Anonymized for teaching. Organization renamed to "Group A"; subsidiaries mapped to functional tags (Auto Units A/B, Consumer Finance, Financing, Logistics, Leasing, Data, Mobility, Distribution, HQ Security/Planning). System names replaced by functional descriptions. Not for performance review or client identification.
This report synthesizes Session B deliverables and 30-day action roadmaps across 11 teams, organized around four change management frameworks: ADKAR for team progress, Kotter for the group roadmap, Prosci stakeholder tools for assessment, and IBM's five principles for AI-specific dynamics. Qualitative findings are derived from self-reported readiness, not rankings.
① This course went beyond tool tutorials. Mapping to IBM (2025) principles in hindsight shows that the 3-line summaries, 2×2 critical engagement matrix, accountability tests, and final human authority gates share the same structure (§02). Group A is already driving change with standard global frameworks. ② The ADKAR progress matrix (§05) shows most teams stuck at Reinforcement: teams can ship a demo, but "who intercepts errors, what numbers to track, and who owns risk" remain theoretical. ③ Running 11 initiatives alongside existing digital transformations creates change saturation — we advise HQ Planning to pace deployment via a shared platform (§08).
The shared pain point across 11 teams is not building a demo, but "shipping a demo without knowing who catches errors or what numbers count as success." The next steps require only three things: stop tracking prompt counts or usage volume, and measure actual time saved and caught errors; clean documents into AI-readable formats before buying heavy knowledge management suites; and formally empower process rebuilders identified in this workshop to restructure workflows.
Change happens at the individual level. The five workshop modules map directly: Session A = Awareness, Red-Team Dual Inquiry = Desire (self-persuasion), NotebookLM build = Knowledge, Demo delivery = Ability, 30-day list + impact dashboard = Reinforcement.
Organization-wide 8-step roadmap. This workshop covered steps 1–6 (§07). Steps 7–8 (consolidating gains, anchoring in culture) define the next phase. Using Kotter alongside ADKAR is Prosci standard practice: Kotter guides the organization, ADKAR guides individual milestones.
Classifies stakeholders by impact × proximity to identify sponsors and build change teams. Supports 4 diagnostics: change impact, readiness, culture, and change saturation (§08 is critical for this group).
Key differences in AI transformation: start with "Why", anticipate resistance, treat AI as a capability rather than a project, balance governance with speed, and measure outcomes rather than activity. §02 compares these with our workshop design.
Methodologies align internal teams; when reporting to leadership, keep it plain: The bottleneck is never whether the model is smart enough, but unorganized documents and missing human guardrails. The first four steps got everyone aware, willing, and capable of shipping a demo. For the next 30 to 90 days, the only priority is making defensive checkpoints and outcome tracking real.
| IBM Principle | Workshop Design Feature | Evidence (Deliverables) |
|---|---|---|
| Start with "Why" | 3-line conclusions (why it matters / before vs after / next step), Investor Q1 "give one concrete number" | 11 team templates submitted; 10 completed |
| Anticipate resistance with predictive analytics | Critical engagement 2×2, Q1 "adversarial disengagement" — veteran silence is modeled in advance, not discovered after launch | Leasing CDP, Data metrics dictionary scenarios |
| AI is a capability, not a project | Copilot → Agent → Multi-Agent ladder: AI evolves decisions, workflows, and roles over time, not as a one-off project | Auto Unit A · Quality Agent prototype |
| Balance governance and speed | Accountability A/B, final human authority, confidence ladder, exception governance loop — embedded upfront without slowing delivery | Consumer Finance confidence ladder, Logistics policy baseline |
| Measure outcomes, not activity | Vanity metrics 2×2: "track business outputs, not question counts" maps directly to IBM's "avoid seat count or login frequency metrics" | Leasing outcome boundary, Mobility token cost analysis |
This table can serve directly as slide one for executive briefings: proving this was not a vendor tool demo, but structured capability building aligned with global standards.
This mapping is the cleanest opening for executive updates — proving this was structured capability building, not a vendor showcase. Immediately stop ranking employee prompt counts or token usage leaderboard; those are costs. Shift focus to business KPIs validated in the demos (e.g., ticket reduction, error interception rate).
| Cluster | Unit · Scenario | Impact × Proximity | Recommended Engagement |
|---|---|---|---|
| Governance Core | HQ Security (Security Assistant), HQ Planning (Shared AI Platform) | Group-wide × Highest | Natural fit for the Change Management team — already managing group standards and pipelines; recommend formal governance mandate and methodology training. |
| High-Consequence Frontline | Consumer Finance (Loan Docs), Financing (Vendor Knowledge Base), Logistics (Expense Claims) | Legal / Financial Risk × High | Deep involvement in solution design: their confidence ladders and exception loops should become group-wide templates. |
| Operations Frontline | Distribution (Competitor Q&A), Mobility (Support Self-Service), Leasing (CDP Onboarding) | Customer Facing × Medium | Co-create outcome metrics — this group is most vulnerable to vanity usage KPIs. Measure outputs, not questions. |
| Manufacturing Pioneers | Auto Unit A (Quality Agent), Auto Unit B (Owner's Manual), Data Unit (Metrics Dictionary) | Quality / Decision × Medium | Quality Agent is the group's only Agent prototype; use it as the benchmark for final human authority. Auto Unit B must first complete Awareness → Desire. |
Prosci Sponsor Identification note: all above are execution layers. Executive sponsors must be explicitly named and publicly committed in the next phase to address internal leadership friction first.
HQ Security and HQ Planning are natural change hubs: officially authorize them to manage component reuse and review pipelines so subsidiaries don't reinvent the wheel. Guardrails from high-risk teams (Consumer Finance, Financing, Logistics) should become group standards. For teams needing cross-department restructuring, stop sending them to classes; assign dedicated builders with authority to modify processes.
§03 identifies positions; this section defines resource allocation. Based on Mendelow (1991), 8 stakeholder groups are mapped across four quadrants with distinct engagement strategies.
Quadrant strategies detailed below. Point positions represent qualitative consultant assessments; red borders indicate direct targets for §09 actions.
High influence but low participation; the main source of unexpected friction (e.g., Distribution Q&A stalled on legal review). ⚠ Having sponsors in this quadrant is the project's biggest risk: the goal is to move ① into "Manage Closely" via public endorsement.
Priority group. ④ and ⑤ are natural change team leads (§03); ⑥ controls implementation resources. Include them in component review loops to trade ownership for commitment — turning them into change champions rather than gatekeepers.
Low priority, standard notices only. Watch one-way flows: as AI tools enter daily routines, some move right into "Keep Informed" (e.g., Financing's vendor tech transfer partners require cooperation).
Highly impacted but lack decision authority. ⑧ serves as the change network; keep them updated to maintain momentum. ⑧' requires special handling: 2×2 Q1 groups rarely voice overt opposition; they stay silent and bypass the tool. Focus groups recover their domain judgment far better than surveys (IBM: anticipate resistance).
Reading: Vertical = Power (top high), Horizontal = Interest (right high). Red highlights indicate targets requiring active repositioning: move sponsors to "Manage Closely", keep disengaged veterans from sliding to "Monitor". Update quarterly alongside saturation reviews (§08).
Key action from the matrix: Move Executive Sponsors into "Manage Closely" via public backing, or legal and financial gates will become silent blockers. For veteran frontline staff, skip survey forms — run direct focus groups to absorb their domain heuristics. Include subsidiary IT leads in component reviews to secure buy-in.
| Team | A Awareness | D Desire | K Knowledge | A Ability | R Reinforcement | Bottleneck Diagnosis |
|---|---|---|---|---|---|---|
| Auto Unit A · Quality Agent | ✓ | ✓ | ✓ | ✓ | ◐ | Final human review remains conceptual; multi-source rating variance unresolved in NotebookLM |
| Consumer Finance · Loan Approval | ✓ | ✓ | ✓ | ✓ | ◐ | RACI submitted; key-field review checklist pending formal sign-off |
| HQ Security · Security Assistant | ✓ | ✓ | ✓ | ✓ | ◐ | Subsidiary self-service adoption threshold unset; automated re-indexing triggers pending |
| HQ Planning · Shared Platform | ✓ | ✓ | ✓ | ✓ | ◐ | Component pipeline fully designed; live Security Gate execution needs proof |
| Logistics · Expense Claims | ✓ | ✓ | ✓ | ✓ | ◐ | Exception governance loop mature; monthly review cadence needs first cycle run |
| Leasing · CDP Onboarding | ✓ | ✓ | ✓ | ✓ | ◐ | Accountability A picked correctly, but output metrics undefined — classic R bottleneck |
| Data Unit · Metrics Dictionary | ✓ | ✓ | ✓ | ◐ | ◐ | Single author without review; new customer definitions lack cross-department consensus |
| Financing · Vendor Knowledge Base | ✓ | ✓ | ✓ | ◐ | ◐ | Tacit knowledge extraction on track; coverage rate and update schedule undecided |
| Mobility · Support Self-Service | ✓ | ✓ | ✓ | ◐ | ☐ | Rate card linting only conceptual; consumer token budget cap unset |
| Distribution · Sales Q&A | ✓ | ✓ | ◐ | ◐ | ☐ | 3-criteria template blank, Q&A script lacks legal sign-off — official path trailing shadow tools |
| Auto Unit B · Owner's Manual | ✓ | ◐ | ◐ | ☐ | ☐ | 3-line summary blank, before/after evidence absent — must revisit Desire; run 1-on-1 interviews to uncover real concerns |
✓ Achieved | ◐ Partial | ☐ Not Started. Basis: Session B submission completeness and 30-day checklist items. Prosci rule: If any ADKAR step is 0, overall progress is 0. Auto Unit B priority is not building a demo, but resolving Desire.
Non-negotiable defensive checkpoints and first actions for next week:
| Team | Content Readiness Structured Docs | Mechanism Readiness Human-AI Guardrails | Accountability Readiness RACI / KPIs | Quant Readiness Measurable Impact |
|---|---|---|---|---|
| HQ Security | High | Med | Med | Med |
| HQ Planning | High | High | High | Med |
| Consumer Finance | High | Med | High | Med |
| Logistics | High | High | High | Med |
| Auto Unit A | High | Med | Med | High |
| Leasing | High | Med | Med | Low |
| Data Unit | Med | Med | Med | Low |
| Financing | Med | Med | Med | Low |
| Mobility | Med | Med | Low | Low |
| Distribution | Low | Low | Low | Low |
| Auto Unit B | Low | Low | Low | Low |
Group-wide read: "Content Readiness" is highest, "Quant Readiness" is lowest — reflecting the R bottleneck across four dimensions. Corresponds to Prosci's readiness diagnostic; formal impact and culture reviews should follow in the next phase.
The heatmap plainly shows group status: Our document assets are well-prepared, but accountability and quantitative metrics are weakest. Do not rush broad rollout. Fix three items first: who owns AI answers, escalation steps for errors, and the exact business number defining success.
Kotter Steps 7–8 succeed by avoiding vague culture talk. Directly publish proven artifacts ("final human review", "exception loops", "confidence ladders") across subsidiaries via the shared platform, and formally write "measure outputs, not prompt counts" into policy. Learn management discipline and rollout sequences from top peers, not their budget scale.
Prosci's Change Saturation diagnostic warns: individuals have a ceiling for concurrent change. Teams already carry active digital transformation projects; frontline and middle managers (critical for adoption) bear overlapping pressure. Saturation shows up not as open resistance, but superficial compliance — leading to 2×2 Q4 "performative expansion".
HQ Planning's component pipeline acts as a pacing valve: centralize rollout timing decisions in the pipeline rather than letting subsidiaries start uncoordinated projects. Audit portfolio saturation quarterly (active changes × audience overlap); queue projects when thresholds are exceeded.
Saturation manifests as superficial compliance from frontline managers. Every AI project must define ROI thresholds and sunset conditions before launch — if expected results aren't measured within 30 days, sunset it immediately. Land a few solid projects rather than jamming dozens of parallel trials.
Reinforcement is not launching a dashboard, but recording the first real error catch: Quality Agent evaluator rejecting an AI recommendation, Consumer Finance catching a handwriting error, or Logistics reviewing an exception claim. Require one documented catch per team in 90 days — the MVP of measuring impact over activity.
Per Prosci: appoint executive sponsors with public backing; form the change team from HQ Security, HQ Planning, and cluster reps, complete ADKAR training, and manage cross-unit reuse (Kotter Step 7).
Rebuild Desire for Auto Unit B via 1-on-1 interviews; secure legal clearance for Distribution before expanding models — shadow tools won't wait; every week official tools lag, risks multiply.
This report's heatmap is qualitative; follow up with formal change impact and culture diagnostics, establishing quarterly saturation reviews for change tracking.
The 90-day focus: Track "evidence of the first real error intercept", not dashboard launch dates (e.g., evaluator rejecting an AI prompt, or loan officer catching a flawed OCR). Identify seed champions who actively restructured workflows and give them cross-department mandate. Official paths must be faster and more accurate than private workarounds, or shadow workflows will persist.
Web sources (Prosci/IBM/Moveworks) should be verified prior to presentations; citations reflect published framework versions.