
Digital transformation
Change management for digital transformation: ADKAR and Kotter
Use ADKAR and Kotter to understand why people struggle with a new system, choose the right support and check whether the change works.
Imagine a purchasing system that goes live on time. Three months later, people still keep a separate spreadsheet because the new approval route cannot handle the exceptions they see every day.
Change management prepares and supports people to work differently. That includes explaining the reason, involving them in the design, providing practice and resolving the conditions that pull them back to the old method.
Kotter's eight steps help organise the wider change. ADKAR helps identify what an individual needs to use it. Together, they give the team a way to investigate the spreadsheet problem instead of treating every workaround as resistance.
Two models, two jobs
| Kotter's 8 steps | ADKAR | |
|---|---|---|
| Level | The organisation | The individual |
| Answers | "What must leadership do, in what order?" | "Where is this person or team stuck?" |
| Origin | John Kotter, Harvard Business Review, 1995 | Jeff Hiatt, Prosci, developed in 1996 |
| Best used for | Planning the program's change activities | Diagnosing resistance and targeting help |
The models ask different questions. An organisation can have a clear plan while a particular shift still lacks the support to use it.
Kotter's eight steps, in a digital program
| Step | What it looks like in practice | Warning sign it was skipped |
|---|---|---|
| 1. Create a sense of urgency | Show the cost of today's process in numbers: hours, errors, lost customers | People ask "why are we doing this?" at go-live |
| 2. Build a guiding coalition | A sponsor with authority, plus respected people from each affected team | The program is seen as "an IT project" |
| 3. Form a vision | One sentence on how work will be different, and why that's better | Every presentation explains it differently |
| 4. Communicate the vision | Repeated, by line managers, in team meetings, not only by email | Teams hear about changes from rumours |
| 5. Remove obstacles | Fix the old incentives, reports and approvals that pull people back | "The new tool is fine, but my boss still wants the Excel" |
| 6. Create short-term wins | A visible, measured improvement within the first few months | Momentum fades after launch |
| 7. Build on the change | Use early wins to take on the harder processes | The program declares victory at go-live |
| 8. Anchor it in the culture | New ways become the standard: in onboarding, KPIs and job descriptions | Six months on, the old way creeps back |
Kotter's original argument, drawn from more than 100 companies, was that skipping steps creates only the illusion of speed. He has since reworked the steps into eight "accelerators". For our purchasing system, step 5 deserves a close look: changing the screen will not help if an unresolved approval rule still sends people to a spreadsheet.
ADKAR: what each person needs
- Awareness of why the change is needed. Do I understand the reason?
- Desire to take part. Do I want to, and what's in it for me?
- Knowledge of how to change. Do I know what to do differently?
- Ability to do it in real work. Can I actually do it, at full speed?
- Reinforcement to keep it going. Do the support, feedback and working rules help me continue?
Someone may know which buttons to press without understanding why the process changed. Someone else may support the change but lack time to practise. Those situations need different help.
Find the barrier point
Prosci uses the barrier point to identify the earliest ADKAR element holding a person back. In the fictional assessment below, we use scores from 1 to 5 and flag the first score of 3 or lower.
Use a score to start a conversation. A team average can hide very different experiences, so ask about the task, shift and support behind it.
Reading the chart:
- Finance has no flagged score in this example. Check actual use and keep support available.
- Sales has its first flag at Awareness. Discuss the reason for the change before assuming another training session will resolve it.
- Warehouse understands and wants the change, but is stuck at Knowledge. They need hands-on training on their actual tasks.
- HR knows how, but is stuck at Ability. They need practice time and someone on hand in the first weeks.
A shared rollout plan can include these different responses. It needs enough detail to give each group the help it actually needs.
Match the action to the barrier
| Barrier | Typical symptom | Help to consider | Often insufficient alone |
|---|---|---|---|
| Awareness | "Why are we even doing this?" | The sponsor and direct managers explaining the reason, with numbers | A newsletter without a chance to ask questions |
| Desire | "I get it, but not for me" | Answering "what's in it for me", involving people in the design, removing personal downsides | More facts about the business case |
| Knowledge | "I don't know how to do my job in it" | Role-based training on real tasks, just before go-live | Generic training months early |
| Ability | "I know how, but it's slow and I get stuck" | Practice, floor support, super-users, time to learn | Assuming training equals ability |
| Reinforcement | Old workarounds come back | Reports and support that use the accepted method; retire old routes after testing the replacement | Declaring victory at go-live |
Check adoption after go-live
Track whether people can complete the work correctly, as well as whether they open the tool. These fictional targets illustrate a review; set the period and thresholds for your process:
| Measure | What it tells you | Example target |
|---|---|---|
| Usage | Share of intended users active each week | Above 90% by week 6 |
| Proficiency | Errors, rework and time per transaction | Time per case back to baseline by week 8 |
| Workarounds | Shadow spreadsheets, emails and manual steps still in use | Old process switched off by week 12 |
| Speed of adoption | How quickly each team reaches the target | No team more than 4 weeks behind |
| Business outcome | The benefit the change was for | Cycle time, cost or error rate from the business case |
Break the measures down by team and shift. Investigate errors and workarounds with the people involved before deciding that low usage is the problem.
Helping employees adopt AI at work
Take a fictional purchasing team trying AI to draft supplier-status updates. The buyer still checks the order record and approves the message. The tool cannot change an order or send an email itself.
That task needs more than a demonstration of a good prompt. The team must know which information it may use, how to check the result and what to do when the evidence is missing.
| Part of the trial | What people practise |
|---|---|
| Permitted data | Use the approved tool and records allowed for this task; leave restricted information out |
| Verification | Compare quantities, dates and supplier claims with the current order record |
| Missing or conflicting evidence | Leave the claim out, flag the gap and ask the responsible buyer |
| Approval | A named buyer reviews the draft before anything reaches the supplier |
NIST's AI RMF calls for trained people, clear oversight responsibilities and monitoring after deployment.
Compare completed, correct updates and total effort with the current method, including checking and corrections. Keep unsuitable cases out of the trial. Higher usage is useful only when the work improves; expanding a tool that creates extra checking would miss the point.
The AI planning example follows the same distinction between preparing a recommendation and authorising a commitment.
Why the 70% and 84% failure claims need context
These figures describe different sources and definitions. They cannot tell you the probability that your rollout will fail.
| Source | What it reports | What the headline misses |
|---|---|---|
| BCG, 2020 | 70% fell short of objectives; research combined 825 executive responses with experience from 70 companies | This includes projects that created value but missed targets, not only projects with no benefit |
| McKinsey, 2018 | 16% reported improved performance and capacity to sustain it; the survey had 1,793 respondents, including 1,521 with transformation experience | Another 7% reported improvement that did not last. Missing the full success definition is not the same as achieving nothing |
The older 84% headline in Forbes comes from a 2016 interview with Michael Gale. It does not provide enough study detail to establish a general failure rate. It also predates the McKinsey survey, so it cannot be explained as simply 100% minus that survey's 16%.
For your own rollout, define success in the process: correct transactions, manageable effort and the outcome the change was meant to improve. The benefits realisation note shows how to follow that beyond launch.
A 90-day plan for your next rollout
- Before launch: name the sponsor, write the one-sentence vision, run a first ADKAR survey, and plan the obstacles you'll remove (reports, approvals, incentives).
- Launch: role-based training right before go-live, super-users in every team, visible sponsor presence.
- Weeks 2 to 6: track usage weekly, run a second ADKAR survey, and act on each team's barrier point.
- Weeks 6 to 12: review the result, retire old routes only when the replacement works, and include the accepted method in onboarding and KPIs.
Give the operating owner time and resources for that follow-up. If the purchasing team still needs its spreadsheet, find the unresolved case and fix or accept it before calling the new process complete.
Sources
- Kotter, J. P. (1995). Leading change: why transformation efforts fail. Harvard Business Review, March to April 1995.
- Kotter Inc. The 8 steps for leading change.
- Hiatt, J. (2006). ADKAR: A Model for Change in Business, Government and Our Community. Prosci.
- Prosci. ADKAR barrier points, individual assessment and the first score of 3 or lower.
- Prosci. The Prosci ADKAR Model.
- NIST. AI RMF Core, Govern 2.2 and 3.2, Map 2.1 and Manage 4.1. Training, task definition, oversight and monitoring.
- Bruce Rogers, Forbes (2016). Why 84% Of Companies Fail At Digital Transformation, interview with Michael Gale; the article does not establish a universal failure rate.
- Boston Consulting Group (2020). Flipping the odds of digital transformation success.
- McKinsey (2018). McKinsey's survey of digital transformation practices.
The teams, scores, purchasing scenarios and adoption targets are fictional. The AI trial is proposed guidance, not an observed performance result.