Change Management and Adoption
Executive view
Fund adoption as part of the investment—not 5% leftover budget. Ask for 30/90-day metrics, sponsor coalition health, and resistance plan with named owners—not vanity login counts.
Decision lens: If adoption stays at 40% for 90 days, do we kill, pivot or invest in embed—and who decides?
Technical view
Embed AI in systems users already use; instrument task success, override rate and time-on-task; feed resistance signals back to UX and eval teams.
Champions are not marketing—they need early access, influence pathways and time allocation in performance goals.
Why this matters
Enterprise AI programmes routinely report “successful pilot” on technical metrics while benefits owners see no P&L movement. Root cause is rarely model accuracy alone—it is workflow mis-fit, missing training, supervisor mistrust, perverse KPIs, or tools living in a tab users never open. A global insurer deployed a claims copilot to 2,500 adjusters; 78% had logged in at least once after 30 days, but only 22% used it on closed claims (operational log). Median handle time unchanged. Investigation: tool opened in separate browser; supervisors penalised slow handle time during “learning”; no champions in team stand-ups; training was a 20-minute video with no hands-on. Re-embed in CRM, supervisor incentives aligned, 12 champions trained, shadow shifts—by day 90, 61% claims assisted, handle time −19%. Cost of first failed adoption wave: £340k delivery spend with negligible benefit.
Change management is how AI Solution Engineering connects to value. Topic 25 covers plan and gates; topic 26 covers people and behaviour. Topic 23 covers trust UX; topic 26 covers organisational trust—sponsors, supervisors, unions, power users.
Regulated and unionised environments add consultation obligations—surprise launch is both unethical and programme suicide. Adoption metrics without ethics (keystroke monitoring, hidden quality scoring) create resistance that looks like “user error” but is rational pushback.
Pair with Adoption and change, Stakeholder Management, User Experience and Human Factors, and Communication and Executive Articulation.
Learn
Adoption vs activation vs benefit realisation
| Term | Definition | Example metric |
|---|---|---|
| Activation | First meaningful use | Completed one assisted task |
| Adoption | Habitual use on target workflow | ≥50% cases assisted/week |
| Benefit realisation | Measured outcome improvement | Handle time −20% at ≥55% adoption |
Vanity metrics: Logins, page views, “users enabled.” Action metrics: Assisted task rate, override rate, time-on-task vs baseline, quality/error rate.
Impact assessment — whose job changes how
Before training slides, document role impact:
| Role | Today | Future | Delta | Support needed |
|---|---|---|---|---|
| Contact agent | Search 6 systems | Review AI draft + cite | Less search, more verify | Shadow shifts |
| Supervisor | Listen to calls | Review override reports | New QA signals | Playbook |
| Policy owner | Update PDFs | Own corpus + eval failures | New accountability | Corpus workflow |
Publish impact summary to works council / union where required—what changes, what does not (e.g. no automated redundancy decision).
Readiness — ADKAR per persona
ADKAR (Prosci): Awareness, Desire, Opportunity (Knowledge + Ability in some variants), Reinforcement. Use per persona, not one org-wide checkbox.
| ADKAR element | Question | Tactics |
|---|---|---|
| Awareness | Do they know why change? | Sponsor story, peer metrics |
| Desire | Do they want to participate? | WIIFM, supervisor alignment, fear addressed |
| Knowledge | Do they know how? | Role-based training, job aids |
| Ability | Can they do it in flow? | Embed, shadow, office hours |
| Reinforcement | Does culture sustain it? | KPIs, champions, celebrate good overrides |
Readiness gate example: <70% supervisors complete enablement workshop → delay wave 2.
Kotter’s eight steps — programme scale
Kotter complements ADKAR for large transformations:
- Create urgency — cost of delay with numbers
- Build guiding coalition — sponsor + ops + union + IT
- Form strategic vision — bounded AI assist, not hype
- Enlist volunteer army — champions (not conscripts)
- Enable action — remove barriers (embed, SSO, time)
- Generate short-term wins — publish pilot metrics
- Sustain acceleration — wave rollout, don’t declare victory week 4
- Institute change — BAU ownership, corpus/eval in org design
AI pitfall: Step 4 becomes “mandate usage” without step 5 embed—champions burn out.
Champions network — design and protect
Champions are credible peers with time allocated—not random volunteers who get extra work unpaid.
Champion charter elements:
- Selection: respected operators, mix of sceptics converted and enthusiasts
- Time: 4–8 hours/month protected in job plan
- Access: early features, direct line to product/quality
- Role: floor coaching, feedback triage, demo in team huddles—not IT support
- Recognition: supervisor visibility, not gimmick prizes
Ratio: 1 champion per 15–25 users in high-touch roles; 1 per 50–80 in light-touch.
Anti-pattern: Champions are only managers—floor credibility lost.
Training — role-based, embedded, measured
| Training type | Duration | Pass criterion |
|---|---|---|
| Executive | 30 min | Can explain scope, risk, benefit hurdle |
| Supervisor | 2 hr | Can read adoption dashboard + coach overrides |
| User core | 4 hr hands-on | Task success on 3 scenarios in training env |
| Refresher | 30 min | After major UX or policy change |
| Champion | 1 day | Can train peers + escalate quality issues |
Training content must include: scope limits, HITL accountability, report-wrong, refuse states, escalation—not only “how to click.”
Competency check: Simulated wrong answer users must catch before production access (high-risk roles).
Communications — narrative arcs without hype
Message map by stakeholder (link topic 27):
| Audience | Core message | Channel |
|---|---|---|
| Users | “Verify before send; you stay accountable” | Huddle, job aid |
| Supervisors | “Overrides are good scepticism early” | Playbook |
| Executives | “Adoption hurdle tied to £ benefit” | Steering |
| Union | “No covert surveillance; human decides” | Consultation |
Cadence: Pre-launch T-4 weeks awareness → T-2 training → T-0 launch with office hours → T+2 wins story → T+4 honest retrospective.
Resistance — types and responses
Resistance is data, not defiance.
| Resistance type | Signal | Response |
|---|---|---|
| Skill fear | “I’m too old for this” | Shadow shifts, champions |
| Trust fear | “It will get me fired” | Transparency on metrics monitored |
| Workflow | “Extra clicks” | Embed, UX fix, remove duplicate search |
| Political | Middle manager loses span | Engage manager in KPI redesign |
| Ethical | Fairness, bias | Co-design, publish limits, human override |
| Prior trauma | “Last AI project failed” | Differentiate with kill criteria, baseline |
Never: Mock resistance in steering. Do: Name top three resistance drivers with mitigation owners.
Process redesign — don’t digitise broken workflow
AI on top of broken process amplifies waste. Co-design workshops answer:
- Which steps delete (duplicate search)?
- Which steps move (approval earlier)?
- Where human must remain (accountability)?
Example: Claims flow added “review AI draft” but kept manual search “just in case”—+3 minutes. Fix: remove redundant search when confidence High + citation clicked.
Embed strategy — meet users where they work
| Embed pattern | Pros | Cons |
|---|---|---|
| CRM side panel | In flow | Narrow UI |
| Teams/Slack bot | Familiar | Thread noise |
| Desktop overlay | Always visible | Intrusive |
| Separate portal | Fast to build | Low adoption |
Rule: If users live in CRM, CRM embed for phase 1—portal is rarely adoption winner.
Adoption metrics and dashboards
Dashboard minimum (30/90 day):
| Metric | Formula / source | Red threshold example |
|---|---|---|
| Active users | Distinct users with ≥1 assisted task / week | <40% of cohort |
| Assisted task rate | Assisted tasks / eligible tasks | <50% |
| Task success | Completed without rework | <80% |
| Time on task | Median vs baseline | No improvement by day 60 |
| Override rate | Edits before send / drafts | <3% (blind trust) or >40% (quality) |
| Report rate | Reports per 1k sessions | <1 (UI broken) |
| Escalation to human | Handoffs / sessions | Spike >2× baseline |
| Quality | QA sample error rate | Above pre-AI baseline |
| NPS / trust survey | Role-specific | <0 net for agents |
Segment by site, tenure, supervisor—aggregate hides pockets of failure.
30/60/90-day adoption plan template
| Period | Goal | Actions | Decide |
|---|---|---|---|
| Day 0–30 | Activation + trust | Champions live, office hours, shadow | Continue wave 2? |
| Day 31–60 | Habit on target workflow | Supervisor coaching, remove redundant steps | UX fixes prioritised |
| Day 61–90 | Benefit signal | Compare handle time/quality vs baseline | Scale / pivot / kill |
Kill criteria example: If assisted rate <45% at day 90 despite embed fix, pause scale—root cause analysis before spend.
Power users vs sceptics
| Persona | Risk | Engagement |
|---|---|---|
| Power user | Bypass governance with consumer AI | Channel into official tool; early access |
| Sceptic | Block peers informally | Co-design; public override wins |
| Passive | Login once, revert | Supervisor nudge + workflow embed |
| Enthusiast | Over-trust output | Training on verification |
Rotate sceptic converted stories in comms—not only enthusiasts.
Reinforcement and KPI alignment
If KPIs punish learning period, adoption dies. Align:
- Temporary learning buffer on handle time (4–6 weeks)
- Supervisor scorecard includes quality of verification, not only speed
- Celebrate good catches (user reported wrong answer that prevented incident)
Perverse KPI example: Bonus on calls/hour during pilot → agents skip AI → declare “doesn’t work.”
Frameworks and methods
ADKAR checklist template (rollout)
Use for each wave sign-off:
Wave: Retail claims Q3 | Persona: Adjuster
[ ] Awareness — sponsor video aired + team huddle brief
[ ] Desire — WIIFM doc; union FAQ published
[ ] Knowledge — 4hr lab completed; 90% pass quiz
[ ] Ability — CRM embed live; shadow shift completed
[ ] Reinforcement — supervisor playbook active; champion office hours scheduled
Blockers: DLP latency — amber — owner IT — due 12 Aug
Go/no-go: Steering 15 Aug
Kotter + ADKAR combined map
| Kotter step | ADKAR emphasis |
|---|---|
| Urgency | Awareness |
| Coalition | Desire (leadership) |
| Vision | Awareness + Desire |
| Volunteer army | Ability (champions) |
| Enable action | Ability (embed) |
| Short-term wins | Reinforcement |
| Sustain | Reinforcement |
| Institute | Knowledge institutionalised (BAU) |
Influence-interest for adoption stakeholders
Map supervisors and union as high interest even if formal power feels medium—adoption lives or dies there.
RACI for change activities
| Activity | Sponsor | PM | ASE | L&D | Supervisors | Champions |
|---|---|---|---|---|---|---|
| Impact assessment | A | R | C | C | C | I |
| Training content | C | C | C | R | C | C |
| Comms cascade | A | R | C | C | R | C |
| Embed decision | C | C | R | I | C | I |
| Adoption dashboard | I | R | C | I | C | I |
| Resistance plan | A | R | C | C | R | C |
Benefits linkage formula
Benefit £ = (Baseline cost per task − Assisted cost per task)
× Volume × Adoption % × Quality factor
Document each variable as Fact/Assumption/Hypothesis (topic 28).
Wave rollout pattern
- Alpha — champions only (n=15–30)
- Beta — willing site (n=100–300)
- Wave 1 — 20% population
- Wave 2+ — scale with lessons
Each wave has ADKAR gate—not only technical gate.
Hypercare and BAU transition
| Phase | Duration | Focus | Exit criterion |
|---|---|---|---|
| Launch week | 5 days | Floor walking, office hours | Critical incidents <3/day |
| Hypercare | 4–8 weeks | Champions + PM on-site/virtual | Adoption ≥ pilot hurdle |
| Steady state | Ongoing | BAU owner + quarterly refresh | SLO and benefit tracked |
BAU owner must be named before launch—not assigned in week 3 when hypercare ends.
Handover pack: adoption dashboard, champion list, training materials version, comms calendar, resistance register status.
Comms templates — launch, win, honest setback
Launch (users, ≤150 words):
Subject: Policy Assistant live in CRM — you stay in control
From [date], Policy Assistant appears in CRM for [workflow].
It drafts answers from approved policies—you review and approve before customer send.
Training: [link]. Champions: [names]. Office hours: [times].
Report wrong answers in two clicks — helps us fix quality.
Questions: [mailbox]
Win story (steering, with numbers):
Week 4: Site A — 54% assisted tasks; handle time −16% vs baseline (n=420).
Override rate 17% — healthy verification. Zero customer complaints on policy advice.
Honest setback (builds trust):
Week 2: Site B adoption 28% — below 45% hurdle. Root cause: DLP latency + supervisor KPI.
Actions: infra fix 9 Aug; KPI buffer restated by sponsor email 8 Aug. Wave 2 paused until Site B ≥45% two consecutive weeks.
Org design hooks for sustained adoption
Adoption dies when corpus owners and quality responders are nobody’s job.
| Role | Minimum allocation | Accountability |
|---|---|---|
| Product owner (AI) | 0.5 FTE | Backlog, prioritisation |
| Corpus curator | 0.3–0.5 FTE | Refresh, versioning |
| Quality triage | 0.2 FTE | Report-wrong queue |
| Change lead | 0.5 FTE waves | ADKAR, training |
| BAU service owner | 0.5 FTE post-launch | SLO, incidents |
If client cannot staff these, reduce scope—do not launch and hope.
Measuring resistance quantitatively
Track proxy signals—not only surveys:
| Signal | Interpretation |
|---|---|
| Login without assisted task | Activation failure |
| High override + low report | Distrust but no feedback channel |
| Shadow tool tickets | Urgent co-design need |
| Supervisor variance | Local KPI misalignment |
| Helpdesk “how to” spike | Training gap |
Review weekly in change standup (30 min) with PM, UX, L&D.
Cross-BU scale — playbook transfer
Scaling from one BU to three fails when playbook is slides not artefacts.
Transfer pack must include:
- Embed configuration recipe (not “ask IT”)
- Training lab environment and scripts
- Champion onboarding checklist
- Site-specific impact addendum template
- RAID seeds known from wave 1 (DLP, SSO, union)
Budget 2–3 weeks calendar per new BU for localisation—not copy-paste launch email.
Real-world scenarios
Scenario A — Contact centre: embed and supervisor alignment
Context: UK telco; 4,500 agents; policy copilot; standalone web app pilot; 31% weekly active after 30 days; handle time flat.
Diagnosis: 68% of agents never re-opened tab after day 3; supervisors measured AHT without learning adjustment; 14% used consumer GenAI shadow.
Intervention:
- CRM embed with SSO; remove standalone URL
- Supervisor playbook: 6-week learning buffer on AHT
- 180 champions (1:25); office hours 3×/week
- Comms: “Verify before send” campaign with real caught-error story
Outcome (90 days):
- Weekly active 58%; assisted calls 52%
- Median handle time −17% on assisted queue
- Shadow GenAI incidents −74% (security metric)
- Report-wrong rate 4.1% per 1k—healthy loop
Numbers: Benefit case £2.1m/year at 55% adoption; achieved 52% at day 90—on track.
Scenario B — Pharmaceutical field medical: MLR fear as resistance
Context: 800 field medical staff; AI draft assist; resistance from MLR and field directors fearing more review work.
ADKAR focus:
- Awareness: MLR director co-presents “draft only, you approve”
- Desire: Pilot shows −7 days cycle for field, not more MLR load
- Knowledge: Scenario training with intentional bad draft
- Ability: Veeva embed; template library only
- Reinforcement: MLR publishes monthly quality sample results
Outcome:
- Pilot adoption 74% of eligible requests assisted
- MLR cycle 18 → 11 days (−39%)
- Zero off-label auto-send; resistance narrative shifted by month 4
Scenario C — Public sector housing: union consultation and co-design
Context: 1,400 field operatives; scheduling assist; union sceptical on “algorithmic management.”
Change approach:
- Works council Week 0; algorithm inputs published
- Mandatory human override UI (topic 23); no keystroke surveillance
- 24 champions including 4 union health & safety reps
- KPI: first-visit fix, not “AI usage minutes”
Outcome:
- Neutral union statement; pilot 380 users
- Adoption 67% assisted jobs by day 60
- Override 22%—used to improve model, not punished
- Media risk avoided—no “surveillance” headline
Scenario D — Financial services AML: power users and compliance
Context: 220 AML analysts; narrative draft assist; 15 power users built Excel macros + ChatGPT workflows outside audit trail.
Intervention:
- Power users invited to champion council; features prioritized
- Official tool adds export formats macros lacked
- Compliance training: audit trail as enabler not blocker
- Adoption metric: % cases with approved draft in case system
Outcome:
- Shadow tool usage −81% in 12 weeks
- Draft prep time −35%; MLRO escalation time −18%
- Power users become trainers—12 named
Scenario E — Global manufacturer: scale without playbook transfer
Context: Plant A pilot 78% adoption; Plants B/C wave launched with email only; same vendor stack.
Failure: Plant B 19% adoption day 60; local ERP embed different; union not briefed in Germany site.
Recovery:
- 3-week pause; local impact assessment; German works council session
- Champions recruited per plant (1:20); embed recipe documented
- Wave calendar reset with ADKAR gates per site
Outcome:
- Plant B 52% adoption day 90 post-recovery
- Benefits tracker revised—enterprise case delayed 1 quarter but still positive NPV
- Lesson codified in cross-BU scale playbook (mandatory for future waves)
Cost: £95k unplanned change spend vs £400k if scale had continued blind to Plant B.
Practice exercises
Primary exercise — ADKAR checklist for 200 agents (45 minutes)
Brief: Roll out policy assistant to 200 contact-centre agents in one site; union informed not consulted; CRM embed planned; go-live 8 weeks.
Tasks:
- Complete ADKAR checklist per element with specific tactics (not “training”).
- Define champion model: count, selection, time allocation.
- Draft 30/60/90-day adoption metrics with red thresholds.
- Identify top 3 resistance drivers with mitigations and owners.
- Write supervisor playbook outline (≤1 page bullets): KPI buffer, coaching overrides.
Acceptance criteria:
- Each ADKAR element has dated deliverable
- At least one metric beyond login count
- Kill or pivot criterion at day 90 stated
- Union/informed status documented as Assumption with risk if upgraded to consult
Stretch exercise — Scale wave failure recovery (half day)
Brief: Wave 1 61% adoption success; Wave 2 different BU 29% at day 45; supervisor conflict; benefits tracker red.
Tasks:
- Root-cause analysis: embed, KPI, champion, UX, political (5 whys per).
- Revised wave plan with Kotter steps 5–7 explicit.
- Comms plan T+0 to T+8 weeks with messages per audience.
- Benefits model sensitivity at 29% vs 55% adoption.
- Steering recommendation: invest £120k in change vs pause scale—BLUF memo 7 lines.
Acceptance criteria:
- No villain narrative; structural incentives named
- Recommendation with trade-offs (£, time, risk)
- ADKAR gaps mapped to wave 2 failure
- Dashboard changes specified to detect early warning
Questions you should be able to answer
- Whose job changes, and how—is impact assessment published?
- What does adoption success look like at 30, 60 and 90 days—in metrics not adjectives?
- Where does the tool live in the workflow—is embed decision documented?
- Who are the champions, and is their time protected?
- What training proves ability—not only attendance?
- How are supervisors aligned so learning isn’t punished on KPIs?
- What are the top three resistance drivers and mitigations?
- How do power users and sceptics differ in engagement tactics?
- What kill or pivot criteria apply if adoption stalls?
- How is benefits realisation linked to adoption % in the financial case?
- What did union or works council consultation require—and was it met?
- How do override and report rates inform change and UX loops?
- What short-term win will you publish in week 4?
- Who owns BAU adoption after hypercare ends?
- How does wave 2 learn from wave 1—not repeat the same training video?
Adoption steering pack — what executives need monthly
Unlike technical steering, adoption reviews answer behaviour and benefit questions:
| Section | Content |
|---|---|
| BLUF | Adoption vs hurdle; recommend continue/pause scale |
| Cohort metrics | Active, assisted rate, task success, time-on-task |
| Segments | Sites or roles below red threshold |
| Resistance | Top 3 drivers + mitigation status |
| Champion health | Coverage ratio, burnout signals |
| Change spend | Training, embed, hypercare vs plan |
| Next wave | ADKAR gate date and prerequisites |
Anti-pattern: Showing login charts only—CFO asks “where is the £?” and meeting ends without decision.
ADKAR workshop facilitation guide (half day)
Agenda:
| Time | Activity | Output |
|---|---|---|
| 0:00 | Sponsor opens—why change, why now | Awareness anchor |
| 0:20 | Impact table walkthrough per role | Signed impact summary |
| 0:50 | Resistance brainstorm (silent sticky) | Resistance register v1 |
| 1:20 | Embed options decision | CRM vs Teams vs portal |
| 1:50 | Metrics and hurdles agreement | Dashboard definition |
| 2:20 | Champion nominations | Named list with manager OK |
| 2:50 | Training approach | Role-based modules |
| 3:20 | 30/60/90 plan | Dated actions |
Facilitator captures dissent—especially union or supervisor concerns—for steering visibility.
Union and works council FAQ themes (template)
Prepare answers before consultation—not improvised:
- What tasks change; what stays human-only?
- What data is logged; what is not monitored?
- How are errors corrected and who is accountable?
- What happens if the tool is wrong—disciplinary impact?
- Timeline, training, and opt-out during learning period (if policy allows)?
- Job impact: redeployment policy reference, not promises you cannot keep.
Legal review FAQ before publication.
Kotter step 5 enable action — removing barriers catalogue
| Barrier | Enable action | Owner |
|---|---|---|
| Extra clicks vs old workflow | Remove duplicate search step | Process + UX |
| SSO/login friction | Identity priority in RAID | IT |
| Slow response | Infra scale; cache corpus | Platform |
| Fear of monitoring | Publish “what we don’t log” | Sponsor + legal |
| No time to learn | Protected training hours | HR + manager |
| Broken citations | Quality sprint | Engineering |
Review barrier list at day 14 post-launch—users articulate barriers early.
Adoption experiment design (A/B ethically)
When comparing embed patterns or training variants:
- Hypothesis: “CRM embed raises assisted rate vs portal”
- Ethics: No withholding assist from control group in high-risk roles—use phased site comparison instead
- Duration: 4–6 weeks minimum
- Metric: Assisted task rate primary; handle time secondary
- Stop: If control site error rate rises, halt experiment
Document with DPO if personal performance data used in analysis.
Reinforcement tactics catalogue — month 2–6
| Tactic | When | Example |
|---|---|---|
| Leader story | Month 2 | COO email: user caught error via Report wrong |
| Metric transparency | Monthly | Dashboard on team board—not individual surveillance |
| Champion spotlights | Fortnightly | 5 min in huddle—peer demo |
| Process simplification | Month 3 | Remove redundant search step after embed stable |
| Refresh training | After UX change | 30-min lab on new citation UI |
| Benefit share | Quarter end | Steering celebrates handle time with adoption % footnoted |
Reinforcement without KPI alignment fails—tactics must match supervisor scorecard changes.
Change standup agenda (30 minutes, weekly)
- Adoption metrics vs red thresholds (5 min)
- Resistance register updates (5 min)
- Champion signals (burnout, gaps) (5 min)
- Training pipeline status (5 min)
- Embed/UX blockers to RAID (5 min)
- Decisions needed before next steering (5 min)
Attendees: change lead, PM, ASE, L&D rep, product owner—optional union liaison monthly.
30/90-day adoption scorecard (example thresholds)
| Metric | Day 30 target | Day 60 target | Day 90 target | Red if |
|---|---|---|---|---|
| Weekly active users | 45% cohort | 55% | 60% | <35% at day 45 |
| Assisted task rate | 30% eligible | 45% | 50% | <25% at day 60 |
| Task success | 75% | 82% | 85% | <70% any period |
| Median time on task | −5% vs baseline | −12% | −18% | No improvement day 60 |
| Override rate | 10–30% | 10–25% | 8–22% | <3% or >45% |
| Report rate / 1k | ≥2 | ≥2 | ≥2 | <0.5 |
Customize per workflow—publish thresholds before launch to prevent moving goalposts.
Sponsor coalition checklist (Kotter step 2)
- Economic buyer in coalition
- Operations/process owner active
- Risk approver engaged (not only informed)
- IT/platform lead committed to embed
- HR/L&D resourced for training
- Union/work council consulted where required
- Finance accepts adoption-linked benefits model
- Coalition meets monthly 30 min—not only steering spectators
Missing coalition member → adoption risk red on RAID until engaged.
Training module outline — contact-centre assist (example)
| Module | Duration | Content | Assessment |
|---|---|---|---|
| 1. Why and scope | 30 min | Sponsor video; what tool does/does not | Quiz 5 Q |
| 2. Trust UX | 45 min | Confidence, citations, refuse | Spot error in sample |
| 3. Workflow lab | 90 min | CRM embed exercises | 3 scenarios pass |
| 4. Report and escalate | 30 min | Report wrong; supervisor path | Submit mock ticket |
| 5. Accountability | 15 min | Approve before customer send | Signed acknowledgement |
Total 3.5 hours hands-on before production access—video-only fails high-risk roles.
Resistance heat map (quarterly refresh)
Plot stakeholders and sites on influence × stance:
- High influence / blocker: manage closely; 1:1 with sponsor present if needed
- High influence / supporter: use in comms and steering
- Low influence / blocker: peer champions; avoid public battle
- Low influence / supporter: seed champion pool
Sites with <40% adoption at day 60 get explicit entry on heat map—even if executive sponsor is green.
Negative cases — when change management fails
Launch and pray
Symptom: Go-live email; no champions; no office hours; embed missing.
Impact: <25% habitual use; benefits case dead; “AI doesn’t work here.”
Fix: ADKAR gate before launch; embed non-negotiable for phase 1.
Training video theatre
Symptom: 20-minute generic video; no hands-on; no competency check.
Impact: Users click through; errors rise; trust collapses after first mistake.
Fix: Role-based labs; simulated error detection; supervisor sign-off.
KPI sabotage
Symptom: Speed metrics unchanged during learning; supervisors warn “don’t slow down.”
Impact: Secret non-use; shadow tools; public failure narrative.
Fix: Learning buffer; quality metrics in scorecard; sponsor message to managers.
Champion burnout
Symptom: Volunteers with no time; blamed for IT issues; no escalation path.
Impact: Champion network collapses; resistance organises.
Fix: Charter, protected hours, exec sponsor access.
Surveillance adoption
Symptom: Dashboards track keystrokes, idle time, “AI reliance score” for discipline.
Impact: Union block; media risk; ethical breach.
Fix: Aggregate operational metrics only; transparency; consultation.
Scale before stabilise
Symptom: Wave 2 launched while wave 1 at 35% adoption.
Impact: Multiplied failure; wasted licence and change spend.
Fix: Wave gates on adoption hurdles; fix root cause first.
Benefits fiction
Symptom: 100% adoption assumed in business case; no measurement plan.
Impact: CFO blocks phase 2; programme labelled failed despite working tech.
Fix: Adoption sensitivity in case; dashboard from day 1.
Practice checklist
- I completed ADKAR checklist for a 200-agent rollout in writing
- I defined metrics beyond logins (assisted rate, task success, time-on-task)
- I identified champion ratio and protected time allocation
- I linked adoption % to benefits formula with sensitivity
- I documented top resistance drivers without blaming users
- I paired change plan with trust UX (topic 23) and delivery gates (topic 25)
- I opened Adoption and change for embed and operating patterns
Related playbook content
- Adoption and change — embed, training, hypercare and BAU ownership
- Consulting and client engagement — sponsor coalition and executive narrative
- User Experience and Human Factors — trust calibration and HITL
- Delivery and Programme Management — gates and dependencies for training tracks
- Stakeholder Management — union, supervisors, user council
- Communication and Executive Articulation — BLUF for adoption investment asks
- How to use this Learning Map
- 8D Framework
Discussion
Comments
Share feedback or questions about this page. No account required.
Loading comments…