Stakeholder Management

Executive summary

Stakeholder management for AI programmes is not a one-off RACI slide—it is continuous alignment of economic buyers, users, risk approvers and delivery teams whose incentives conflict by design. You map power and interest, separate roles (who pays vs who uses vs who can say no), resolve conflicts before architecture gates, secure active sponsorship, and influence without authority when you own neither budget nor headcount.

When to use this page

• At programme kick-off before discovery workshops commit to scope or timelines.
• When a pilot stalls at “waiting for security/legal” or “business not engaged.”
• When C-suite, BU and control functions disagree on data use, autonomy level or vendor.
• Before production launch approvals that touch model risk, privacy and operational ownership.
• When you are a consultant or platform engineer with delivery accountability but no direct reports.

Executive view

Confirm named economic buyer, active sponsor and risk approver for each phase gate—not a generic “steering committee.” Ask monthly: what decision is blocked, and who has not yet said yes?

Decision required: Are decision rights documented (RACI/RAPID) for the top three contentious topics this quarter?

Technical view

Build stakeholder maps that include data owners, model risk, SRE and union/work council where relevant—not only IT and business leads.

Escalate conflicts with options and pre-reads; never surprise CISO or General Counsel in public forums.

Why this matters

AI programmes cut across every power centre in the enterprise. Business units want speed and customer impact. Finance wants measurable ROI. Cybersecurity wants data containment and auditability. Legal wants liability boundaries. HR and works councils worry about roles. Procurement wants competitive tender. Vendors want expansion. Users want tools that do not add admin.

An AI Solution Engineer sits at the intersection—often without authority to hire, fire, fund or approve. Delivery fails when stakeholders are “informed” too late, when the wrong person is treated as decision-maker, or when conflict is smoothed over in workshops and explodes at the production gate.

Strong stakeholder management produces:

• A living map of who cares, who decides, who can block
• Decision rights clear enough that teams do not re-litigate the same topic weekly
• Sponsors who remove blockers and communicate trade-offs upward
• Escalation paths that resolve conflict without destroying trust

Weak stakeholder management produces architecture rework, pilot purgatory, blame between “the business” and “IT,” and consultants who become permanent meeting facilitators.

Learn

Stakeholder categories in AI programmes

Group stakeholders by function in the decision, not only by job title.

Category	Typical roles	What they optimise for
Economic buyer	BU P&L owner, CFO delegate, division president	ROI, budget, strategic fit
Executive sponsor	CIO, CDO, CAIO, COO, BU leader	Programme success, reputation, pace
Process owner	Operations director, product owner	Workflow outcome, SLA, quality
End users	Agents, analysts, clinicians, engineers	Daily usability, workload, career impact
Technical delivery	Architecture, data, AI engineering, SRE	Feasibility, maintainability, standards
Risk approvers	CISO, DPO, legal, compliance, model risk	Control, liability, regulatory fit
Enablers	Procurement, vendor management, HR, L&D	Contract, capacity, skills, change
Influencers	Union, works council, clinical governance, audit	Fairness, safety, precedent

Missing any category from discovery interviews creates late veto risk.

Economic buyer vs user vs risk approver

These three roles are often different people. Conflating them is the root cause of “we thought business approved” when only a product manager attended demos.

Economic buyer

• Controls budget release and benefit realisation accountability
• Signs off business case and scale decision
• Asks: “Is this worth prioritised capital vs other initiatives?”

User (and user management chain)

• Performs or supervises the workflow daily
• Validates usability, exceptions, shadow processes
• Asks: “Will this help me hit my targets or add friction?”

Risk approver

• Can halt or constrain deployment regardless of business enthusiasm
• Owns policy interpretation: data residency, model risk tier, autonomous decisions
• Asks: “Can we defend this to regulator, auditor, board?”

Engagement rule: For each gate (pilot start, prod launch, scale), document three names—buyer, user representative, risk approver—and whether each has consulted, agreed, or decided.

Power / interest mapping

The power/interest grid (Mendelow) prioritises engagement effort.

	Low interest	High interest
High power	Keep satisfied (brief, no surprises)	Manage closely (active partnership)
Low power	Monitor (minimal effort)	Keep informed (regular updates)

Power indicators for AI:

• Budget sign-off over £X
• Veto on data classification or external model use
• Owns production on-call or operational SLA
• Controls workforce policy or union relationship
• Regulatory interface (named accountable individual)

Interest indicators:

• Workflow directly changed
• Headcount or role impact
• Customer-facing reputation exposure
• Personal KPI tied to outcome

Refresh the map quarterly or when org restructures, vendor changes, or regulatory events land.

Stakeholder maps — practical format

A useful map is one page, not a 200-row spreadsheet.

Minimum fields per stakeholder:

• Name, role, organisation
• Power (H/M/L), interest (H/M/L)
• Stance (champion / supporter / neutral / sceptic / blocker)
• Primary concern (one line)
• Required engagement (weekly / gate-only / exception)
• Last touchpoint and next action

Visual: Grid quadrant + short narrative on top five relationships to manage closely.

Anti-pattern: Map created in week one, never updated when blocker becomes champion after successful demo.

RACI and RAPID — decision rights

Use RACI for ongoing activities; RAPID for single major decisions.

RACI: Responsible, Accountable, Consulted, Informed

RAPID: Recommend, Agree, Perform, Input, Decide

Model	Best for	AI example
RACI	Recurring processes	Monthly model performance review
RAPID	One-off high-stakes decisions	Approve production RAG with customer PII

Rules that prevent RACI theatre:

• Exactly one Accountable (RACI) or Decide (RAPID) per row
• “Consulted” is not veto unless explicitly documented
• Informed stakeholders receive read-ahead 48 hours before steering
• RACI published where teams can find it; version controlled

Example RAPID — approve production RAG launch:

Role	RAPID	Name
BU operations lead	Recommend	Launch scope and rollout wave
AI engineering lead	Perform	Operate service to SLA
CISO	Decide	Data egress and security controls
DPO	Agree	Privacy impact and retention
Legal	Input	Contract and liability
Contact-centre manager	Input	User readiness and training

Conflict — predictable fault lines

AI programmes hit the same conflicts repeatedly. Name them early.

Fault line	Typical positions	Resolution approach
Speed vs control	BU wants Q2 launch; CISO wants private deployment	Timeboxed spike; phased launch; explicit risk acceptance by buyer
Central vs federated	Group AI platform vs BU shadow IT	Architecture board decision; chargeback or guardrails
Autonomy vs human-in-loop	“Deflect 40%” vs model risk cap	Suitability ladder; tiered automation by use case
Public LLM vs private	Cost vs data residency	Decision matrix on data classification; no ad hoc paste
Build vs buy	Engineering pride vs time to value	Business-case comparison; TCO not licence alone
Vendor vs incumbent	Microsoft shop vs specialist	Proof criteria; exit strategy; integration cost

Conflict protocol:

1. Name issue without attributing bad faith
2. Document options with trade-offs (time, cost, risk, value)
3. Identify Decide role; do not vote in workshops
4. Record decision, dissent, and review date in decision log
5. Communicate downward within one week

Sponsorship — active vs nominal

Nominal sponsor: Name on deck; forwards emails; absent from trade-off meetings.

Active sponsor: Removes blockers, attends gates, communicates “why this, why now” to peer executives, accepts public ownership of failure.

Sponsor effectiveness checklist:

• Attends steering monthly (or delegates with decision authority—rare)
• Responds to escalations within agreed SLA (e.g. 5 business days)
• Communicates programme priority to conflicting BUs
• Owns benefit realisation metric, not only delivery date
• Pairs with second-line sponsor for continuity (vacation, reorg)

If sponsor is nominal after two escalations, raise programme risk to buyer explicitly.

Influence without authority

Consultants, solution engineers and architects often lead without line management. Influence levers:

Credibility — Accurate summaries, no surprise, admit unknowns, bring evidence not hype.

Pre-work — Pre-reads that make executives efficient; options framed for decision not discussion.

Coalitions — Align process owner and respected user champion before steering.

Reciprocity — Help risk teams with control language they can reuse; they help you at gates.

Escalation discipline — Escalate with options and recommendation, not open complaints.

Transparency — Decision log visible; dissent recorded; no back-channel rewrites.

What not to do: End-run blockers; commit dates without Perform role agreement; confuse enthusiasm in a demo with organisational decision.

Steering vs working groups

Forum	Purpose	Cadence	Attendees
Steering	Decisions, escalations, benefit tracking	Monthly	Sponsor, buyer, Decide roles, programme lead
Working group	Backlog, design, risks, demos	Weekly	R/P roles, SMEs
Control forum	Security, privacy, model risk	Gate-based	CISO, DPO, model risk, architecture
User council	Workflow, adoption, exceptions	Fortnightly	Users, supervisors, change

Steering packs: BLUF, decisions needed, RAG status, blockers with owner, financials vs plan, next gate date. Never first-read a contentious topic in the room.

Frameworks and methods

Stakeholder engagement plan (SEP)

One-page plan per major stakeholder (top 8–12):

• Current stance → desired stance
• Key message (what they need to hear)
• Owner of relationship (named person, not “programme”)
• Tactics (1:1, demo, site visit, risk workshop)
• Success signal (behaviour change, signed artefact)

Review SEP at steering; update after reorgs.

Three yeses before production

Production launch requires documented:

1. Business yes — Buyer and process owner accept outcome and operating model
2. User yes — Training complete; support model agreed; union consulted if required
3. Risk yes — CISO/DPO/legal/model risk controls in place or risk accepted in writing

Missing any “yes” is a programme risk, not a “soft” delay.

Escalation ladder

Level	Trigger	Forum	SLA
L1	Task blocked >5 days	Working group lead	2 days
L2	Cross-function conflict	Programme manager + sponsors	5 days
L3	Budget, policy, or launch veto	Steering	Next scheduled or extraordinary
L4	Regulatory or reputational exposure	Executive committee	As defined by client

Each escalation includes: issue, impact, options, recommendation, Decide role.

Coalition map for AI adoption

Identify champions, swing voters, and blockers among:

• Frontline supervisors (often stronger than executives on adoption)
• IT operations (own on-call pain)
• Middle management (fear of metric change)
• Peer BUs (fear of unfunded mandate)

Plan wins: pilot with willing BU; publish metrics; rotate user council chair from sceptic team after co-design.

International and matrix organisation dynamics

Global banks, pharma and manufacturers often have matrix reporting—BU P&L vs group CTO vs country regulator. Stakeholder maps need a layer per dimension:

Dimension	Example stakeholder	Decision scope
Group	Group CDO	Platform standards, group funding
BU	Retail COO	BU pilot priority, local ROI
Country	UK Country CEO	Local regulatory acceptance
Function	Group CISO	Global security policy

Fault line: BU funded pilot on group-prohibited data path. Fix: RAPID with group Decide on standards before BU Decide on rollout—sequence decisions explicitly in steering pack.

Working with procurement and third parties

Procurement is often labelled “blocker” when they are Perform on competitive process. Engage early with:

• Category strategy (build vs buy vs partner)
• Evaluation criteria weighted toward client outcomes, not vendor features
• Contract clauses on data use, model retraining, exit portability

Influence tactic: Provide procurement a decision-grade options paper identical to executive format—reduces cycle time because they can paste into tender docs.

Union and works council engagement

In UK/EU contact centres and public sector, works councils may have consultation rights on monitoring and performance metrics tied to AI.

Do:

• Engage early with formal agenda items: what is measured, what is not, human override
• Bring user council proposals, not fait accompli
• Document “no keystroke surveillance” where applicable

Do not:

• Surprise union with launch announcement
• Promise “no job impact” you cannot guarantee—frame task change and retraining

Stakeholder map should include union interest even when power feels low—reputational power is high.

Operating-model notes

Discovery → Stakeholder map v1 + SEP for top 10
Design → RACI/RAPID for build and launch decisions
Pilot → User council + weekly working group
Pre-prod → Three yeses checklist
Scale → Refresh power/interest; union/work council if headcount impact

AI Solution Engineer typically facilitates the map and maintains the decision log; programme manager owns cadence and RAID; sponsor owns escalations above L2.

Real-world scenarios

Scenario A — Global bank: public LLM vs private deployment

Context: UK retail bank; 12,000 contact-centre staff; pilot copilot for policy Q&A. BU sponsor wants Q2 launch using public SaaS (£180k/year). CISO blocks customer PII in external API. BU IT cost centre already over budget.

Stakeholder landscape:

Stakeholder	Power	Interest	Stance
BU COO (buyer)	High	High	Champion — KPI: handle time
CISO	High	Medium	Blocker on public LLM with PII
Contact-centre union	Medium	High	Sceptic — monitoring, headcount
CIO	High	Medium	Neutral — wants standard platform
DPO	High	High	Consulted — DPIA required

Intervention:

• RAPID workshop: CISO Decide on data path; BU COO Recommend on rollout; CIO Agree on platform standard.
• Options paper: (A) public with strict redaction — DPO flags residual risk; (B) private EU deployment — +£420k/year, +10 weeks; (C) on-prem corpus only, public model prohibited — +£650k capex.
• Economic buyer accepts B with phased rollout; union Input via works council briefing on monitoring policy (no keystroke surveillance; aggregate QA only).

Outcome: Launch slips 10 weeks; programme survives because conflict resolved before build complete. Decision log published; steering records COO acceptance of cost uplift.

Numbers: Private deployment TCO year one £580k vs public £180k; projected handle-time benefit £720k/year—still positive ROI at buyer hurdle rate.

Scenario B — Pharmaceutical field medical — MLR and medical affairs

Context: Top-10 pharma; 800 field medical staff; idea to use GenAI to draft initial medical information responses. Medical Legal Regulatory (MLR) average cycle 18 days; field requests growing 22% YoY.

Stakeholder landscape:

• Medical affairs VP (buyer) — wants faster responses, fears off-label risk
• MLR director (risk approver) — Decide on content policy; blocks autonomous send
• Field medical directors (users) — high interest; fear tool creates more review work
• IT/digital health — Perform; integration with Veeva and audit trail
• Country affiliates — low power centrally, high power locally for launch

Conflict: Digital vendor sold “80% automation” to buyer; MLR heard “80% less review”—blocker formed pre-workshop.

Intervention:

• Stakeholder map surfaces MLR as Decide, not “consulted”
• RACI for “draft medical response”: AI Responsible for draft; medical writer Accountable; MLR Consulted on template library only
• Pilot scoped to pre-approved snippet assembly, not free generation—autonomy capped

Outcome: Pilot approved in two countries; MLR cycle on pilot queue drops from 18 to 11 days (39%); scale gated on audit sample of 500 outputs with zero off-label findings.

Scenario C — Public sector housing repairs — influence without authority

Context: Regional housing association; 1.2m properties; contractor allocation AI pilot suggested by central digital team. No budget authority for solution engineer (external consultant). Economic buyer: Director of Repairs. Blocker: Union branch secretary (overtime allocation fairness).

Constraints: Consultant cannot escalate to CEO without buyer; media sensitivity on “AI replacing trades.”

Approach:

• Power/interest: union high interest, medium power—move from “keep informed” to “manage closely”
• Co-design sessions with union health and safety rep; algorithm inputs published; human override mandatory
• User champion: senior scheduler with 20 years tenure—credibility with peers
• Pre-read to buyer: three options including no-AI workflow fix ranked by cost

Outcome: Buyer chooses hybrid—AI recommends routes; scheduler decides; union neutral public statement; pilot proceeds without CEO escalation.

Lesson: Influence without authority depends on correct Decide map and credible user champions, not executive charisma alone.

Scenario D — Energy retailer — group AI CoE vs BU impatience

Context: UK energy retailer; group AI CoE mandates Azure OpenAI standard; BU smart-meter disconnection team wants niche vendor for field-agent voice offline mode. £24m annual cost of failed first visits; 1,400 field agents.

Stakeholders:

• Group CTO (high power, medium interest) — standardisation
• BU field operations director (buyer, high/high) — visit success KPI
• Field agents (users, low power, high interest) — offline capability sceptical of cloud-only
• Vendor sales (influencer, medium/low) — bypass CoE via BU exec lunch

Conflict: BU signed vendor POC contract without architecture board—CoE threatens to withdraw funding for BU’s other AI backlog.

Intervention:

• Emergency RAPID: Group CTO Decide on approved patterns; BU director Recommend on use case; field ops lead Perform
• Option hybrid: approved group platform for online; edge cache pattern for offline read-only job pack—not niche vendor full stack
• Decision log published; vendor POC scoped to evaluation only, not production

Outcome: Avoids dual standards; BU accepts 8-week engineering spike; visit rework baseline −8% in pilot region over 16 weeks (from 22% to 14% failed first visit rate).

Stakeholder lesson: Influencers with commercial intent belong on the map explicitly—otherwise they reshape decisions outside governance forums.

Practice exercises

Primary exercise — RACI for production RAG launch (60 minutes)

Brief: Enterprise insurance firm launching internal claims-policy RAG for 2,500 adjusters. Roles: Claims COO, CISO, DPO, Head of Claims Operations, AI Platform Lead, Vendor Account Manager, Model Risk, Union rep (inform only, not Decide).

Tasks:

1. Draw power/interest map; mark top three “manage closely.”
2. Complete RACI for: (a) weekly prompt change, (b) corpus update, (c) production launch, (d) incident response for harmful output.
3. Complete RAPID for production launch decision only.
4. Write escalation L2 memo: CISO wants private cluster; Claims COO wants SaaS for speed—three options, recommendation, Decide role.

Acceptance criteria:

• One Accountable/Decide per RACI row
• Union correctly Informed, not Consulted, unless client policy says otherwise (document assumption)
• Memo ≤400 words, BLUF first line

Stretch exercise — Sponsor recovery plan (half day)

Brief: Six-month pilot; nominal sponsor (CIO) disengaged; CISO issued conditional approval with 90-day expiry; BU usage at 34% vs 70% adoption target; procurement questioning vendor renewal.

Tasks:

1. Forensic stakeholder map: who changed stance since kick-off?
2. SEP for CIO, CISO, BU process owner—move stance one notch toward champion/supporter
3. Steering pack outline for extraordinary session: decisions needed, kill/continue/pivot recommendation
4. Decision log entries for last three conflicts (reconstructed plausibly)
5. 90-day influence plan without authority—named tactics per stakeholder

Acceptance criteria:

• Three yeses assessment with red/amber/green
• Explicit programme risk if sponsor remains nominal
• No villain narrative—structural incentives named

Questions you should be able to answer

1. Who is the economic buyer for this phase, and who is merely enthusiastic?
2. Who can say no with regulatory or policy force, even if the buyer says yes?
3. Who represents end users with authority to speak for workflow exceptions?
4. Where are decision rights documented today, and are they version controlled?
5. Which stakeholders moved quadrant since last map update?
6. What is the top unresolved conflict, and who is the Decide role?
7. What does the sponsor do monthly that proves active sponsorship?
8. Who must be Consulted vs Informed for the next gate—and what is the pre-read deadline?
9. What happens if CISO and buyer disagree after steering—escalation path?
10. Which middle managers fear metric change, and how are they engaged?
11. Where does union or works council fit for monitoring and role impact?
12. What decisions were made in rooms where risk approvers were absent?
13. Who owns on-call and operational SLA post-launch—Perform role named?
14. How do you influence a blocker without end-running them?
15. What would steering need this month to unblock the critical path?
16. How do you re-engage a nominal sponsor without burning the relationship?
17. When did you last update the map after an org restructure or vendor change?

Monthly steering cadence — stakeholder rhythm

Week	Stakeholder activity
1	SEP review; 1:1 with swing voters; pre-read draft
2	Working group; user council; update RACI conflicts
3	Risk/control touchpoint if gate within 30 days
4	Steering pack issued 48h before; decision log updated

Consistency beats heroics—blockers expect predictable engagement.

RACI worked example — weekly model performance review

Activity	AI engineering	Model risk	BU owner	MLOps/SRE	Legal	Steering
Run eval suite	R	C	I	A	I	I
Interpret drift metrics	R	C	C	C	I	I
Approve prompt/model change	C	C	C	R	C	A
Communicate incident to regulators	I	C	R	C	A	I
Decide rollback vs hotfix	C	C	C	R	I	A

Exactly one A per row. Model risk Consulted on approval—not Accountable—unless client policy explicitly delegates.

Influence scripts — phrases that work without authority

Situation	Weak phrase	Strong phrase
Blocker raises new concern in steering	“We’ll take that offline”	“CISO, we need Decide on data path today to hold build—Options A/B on slide 4; recommend B.”
Sponsor silent	“Any thoughts?”	“[Name], we need your call on defer vs proceed—default is defer if no response by Friday.”
BU bypasses CoE	“That’s against policy”	“Architecture board Decide required—happy to expedite RAPID this week with options memo.”
User scepticism	“Training will fix it”	“What would convince you in 30 days—shadow shift, metric, override rate?”

Scripts are not manipulation—they name the decision respectfully.

Shadowing and discovery — stakeholder signals from observation

When interviews contradict reality, shadowing (with permission) reveals true stakeholders:

• Who do users actually call when the system fails? (Often unofficial power)
• Which dashboard does the supervisor refresh hourly? (Hidden KPI)
• Where does work leave the official system? (Spreadsheet = stakeholder not on map)

Spend 4–8 hours shadowing before finalising RACI for production launch—users reveal blockers executives never mention.

Post-launch stakeholder shift

After go-live, power often moves:

Pre-launch power centre	Post-launch shift
Sponsor / buyer	Operations / service owner
Architecture	SRE / on-call
Vendor sales	Support + account team
Risk approver	Audit / incident response

Refresh map at 30/60/90 days post-launch; SEP targets shift from “approve” to “sustain adoption” and “operational acceptance.”

Reading stakeholder stance changes

Track stance transitions in your decision log—they predict programme health:

From → To	Interpretation	Action
Neutral → Supporter	Demo or 1:1 built trust	Ask them to co-present to peer
Supporter → Sceptic	Incident or scope creep	Root-cause 1:1; revise SEP
Sceptic → Blocker	Missed consultation	Escalate L2 with options; never ambush
Blocker → Neutral	Face-saving path offered	Document concession publicly

Stance is dynamic; maps that freeze week-one opinions become fiction by month three.

Gift of pre-work — reducing executive load

Stakeholders with high power and low interest (CFO, some board members) engage when you reduce load:

• One-page summary with three numbers they will repeat internally
• Explicit “no decision needed today—information only” when true
• Red/green dashboard vs last steering—no narrative rehash
• Appendix for their analyst to drill

This is stakeholder management through communication design—pairs directly with topic 28.

Onboarding new stakeholders mid-programme

Reorgs and role changes are guaranteed on 12-month AI programmes. When a new CISO or BU lead arrives:

1. Brief within 10 days — BLUF programme status, next gate, one risk they own
2. Do not re-open settled decisions unless they have formal authority to reverse
3. Offer 1:1 before their first steering—never first exposure in group challenge
4. Update RACI within one week; circulate version number

New leaders often challenge to establish credibility—pre-brief converts challenge into informed oversight.

Keep a stakeholder changelog (date, name, old role → new role, impact on RACI) in the programme wiki—consultants who leave mid-programme rely on this handover artefact.

Review the changelog at monthly steering so Decide roles never surprise the room with “I wasn’t aware I owned this.”

Negative cases — when stakeholder management fails

The friendly demo trap

Symptom: Product manager loves demo; treated as economic buyer; real buyer never saw business case.

Impact: Funding stops at scale; “business didn’t adopt” when business never committed.

Fix: Three names on charter—buyer, user rep, risk approver—signed before build.

RACI wallpaper

Symptom: RACI in deck appendix; teams ignore it; five people think they decide.

Impact: Re-litigation every sprint; slow decisions; consultant as full-time facilitator.

Fix: One Decide per decision; publish in team wiki; reference in meeting invites.

Surprise CISO

Symptom: Architecture complete; first CISO review at pre-prod.

Impact: Six-month delay; redesign; blame culture.

Fix: CISO/DPO in discovery and design gates; early DPIA/security assessment.

Sponsor substitution

Symptom: Delegates send juniors without decision authority to steering repeatedly.

Impact: Blockers not cleared; steering becomes status theatre.

Fix: Escalate programme risk; require delegate letter specifying decision limits.

User-washing

Symptom: Single power user attends all sessions; frontline sceptics excluded.

Impact: Low adoption; shadow spreadsheets; public failure at scale.

Fix: User council with rotating reps; site visits; supervisor engagement.

Conflict avoidance

Symptom: Facilitator “finds consensus” without Decide role; issues parked “offline.”

Impact: Identical fight at launch; trust collapse.

Fix: Options + RAPID in room; record decision or named escalation date.

Output

Expected outputs for this topic:

• Stakeholder map — Power/interest, stance, top-five engagement list
• Stakeholder engagement plan (SEP) — Per key stakeholder tactics and success signals
• RACI — Recurring activities with single Accountable
• RAPID — Major gate decisions with single Decide
• Decision log — Issue, options, decision, dissent, review date
• Escalation path — L1–L4 with SLAs and forums
• Steering pack template — BLUF, decisions, blockers, financials
• Three yeses checklist — Business, user, risk for production
• Sponsor effectiveness review — Active vs nominal assessment

Practice checklist

• I can name buyer, user rep and risk approver for my current or practice scenario
• I completed RACI and RAPID for production launch without role conflicts
• My stakeholder map distinguishes power from interest with updated stances
• I wrote an escalation memo with options and a single Decide role
• I identified at least two predictable fault lines and mitigation tactics
• I can explain influence without authority with three concrete levers
• I documented negative cases relevant to a programme I know

Related playbook content

• Leadership: How to Engage with Business and Service-Line Leaders — Executive relationship patterns
• Leadership Business Engagement roadmap — Sequenced leadership capability
• Leadership: Represent the Firm Externally — External stakeholder communication
• Consulting and client engagement — Engagement setup and executive alignment
• Change Management and Adoption — User adoption and resistance
• Communication and Executive Articulation — Steering packs and BLUF
• Responsible AI and Governance — Risk approver artefacts
• Delivery and Programme Management — RAID, cadence and governance
• Project Management — PMBOK® Fifth Edition study map — Classic stakeholder theory
• How to use this Learning Map — Reference-depth study method
• 8D Framework — Lifecycle gates

Supplemental scenario — public sector RAPID conflict on data residency

Context: UK council adult services; vendor proposes US-hosted LLM; CISO Recommend, Director Decide, Sponsor Agree on pilot scope.

RAPID snapshot:

Role	Name (fictional)	Stance	Action
Recommend	CISO	No US processing	Private UK tenant spec
Agree	Sponsor (DCS)	Wants fast pilot	Accept 6-week slip
Perform	ASE	Build eval	UK-only stack
Input	Legal	DPIA conditions	HITL mandatory
Decide	Director	Fund £185k	Approved UK path

Escalation memo (7 lines): Ask Director to Decide UK tenant vs defer; cost of US path £0 infra but infinite legal delay (Assumption); coalition: CISO + Legal + union rep briefed pre-meeting.

Fault line mitigated: "Public cloud vs on-prem" re-litigation stopped by Decide recording: "Approve managed UK region; revisit on-prem if cost +40% at year 2 review."

Outcome: Pilot launched without sponsor override of CISO Recommend role—decision log cited RAPID explicitly; audit trail satisfied scrutiny committee.

Key takeaways

• Economic buyer, user and risk approver are usually three people—map them explicitly every phase.
• Power/interest grids prioritise your week; RACI/RAPID stop decision re-litigation.
• Predictable AI conflicts (speed vs control, public vs private LLM) deserve pre-baked resolution protocols.
• Active sponsors remove blockers; nominal sponsors are a programme risk to escalate early.
• Influence without authority runs on credibility, pre-work, coalitions and disciplined escalation—not charisma.

What to do next

1. Complete the primary RACI/RAPID exercise and file it in your pattern library.
2. Continue with Communication and Executive Articulation for steering narratives.
3. Apply the Consulting and client engagement Guide on your next kick-off.
4. Cross-check Change Management and Adoption for user council design.