---
name: playbook-engagement-kickoff
type: playbook
description: >-
  The prescribed sequence for kicking off and running a Sonar engagement — lock
  the spine and actor, model the audience and where we win, map content and
  belief, blueprint the service, synthesise opportunities and gaps, prioritise,
  and plan. Read and follow this whenever an engagement starts, so the work is
  repeatable and consistent across clients and every step pulls the right
  convention, standard, and skill. Composes Sonar's governance and skills; the
  Finding → Theme → Opportunity → Initiative ladder is the common currency
  throughout. Surfaced by the a1.Muse engagement.
---

# Engagement kickoff playbook

The spine of a Sonar engagement: the order in which the work is done, what to pull at each step, and which skill runs it. A playbook is read and followed, not run — it composes governance and judgement and calls the skills, shedding any step into a skill as that step is automated. Strong candidate to also render as an A1 practice playbook for client-facing use; the canon lives here.

> *This is the canonical copy of this playbook. Any human-readable A1 client version is a downstream rendering — edit here.*

## When to use

At the start of any engagement that needs to go from "we've been hired" to "here is prioritised, evidence-backed direction". The sequence is consistent; the depth flexes to the engagement.

## The common currency

Every step produces or climbs the **Finding → Theme Summary → Opportunity → Initiative** ladder (`convention-audit`). Whatever the step, its output lands on a rung of that ladder, so the engagement accumulates one comparable body of work rather than disconnected deliverables.

## The sequence

1. **Lock the spine and actor.** Agree the engagement's scope, the primary actor (whose experience we're improving), and the stage spine the work will hang on. *Pull:* `convention-user-modelling` for the actor; `convention-audit` for the spine. *Decide:* quality audit vs conversion diagnostic.
2. **Model the audience and where we win.** Produce the needs-based Segment, Persona, JTBD layers, and Forces; state where this venture wins for this actor. *Run:* `skill-user-modelling`. *Pull:* `convention-user-modelling`.
3. **Map content and belief.** Lay the user's beliefs and the content/messages they meet against the stage spine — what they need to believe to progress, and whether the experience supplies it.
4. **Blueprint the service (the diagnostic).** Build the service blueprint and, for funnel work, the conversion diagnostic — connect front-stage moments to the stack behind them and locate the leaks. *Run:* `skill-service-blueprint`, `skill-conversion-diagnostic` (or `skill-discovery-audit` for a quality object). *Pull:* `convention-audit`, `convention-user-modelling`.
5. **Synthesise opportunities and gaps.** Cluster findings into Theme Summaries with root cause and HMWs; reframe as Opportunities; mark the known/assumed/unknown gaps. *Pull:* `convention-audit`, `standard-prioritisation` (confidence overlay).
6. **Prioritise.** Score and sequence the opportunities into Initiatives; split the research-first track. *Run:* `skill-prioritise`. *Pull:* `standard-prioritisation`.
7. **Plan.** Turn the prioritised Initiatives into a sequenced plan with owners and the research track called out, framed for the client's decision.

## Boundaries

Sonar diagnoses and specifies. Hand visual production to Pixel, GTM execution to Reach, and build to Forge. The playbook produces direction and a plan; it does not implement, and nothing commits without Aled's (or the client's) decision.

## Quick checklist

- [ ] Spine and actor locked before any modelling?
- [ ] Why (JTBD, Forces) settled before How (blueprint, plan)?
- [ ] Every output landed on the Finding → Theme → Opportunity → Initiative ladder?
- [ ] Confidence tagged throughout, with high-stakes unknowns on the research-first track?
- [ ] Prioritisation derived per `standard-prioritisation`, not ad hoc?
- [ ] Handoffs to Pixel / Reach / Forge named where they apply?
- [ ] Tone clean per `cos.tov`?