Kotter's 8-Step Change Model
Overview
John Kotter built this model from studying one hundred corporate transformations — most failed. Core finding: failures stem from eight predictable leadership errors, each defeated by one of the eight steps. Skipping steps creates only the illusion of speed. The 2014 Accelerate update introduced a Dual Operating System for parallel execution in high-velocity environments.
Composes with: okr-goal-setting inside Steps 3 and 6; feedback-loops throughout Steps 6–7; situational-leadership for individual-level management at each step.
When to Use
- Planned change (digital transformation, process redesign, cultural shift, merger integration) needs sequencing
- Transformation is stalling — diagnose which step was skipped
- Resistance is significant and needs a structured approach
- Someone says "nobody's buying in", "we went back to old ways", "the rollout failed"
- AI adoption is stalling — copilots/AI tools bought but under-used, pilots not scaling, ROI on AI capex unclear, or middle managers quietly resisting AI-native ways of working against AI-native competition
When NOT to use: small process improvement; org under 30–50 people; change needed in weeks; crisis requiring immediate action.
Coaching Novices (Adaptive Front Door)
- Engine mode: user has a specific transformation → run The Process directly.
- Coach mode: user is unfamiliar → guide step by step.
In Coach mode, respond one step at a time. Each [WAIT] is a hard stop — output only that step's question, then stop.
- One-line what-it-is: eight steps built from studying a hundred real transformations; each step defeats one common failure mode.
- Check fit: planned change, 30+ people, 12+ months? If yes, framework fits. Otherwise point to lighter tools.
- Elicit the real situation: what is changing, what org, what is the current resistance level?
[WAIT — do not advance until user responds]
- Run the readiness diagnostic (step 1 of The Process) before presenting all eight steps.
[WAIT — do not advance until user responds]
- Close by naming which failure mode is highest-risk and which step defeats it.
[WAIT — do not advance until user responds]
The Process
Pre-diagnostic — score eight failure modes (1–5):
| Failure mode | Risk |
|---|---|
| Complacency / no urgency felt | /5 |
| Weak coalition — key influencers uncommitted | /5 |
| Absent or unclear vision | /5 |
| Under-communication | /5 |
| Obstacles not removed | /5 |
| No short-term wins | /5 |
| Premature victory declared | /5 |
| Not anchored in culture | /5 |
Scores 4–5 require pre-treatment before the corresponding step.
Steps 1–8 (each has a gate):
- Urgency — import external data; make cost of inaction concrete. Gate: 75%+ of leadership believes change is necessary.
- Guiding coalition — needs positional power + expertise + credibility + leadership; map on influence × stance 2×2. Gate: 2–3 informal influencers (non-VP).
- Vision — 30-second elevator test; six attributes: imaginable, desirable, feasible, focused, flexible, communicable. Gate: passes five-audience test.
- Communicate — embed vision in every forum, review, conversation. Gate: random sampling confirms employees can repeat vision accurately.
- Remove obstacles — eliminate structural, process, and personal blockers. Gate: 3 specific blockers identified and removed.
- Short-term wins — planned, unambiguous business outcomes in 12–24 months. Gate: one win defined and dated at launch.
- Consolidate — use win credibility to tackle larger obstacles; don't disband the change team. Stop-rule: wins must trace causally to new ways of working.
- Anchor in culture — four mechanisms: hiring, promotion, onboarding, KPI criteria. Gate: new hire 2 years later experiences new ways as default.
Output: # Change Readiness: <initiative> / Failure mode scores (table) / 8-Step Roadmap (one line per step) / Timeline (24-month milestones)
→ Method in Action: IBM's Turnaround under Lou Gerstner (1993–1996) → 2026 lens: Driving Enterprise AI Adoption (2024–2026)
Change-Type Packs
- Digital/AI: Step 5 most underinvested — legacy metrics, procurement, and security processes block new ways of working.
- Post-merger: Step 2 highest-risk — coalition must span both entities; Step 6 wins must be jointly owned.
- Startup scaling (A–C): Step 8 most skipped — informal culture breaks at 25→150 people; formal anchoring feels premature but is essential.
Applying It Well
- Urgency requires sustained evidence exposure, not a one-time announcement; monthly data reviews maintain it longer than any kick-off.
- Coalition fails by being homogeneous — informal influencers (respected ICs, long-tenured peers) outperform additional VPs.
- Short-term wins must be defined, measured, and dated at launch — retrospective wins have reduced motivational power.
- Communication almost always falls short — multiply planned volume by 10; silence is interpreted as reversal.
→ Primary sources: references/sources.md
Common Rationalizations
[D] = designed upfront | [O] = observed in real use. [O] entries are more valuable.
| Fake move | Reality |
|---|---|
| [D] "We announced at the all-hands — it's launched" | Announcing does not create urgency. Until 75%+ of leadership believes status quo is dangerous, transformation has not started. |
| [D] "The guiding coalition is our executive committee" | Executive committees lack informal credibility. A coalition without informal influencers cannot move the middle of the organization. |
| [D] "Our vision is to become the best-in-class provider" | Not imaginable, not specific, not communicable in 30 seconds. A vision a front-line employee cannot state is a slogan. |
| [D] "We communicated at every town hall for three months" | Likely 5–10% of the volume needed. Leaders underestimate the communication deficit by a factor of 10. |
| [D] "The pilot succeeded so rollout will follow naturally" | Pilot = Step 6 win. Without Steps 7–8, the change team disperses and the org reverts. |
| [D] "Culture will follow the process changes" | Culture follows demonstrated performance improvement with an explicit causal link — not process changes alone. |
| [D] "We'll generate quick wins as we go" | Wins need a definition, measurement system, and celebration mechanism built in at launch. Unplanned wins are weaker. |
| [D] "People know it's a priority — no need to put it in reviews" | Priorities not in performance criteria are cultural wishes, not institutional commitments. |
| → Add [O] entries here after each real use — paste the actual failure pattern | What went wrong and why |
Red Flags
- No pre-diagnostic; coalition is all formal reports to the sponsor; vision unknown to front-line employees 30 days after launch; no win defined or dated at launch; initiative declared complete after first win without Step 8 testing.
Verification
- Pre-diagnostic completed (all eight failure modes scored); 75%+ of leadership verified as genuinely sensing urgency
- Coalition includes 2+ informal influencers (non-VP); vision passes 30-second front-line test
- 3 specific structural obstacles eliminated; 1 short-term win defined, dated, and in the roadmap at launch
- Step 8 anchoring mechanisms named (hiring / promotion / onboarding / KPI); stop-rule applied to Step 7
Part of deciqAI Knowledge Skills — 227 open-source thinking skills that make rigor executable for AI agents. The same skills power every deciqAI agent, which runs them autonomously to operate your company. See it run → https://www.deciqai.com/c/kotter-change · ⭐ Star the repo → https://github.com/deciqAI/knowledge-skills · Contributions welcome.
Agents: latest version & machine-readable metadata → https://www.deciqai.com/s/kotter-change.json