The Discovery Playbook.

Goal: Land the sprint in the client’s organisation. Confirm access. Align on outcomes. Set the decision log rhythm.

Agenda (2-hour kickoff):

• 00:00–00:15 Introductions. Executive Sponsor opens; Sprint Lead sets frame.
• 00:15–00:45 Workflow context. Workflow Owner narrates “what this thing does, and why it hurts.” No slides. Verbal only. Sprint Lead takes narrative notes.
• 00:45–01:15 Outcomes conversation. Sprint Lead facilitates: “If this sprint succeeded perfectly, what would be different 90 days from now?” Capture 3–5 outcomes verbatim.
• 01:15–01:40 Logistics. Confirm access is live, calendar holds for all gates are set, communication channels are open, shadow sessions are approved.
• 01:40–02:00 Close. Sprint Lead reads back outcomes. Executive Sponsor signs off verbally. Next-steps calendar confirmed.

Afternoon (async): Sprint Lead opens the Decision Log, opens the Risk Register v0, files the kickoff memo with read-back outcomes to Executive Sponsor and Workflow Owner, and drafts the Week 1 interview schedule.

Output of the day: Kickoff memo, outcomes locked verbatim, Decision Log entry 001 (engagement framed), Risk Register opened.

Red flags to watch: Executive Sponsor sends a delegate — stop and rebook. Workflow Owner describes the workflow in generalities (“we, you know, process things”) — push for specifics or schedule a second walkthrough. Access has not landed — file a blocker in the Risk Register before leaving the room.

Goal: Produce a first-pass step inventory. Whiteboard the workflow end-to-end with the Workflow Owner so Studio has a shared mental model before going deep.

Agenda (3-hour whiteboard session):

• Present: Sprint Lead, Agent Engineer, Workflow Owner. No other stakeholders — this is the canonical walkthrough and needs focus.
• Method: Start at the trigger. Walk the workflow forward, step by step. For every step ask: who, what, input, output, decision rules, elapsed time, failure modes. Draw on a whiteboard or in a shared canvas.
• Tool deployed: D-01 Workflow Audit. Enter each step as a row in the spreadsheet during the session — do not retrofit afterward.
• Checkpoint: At the end, Sprint Lead reads the workflow back. Workflow Owner must explicitly confirm or correct each step.

Afternoon: Sprint Lead formalises the whiteboard into D-01, photographs the whiteboard, files both in the shared drive, and posts a short Slack note summarising Day 2 to the client channel.

Output of the day: Workflow Audit v0.3 (all steps inventoried, some attributes still empty), whiteboard photograph archived, Decision Log entry 002 (workflow canonical walkthrough complete).

Red flags: If the walkthrough cannot finish in 3 hours, the workflow is wider than qualified. Document the scope delta in the Risk Register and raise it at the Day 5 checkpoint.

Goal: Triangulate the walkthrough. The Workflow Owner knows what the workflow should be. Stakeholders around them know what it actually is. Listen for the delta.

Who to interview (the default set):

• The person who supplies the inputs — upstream dependency. Learn what the input looks like on a bad day.
• The person who receives the outputs — downstream consumer. Learn what failure costs downstream.
• A peer of the Workflow Owner — sometimes skipped if no peer exists; always interview if one is available, because peers are honest about ambiguity.

Tool deployed: D-04 Stakeholder Interview Guide (lighthouse-stakeholder-interview-guide.docx). Run the full 26-question template. Do not paraphrase; record the verbatim answer. If permitted, record the session.

Output of the day: 2–3 interview note files, a verbatim quote log (one spreadsheet with quote, speaker, context tag, and relevance score), Decision Log entry 003 (stakeholder perspectives integrated).

Red flags: Interviewees describe a fundamentally different workflow than the Workflow Owner did. This is common and critical — raise at Day 5. Interviewees describe the workflow in purely negative terms (“it’s a disaster, I can’t believe we still do it this way”). This is a cultural signal, not a workflow signal — note it, but do not over-weight.

Goal: Watch the workflow happen in real time. Shadow data replaces assumptions with observations. This is often where the most important findings appear — the client has usually forgotten what they actually do.

Method:

• Pair with the Workflow Owner or a designated operator. Ask them to work as they normally would — not to narrate or teach.
• Take dual notes: left column “what happened”, right column “what surprised me.” The right column is the gold.
• Pilot the stopwatch: start tracking D-02 in this session. The full stopwatch pass happens Day 6, but Day 4 is where the timing method is tested and the right atomic units identified.
• Note the switches: Every time the operator switches tools, apps, windows, browser tabs, or systems, capture it. Tool-switch volume is one of the most reliable predictors of agent value.

Tool deployed: D-01 Workflow Audit (fill in the “observed variance” column), D-02 Stopwatch Protocol (pilot mode).

Output of the day: Shadow notes, pilot stopwatch data on 2–3 instances, observed variance notes added to Workflow Audit, Decision Log entry 004 (field observations complete).

Red flags: The operator cannot work as normal because “today is unusual.” Reschedule — do not watch an unusual day. The operator performs a hidden step that was not in the walkthrough (common — this is why shadow exists). Document and raise at Day 5.

Goal: Share what Studio has observed. Let the Workflow Owner correct misunderstandings before they are compounded into the Gate 1 recommendation. Lock the Workflow Audit at v0.5.

Agenda (90 min):

• 00:00–00:20 Sprint Lead presents the Workflow Audit as observed. No interpretation yet — just the decomposition.
• 00:20–00:50 Workflow Owner corrects, adds, or pushes back. Sprint Lead edits the audit live.
• 00:50–01:10 Sprint Lead surfaces the top 5 observed variances (where observation diverged from Day 2 narrative). Discuss each.
• 01:10–01:25 Preview of Week 2: stopwatch, bottleneck scoring, Fit Assessment. Workflow Owner commits calendar.
• 01:25–01:30 Close. Workflow Audit v0.5 locked.

Afternoon: Sprint Lead sends a Week 1 wrap memo to Executive Sponsor with the Workflow Audit v0.5 attached and a short narrative: what we heard, what we observed, where we are so far. Raise any scope or access risks explicitly.

Output of the day: Workflow Audit v0.5 (locked), Week 1 memo, Decision Log entries 005 (variances resolved) and 006 (Week 2 plan confirmed), Risk Register updated.

Red flags: If the Workflow Owner rewrites more than 30% of the Workflow Audit, Week 1 listening was insufficient. Extend Week 2 Day 6 to include a second walkthrough — do not proceed to stopwatch on an unstable base.

Goal: Run the full stopwatch protocol. Measure cycle time, throughput, error rate, rework rate, and human-hour cost across 10–15 instances of the workflow.

Method:

• Sampling: Target 10–15 instances minimum. If workflow throughput is low (fewer than 10/week), instrument historical data from logs or spreadsheets instead of waiting.
• Atomic units: Use the atomic unit list agreed on Day 4. Every step must have a start and stop marker.
• Parallel measurements: Where possible, instrument alongside a self-report from the operator. The delta between observed and self-reported time is itself a finding (usually 30–50% understatement).
• Tool deployed: D-02 Stopwatch Protocol, logged directly into lighthouse-baseline-metrics.xlsx.

Output of the day: Baseline Metrics v0.5 with 10–15 rows, per-step duration distributions, observed error rate, observed rework rate, Decision Log entry 007 (baseline instrumentation complete).

Red flags: The workflow cannot run 10 instances in a day and the historical log is unusable. Reduce sample to 5–7 high-confidence instances and document the limitation in the Discovery Report — do not fabricate synthetic data.

Goal: Turn the baseline numbers into a prioritised bottleneck list. Which steps are candidates for automation? Which are candidates for agent-initiated human review? Which are untouchable?

Method:

• Tool deployed: D-03 Bottleneck Scoring Matrix. Score each workflow step on five dimensions: time cost, error cost, tedium (operator-reported), automatability (Studio-assessed), and risk (Governance Architect-assessed, if present).
• Workshop flow: Sprint Lead pre-populates the time cost column from Day 6 data. Workshop participants score tedium and automatability live. Governance Architect scores risk async.
• Output ranking: Sum to a composite score, sort descending. Top 3–5 rows are the expected candidates for agent redesign.

Output of the day: Bottleneck register (scored), top-3 agent candidates identified, Decision Log entry 008 (priorities agreed), Risk Register updated with risk column scores.

Red flags: The top bottleneck is a step the Workflow Owner considers emotionally “their real work.” This is a common and serious signal — the workflow redesign may conflict with role identity. Raise at Day 8 and flag in the Discovery Report.

Goal: Score the engagement against the Fit Assessment Rubric. Produce the Go / Conditional Go / No-Go recommendation that will be presented at Gate 1.

Method:

• Tool deployed: D-05 Fit Assessment Rubric (lighthouse-fit-assessment-rubric.xlsx). 12 criteria across workflow fit, data fit, organizational fit, commercial fit. Score each 1–5 with evidence.
• Participants: Sprint Lead (primary). Agent Engineer and Governance Architect contribute to data fit and risk scoring. Executive Sponsor is not in the room.
• Evidence discipline: Every score must cite at least one Discovery artifact (a specific interview quote, stopwatch row, Workflow Audit step, or Bottleneck score). Scores without evidence are replaced with question marks until evidence is found.
• Recommendation logic: 36+/60 → Go. 30–35/60 → Conditional Go (specify the condition). Below 30 → No-Go.

Output of the day: Completed Fit Assessment Rubric, draft recommendation, Decision Log entry 009 (recommendation drafted).

Red flags: Your personal read conflicts with the rubric score. The rubric wins — it exists to override gut feel. If you think the rubric is wrong, document the specific criterion you would change and why, and raise it internally before Gate 1.

Goal: Assemble the Discovery Report and the Gate 1 deck. Preview the allocation framework for Redesign so the client arrives at Gate 1 with a mental model for what comes next.

Method:

• Template: Use the Discovery Report template (lighthouse-discovery-report.docx). Eight sections: workflow overview, baseline summary, bottleneck analysis, fit assessment, recommendation, risks, proposed scope for Redesign, commercial implications.
• Allocation preview: For the top 3 bottlenecks, sketch preliminary allocation: Fully Autonomous, Agent-Initiated Human-Approved, Human-Led Agent-Assisted, Human Only. This is a preview, not a commitment — Redesign does the real allocation work.
• Gate 1 deck: 10-slide deck that mirrors the Report’s structure. Deck is the meeting; Report is the record. Deck includes the scored rubric as an appendix slide.
• Circulation: Both Report and deck go to Executive Sponsor and Workflow Owner at least 24 hours before Gate 1. This is non-negotiable — surprises at a gate are sprint failures.

Output of the day: Discovery Report v1.0, Gate 1 deck circulated, allocation preview drafted, Decision Log entry 010 (G1 materials locked).

Goal: Present Discovery findings. Secure Go / Conditional Go / No-Go decision from Executive Sponsor and Workflow Owner. Close Phase 1.

Agenda (90 min):

• 00:00–00:05 Sprint Lead re-reads the outcomes from Day 1. Re-anchors the meeting.
• 00:05–00:25 Workflow as observed. Walkthrough of the Workflow Audit, with 2–3 verbatim stakeholder quotes.
• 00:25–00:40 Baseline numbers. Cycle time, throughput, error rate, human-hour cost.
• 00:40–00:55 Bottlenecks and allocation preview. Where agents help, where they do not.
• 00:55–01:10 Fit Assessment. Rubric score, criterion-by-criterion evidence, risks.
• 01:10–01:20 Recommendation. Go / Conditional Go / No-Go with explicit reasoning.
• 01:20–01:30 Decision. Executive Sponsor signs off or requests changes. Workflow Owner concurs or dissents explicitly.

Output of the day: Signed Gate 1 decision (captured in the Decision Log and in writing to Executive Sponsor), Phase 1 closed, Redesign Playbook activated for Monday.

Red flags: Executive Sponsor defers to “think about it and come back next week.” This is common and acceptable once. Block a follow-up window within 5 business days. If the decision keeps sliding, re-open qualification — this is an organizational-fit signal.