Draft the Stakeholder Map and Engagement Plan, Then Correct It from Lived Knowledge

Workflow area: Elicitation & Discovery · Tier: Foundation · Hands-on time: ~25 minutes

Stakeholder analysis is usually done once, badly, in a template nobody reopens. The blank-page cost is the reason: mapping fifteen people across roles, influence, interest, and cadence is tedious enough that analysts write three rows and move on. AI removes the blank page. It cannot know that the procurement director blocks everything she was not consulted on early, or that the "sponsor" delegates every decision to a deputy — but when a full draft is in front of you, those corrections take seconds each. The corrections are the analysis; the draft is just the provocation.

When to use — and when not to

• Use it at project kickoffbefore elicitation starts, so the engagement plan drives who you talk to and in what order, not the other way around.
• Use it when inheriting a project mid-flightregenerate the map from the current charter and compare it to how engagement is actually happening; the gaps are your risk list.
• Use it before a governance resetwhen steering meetings have decayed into status theater, a fresh RACI and cadence proposal gives the conversation structure.
• Skip it for a two-stakeholder changea small enhancement with one sponsor and one SME needs a sentence, not a grid.
• Skip the AI draft when the politics are the projectin a restructuring or a turf war, even a draft grid on the wrong screen can cause damage. Build it manually, share it with no one by default.

The recipe

1. 1.

   Collect the raw material. Project charter or brief, org chart extract for the affected areas, and any known names and roles. Where you lack names, roles alone are fine — the draft can run on "Head of Customer Service" placeholders.

2. 2.

   Generate the three artifacts in one pass.

   Copy

   You are a business analyst building a stakeholder analysis.
   Project: [2-3 LINES — objective, scope, what changes for whom]
   Org context: [PASTE ORG CHART EXTRACT OR ROLE LIST]
   Known constraints: [e.g. "sponsor is CFO; project replaces a tool
   the ops team built in-house"]
   
   Produce:
   1. STAKEHOLDER REGISTER — table: role/name, what the project
      changes for them, likely stance (champion / neutral / at-risk
      / unknown), what they can block or accelerate. Mark every
      stance you inferred with (ASSUMED).
   2. POWER-INTEREST GRID — place each stakeholder in one of the
      four quadrants (manage closely / keep satisfied / keep
      informed / monitor) with a one-line justification each.
   3. RACI — rows are the project's major decisions and artifacts
      [LIST 5-8, e.g. "scope sign-off", "process design", "UAT
      acceptance"]; columns are stakeholders. Exactly one A per row.
   4. ENGAGEMENT CADENCE — per stakeholder: channel, frequency,
      owner, and the one message that must land with them.
   
   Rules: infer nothing about named individuals' personalities.
   Stances derive only from what the project changes for their role.
   Flag any row where you lack enough context as (UNKNOWN) rather
   than guessing.
   

3. 3.

   Review the sample output for shape. Good output looks like:

   Copy

   STAKEHOLDER REGISTER (extract)
   - Ops team lead — loses the in-house tool their team built;
     stance: at-risk (ASSUMED); can block: adoption, UAT
     participation, tribal knowledge transfer.
   - CFO (sponsor) — funds the project, owns the benefits case;
     stance: champion; can accelerate: escalations, cross-dept
     priority.
   
   POWER-INTEREST GRID
   - Manage closely: CFO, Ops team lead
   - Keep satisfied: CIO (platform standards) (UNKNOWN interest)
   

4. 4.

   Correct it from lived knowledge — this is the actual work. Go row by row and override the model with what you know: real influence versus org-chart influence, history between teams, who shows up versus who delegates, past project scar tissue. Change stances, move people between quadrants, reassign the A on decisions where formal authority and actual authority differ. Keep a small "delta log" of what you changed — those deltas are your genuine political insight, and reviewing them with your project sponsor is a high-value fifteen minutes.

5. 5.

   Pressure-test the RACI. Ask the model to critique its own draft after your corrections: "Given this corrected RACI, which rows have an accountable person who lacks the authority to decide? Where would a disagreement between R and A stall?" Fix what the critique surfaces, then walk the final RACI past the sponsor — a RACI nobody with power has blessed is a wish.

6. 6.

   Put the cadence in the calendar. An engagement plan that lives in a document is decorative. Create the recurring invites, name owners for each touchpoint, and set a monthly reminder to re-review the grid — stances move as the project makes its changes real.

Make it yours

• Claude Project for living stakeholder context: keep the corrected register in a Project and update it after each significant meeting ("ops lead attended UAT planning, constructive — update stance"); the map stays current instead of fossilizing at kickoff.
• Miro AI for workshop versions: generate the power-interest grid as a board and refine it live with your project leadership — placing people is a conversation, and the board makes disagreement visible and resolvable.
• ChatGPT/Gemini for the lightweight variant: on small projects, collapse the prompt to register-plus-cadence only; a grid for six stakeholders can live as one table in the project wiki.
• Program scale: for multi-workstream programs, run the prompt per workstream, then ask the model to merge registers and flag stakeholders appearing in three or more — those are your coordination risks and your steering-committee shortlist.

Quality check

• Every (ASSUMED) stance was reviewedeach one either got confirmed from lived knowledge or stayed flagged as a question to answer through early engagement, never silently promoted to fact.
• The grid reflects actual influenceat least one person moved quadrants during your correction pass; if you changed nothing, you reviewed nothing.
• One A per decision, and the A can actually decidethe most common failure is an accountable party who must escalate anyway.
• At-risk stakeholders have a specific plan"keep informed" is not a plan for someone who can block adoption; they need early involvement with a named owner.
• The delta log existsyour corrections are recorded, because they are the analysis you will want when a stakeholder behaves exactly as predicted in month four.
• Sensitive judgments stay off the shared copythe circulated version carries roles and cadence; candid stance assessments live in your private notes.