tool. Measure whether the bottleneck frees up. If it does, implement. If not, you lost a few weeks rather than a budget.

Before the first workshop: the flow log

For one week, each step notes four moments for each case: when it arrives, when work starts, when it ends, and whether it comes back (incomplete, wrong, follow-up). One line per case, no more.

The AI Canvas friction log records what irritates. The flow log records what waits. They work together: an irritant is not always the bottleneck.

Case	Step	Arrived	Started	Finished	Came back?
S-112	Qualification	Mon 9:00	Mon 10:00	Mon 10:40	No
S-112	Writing	Mon 10:40	Thu 2:00 pm	Thu 4:00 pm	Yes, price to confirm
S-113	Qualification	Mon 11:00	Tue 9:00	Tue 9:30	No

A typical project

1	Brief and desk research	Brief: goal, scope, context
1 and 2	Demand analysis	Demand register
3	Workshop 1: journey map	Current journey map
4	Workshop 2: service blueprint	Blueprint layers 1 and 2
5	Workshop 3: flow layer and synthesis	Layer 3, key insight
6	Workshop 4: ideation and decision matrix	Idea portfolio
6 to 8	Change the rules, without AI	Rules changed, baseline measure
9 and 10	Prototyping, Wizard of Oz	Tested prototype
11 to 16	Implementation and iteration review	Dashboard, next bottleneck

Indicative durations, adjusted to the size of the organization.

What the team gets

The project brief: goal, scope, context.
The demand register, classified as value and failure.
The current journey map and its personas.
The three-layer service blueprint, with the key insight: the bottleneck, its type, the rule involved.
The decision matrix and idea portfolio, with the order of action.
A tested prototype, and its stopping criteria.
A five-measure dashboard.
A forecast of the next bottleneck.

What the approach requires

A leader who carries the project, and who agrees to be mapped like everyone else.
The whole team at the four three-hour workshops, not just managers.
Two to five people who live your service, to build the journey map with them.
A frontline worker acting as guardian of the relationship, with a veto on criterion 5.
The person responsible for personal information, for the assessment required by Law 25.