Phase 5 (Implementation) deep-dive playbook for an initiative shepherd. You are not doing the implementation. You enable teams, maintain consistency, ensure the initiative completes, and step back when it does. Time budget: 2–6 months wall clock, 10–20 hours/month of shepherd time. Composes Skill(running-work-transitions) in bitwarden-delivery-tools for the originating-side Support Period, Pulse Check, Retrospective, and Closure phases of the Work Transition Playbook.
The funnel doc's mental model:
Think of the shepherd as:
- A guide ensuring teams stay aligned with the initiative's vision
- A coordinator managing cross-team dependencies and communication
- A subject matter expert available for questions about the approach
- A reporter keeping leadership informed of progress
- Not a project manager micromanaging day-to-day work
- Not doing code reviews on every PR (unless specifically needed for approach validation)
Establishing Communication Channels (Before Implementation Starts)
Set up the coordination mechanisms before teams begin. The funnel doc specifies:
- Optional dedicated Slack channel (e.g.,
#initiative-typescript-migration). Pin links to: PoC PR, ADR, architecture plan, Jira dashboard. Use for questions, blockers, and learnings across teams.
- Bi-weekly tech-leads sync with all affected tech leads. 30–45 minutes. Round-robin: progress, blockers, questions, cross-team dependencies.
- Optional office hours (1–2 hours per week for drop-in questions) or use the company-wide office hours.
- Monthly stakeholder-sync update for engineering and architecture leadership — 15 min, status / completion percentage / risks / decisions needed.
Set response-time expectations explicitly when you announce the channel: "Questions in Slack — aim for 1 business day response. Architecture concerns — same day."
Kickoff Meeting
When teams are ready to begin (capacity allocated, stories in sprint planning), host a 1-hour kickoff with all teams. Per the funnel doc:
- Recap the initiative. Problem, solution, PoC results — for engineers who weren't in handoff meetings.
- Walk through the approach and key patterns. The PoC PR is the best teaching artifact you have.
- Review cross-team dependencies. Make them visible so teams know who they wait on and who waits on them.
- Introduce communication channels and cadence. What to use where; what response time to expect.
- Answer questions and address concerns.
- Celebrate the start.
Share a resources package in the Slack channel: PoC PR, ADR, architecture plan, FAQ (start empty), your availability and response time.
Supporting Teams During Execution
Four activities, per the funnel doc:
1. Answer Questions
- Respond in the Slack channel within ~1 business day.
- Jump on Meets when text doesn't suffice.
- Clarify edge cases or scenarios the PoC didn't cover.
- Provide examples or references to similar implementations.
2. Review for Consistency (Not Detailed Code Review)
This is the most-violated rule of Phase 5. The funnel doc is unambiguous: "Trust teams for detailed code review – only intervene for approach issues."
- Monitor PRs related to the initiative.
- Review the approach in early PRs from each team to ensure alignment with the PoC pattern.
- Provide feedback only on approach deviation, not on style, naming, test coverage, or anything else the team's own reviewers handle.
- The funnel doc's example pattern: "This looks good but uses callbacks instead of the async/await pattern from the PoC — was that intentional?"
The "Not a reviewer for the team's PRs" line from navigating-the-initiative-funnel applies symmetrically here: tech leads expect you not to be their team's code reviewer. Respect it.
3. Troubleshoot Unexpected Issues
When teams hit problems the planning didn't anticipate:
- Help diagnose: implementation issue, or approach issue?
- For approach issues, escalate to Architecture Council if fundamental adjustment is needed.
- Document solutions in the FAQ so other teams can learn.
4. Unblock Dependencies
- Track which teams are waiting on others (via the dashboard and the bi-weekly sync).
- Coordinate communication between dependent teams.
- Escalate to engineering leadership when blockers can't be resolved at the team level.
- Adjust sequencing if the original plan proves problematic.
Maintaining Cross-Team Consistency
The funnel doc calls this "one of the shepherd's most critical responsibilities." Four sub-activities:
- Early detection of divergence. Notice when teams interpret the pattern differently. Spot legitimate variation vs. misunderstanding. Review 1–2 early PRs per team to catch issues before they multiply.
- Share learnings across teams. Post in Slack when one team solves a common problem. Update the FAQ as patterns emerge. Call out good examples: "Team X's PR #567 shows a clean way to handle this edge case."
- Refine guidance when needed. If teams consistently struggle, the guidance may need improvement. Update the architecture plan or create supplementary guides. Host ad-hoc working sessions if multiple teams hit the same issue.
- Make judgment calls on acceptable variation. Some variation is appropriate based on context (e.g., "Mobile apps use variation A because of platform constraints; web should use the standard pattern"). Some variation is drift that undermines consistency. Document the calls.
Tracking and Reporting Progress
The funnel doc specifies multiple cadences:
Weekly (Your Internal Tracking)
- Review Jira dashboard: completed, in progress, blocked.
- Check the Slack channel for unresolved questions or concerns.
- Note risks or trends (e.g., multiple teams reporting the same issue).
Bi-Weekly Tech-Leads Sync
- 30–45 minute meeting with tech leads from all affected teams.
- Round-robin update: progress, blockers, questions.
- Coordinate on dependencies.
- Identify needs for Architecture Council input.
Monthly Leadership Update
15-minute slot in a stakeholder sync. Cover:
- Status: on track / at risk / blocked.
- Completion percentage (stories done / total stories).
- Revised timeline if needed.
- Escalations or decisions needed.
Example phrasing from the funnel doc:
"TypeScript migration 60% complete, 3 of 6 teams finished their epics. Vault team delayed 2 weeks due to higher priority security fix. Still on track for Q3 completion."
Celebrating Milestones
- First team completes its epic — recognition in Slack / team meeting.
- 50% completion — update in company all-hands.
- Last PR merged — celebrate with everyone involved.
Managing Documentation
Documentation should not wait until the end. Per the funnel doc:
- Draft early. Create documentation structure (outline / skeleton) as Implementation starts.
- Add content progressively. As patterns stabilize and teams produce real implementations, add to the docs.
- Have teams contribute examples from their own implementations.
Documentation lands in two homes per Documentation Patterns. What goes where:
Close-to-code (alongside the team's code, in the repository):
- Framework / pattern
README.md updates as the pattern stabilizes across teams. The PoC's initial framework README evolves with what teams discover during rollout.
- Folder-level notes in each team's adopted area pointing to the framework README and the ADR.
- Inline docs (JSDoc/XML comments for TypeScript/Angular/.NET;
rustdoc for Rust) per the per-stack rubric in Documentation Patterns.
- CLAUDE.md updates at the root and folder levels where the new pattern changes how engineers (and Claude tooling) should work. Use
@ syntax to link the README.md files that carry the canonical pattern.
Centralized (in bitwarden/contributing-docs, rendered at contributing.bitwarden.com):
- ADR updates — final status (Accepted → Implemented if your numbering scheme uses that), lessons learned, actual timeline vs. predicted. Edit the same ADR file you opened during PoC; don't create a new one.
- Migration guide — how to convert from the old pattern to the new. This is logistical / how-to content and belongs centrally so it's findable across repos.
- Contributing guide updates — new standards or patterns that should govern future code beyond this initiative.
- Runbooks — if the initiative involved operational or infrastructure changes, runbook content typically lives here too (or in SRE's home as appropriate).
Timing per the funnel doc:
- Draft technical guide when 2–3 teams have completed implementation.
- Finalize all documentation before marking the initiative complete.
- Plan for documentation to be ready 1–2 weeks before the final PR merges.
Knowledge Transfer
Per the funnel doc, ensure the initiative's outcomes live beyond your involvement. During final weeks:
- Tech talk or brown bag (45–60 min) for full engineering org, possibly at an office hours session or Architecture Council slot.
- Onboarding materials for new engineers.
- Update team runbooks with new patterns.
- Consider recording a video walkthrough.
Tech talk structure (funnel doc):
| Time | Content |
|---|
| 5 min | Approach we took and why |
| 10 min | Demo or code walkthrough |
| 5 min | Results and metrics |
| 5 min | Lessons learned |
| 5 min | Q&A |
The 30-Day Pulse Check (Work Transition Playbook Phase 4)
Composing the playbook from the originating side — this is the load-bearing checkpoint that prevents "we handed it off" from becoming "it was never picked up." The Work Transition Playbook is unambiguous: the 30-day pulse check is the one phase that should not be skipped regardless of how the rest is adapted.
A 15–30 minute conversation, or an async thread. Cover:
- Has the team begun working with the transferred material? If not, what's blocking?
- Unanswered questions or insufficient documentation areas?
- Is the team comfortable with the approach, or working around it?
- Does the support period need adjustment?
If a team hasn't started at all, escalate jointly with the receiving team — not punitively. Capacity issue, priority conflict, or transition gap? Understand before assuming.
Mid-Flight Course Correction
The funnel doc's example: mid-implementation discovery of GraphQL resolver performance issues. The shepherd's response:
- Immediately raise to Architecture Council.
- Pause affected teams' work while investigating.
- Work with one team to test a revised approach.
- Update guidance with the new findings.
- Communicate clearly: "Pause, don't abandon, we're fixing this."
- Extend timeline if needed.
The lesson: good shepherding includes recognizing when to pause, adjust, and communicate — not just pushing forward regardless.
Completion and Closure
Exit Criteria
Per the funnel doc, the initiative is complete when:
- All stories completed and merged to
main.
- All teams' epics marked complete in Jira.
- Documentation written, reviewed, and published.
- Knowledge transfer completed.
- Retrospective conducted.
Retrospective (Work Transition Playbook Phase 5, ~90 Days)
Schedule within 2 weeks of completion, while memories are fresh. 1.5 hours, you + tech leads from all affected teams. Per the funnel doc's agenda:
- What went well? Processes that worked, coordination wins, what to repeat.
- What could have gone better? Friction points, delays, wrong assumptions, what to do differently.
- Process improvements. Communication adjustments, scoping or estimation changes.
- Was the work understood well enough to execute? Which teams were close on estimates? Which were off? What caused variance?
Document findings and action items. Update the funnel process documentation with the learnings — Bitwarden's funnel gets better when shepherds add what they learned.
Closure (Work Transition Playbook Phase 6)
- Mark the BW initiative as complete in Jira.
- Archive the Slack channel (or make read-only as reference).
- Ensure all documentation is findable.
- Update related runbooks or onboarding materials.
- Submit any funnel-process improvements based on learnings.
- Update the ARCH idea status to its final completed status in JPD.
Recognize contributors publicly (all-hands, Slack), in performance reviews, with a case-study post if warranted.
The Work Transition Playbook's framing on closure applies here: don't linger as a "just-in-case" reviewer past closure — that's a soft form of refusing to let go.
Impact Measurement (3–6 Months After Completion)
Per the funnel doc, revisit the success metrics defined during Scoping:
- Quantitative metrics (if defined): bug reduction before/after, performance improvements, time savings. Example from the doc: "Predicted 30% reduction in error handling bugs; actual reduction: 42%."
- Qualitative feedback: survey teams — is the new pattern better than the old? What's working, what's not?
- Adoption tracking: Is new code following the pattern? Are teams defaulting to the new approach? Drift back to old patterns?
Document results:
- Update the ADR with actual vs. predicted outcomes.
- Add impact summary to the architecture plan.
- Share results with the engineering org.
Updates to the BW Initiative
During Implementation (see Idea-Based Initiatives):
- Status tracking is primarily through child epics and their stories — the initiative itself doesn't need frequent field updates.
- Comments: Significant coordination events — cross-team dependency resolution, mid-course corrections, milestone achievements, escalations. Comments become the narrative timeline used at retrospective.
- Links: Add new related work as it emerges — operational tickets, bug reports, documentation pages, adjacent initiatives.
- Description: Update only if scope or approach changed materially.
- ARCH idea status: Update to "5️⃣ Implementation" at kickoff, then to its final status at completion.
Common Mistakes
- Doing detailed code review. You are not the team's reviewer. Approach-alignment only.
- Letting drift compound. Catch divergence in PRs 1–2 per team, not PRs 10–20.
- Skipping the 30-day pulse check. The Work Transition Playbook is explicit — this is the load-bearing checkpoint. Skip it and silent failure modes become invisible.
- Treating documentation as an end-phase artifact. Patterns drift between Implementation and writeup. Document progressively.
- Quietly resuming work because a team isn't picking it up. Per the playbook, that's a leadership conversation, not a heroism opportunity.
- Lingering past closure. Hand back to the team's regular cadence and step away. The signal matters.
- Skipping the retrospective. It's the only mechanism that improves the funnel itself.
Reference
- Software Initiative Funnel §5 — canonical phase description, examples of successful, troubled, and course-correcting implementations.
- Work Transition Playbook — canonical six-phase transition; Phase 5 of the funnel is the originating-side support-period-through-closure portion.
- Documentation Patterns — close-to-code vs. centralized
contributing-docs, per-tech-stack best practices, CLAUDE.md conventions.
- Idea-Based Initiatives — how to update the BW Initiative through Implementation.
- Related:
Skill(shepherding-an-initiative) for the umbrella playbook; Skill(running-work-transitions) (in bitwarden-delivery-tools) for the originating-side support-period guidance; Skill(scoping-and-handing-off-to-teams) for the phase that hands work into this one.