Kotter's change model is useful for AI adoption when a change needs several teams to agree on a purpose, remove obstacles and support new ways of working. Its value lies in those decisions. A coalition with no authority to change the workflow, or a launch celebration with no continuing support, leaves the hard work unfinished.
Start with one defined change. For a commercial printer, that might be using an approved assistant to prepare production job sheets from customer briefs. Account service, production planning and IT need to agree on what the assistant may use, what a planner must check and how incomplete jobs reach the production floor. That is a coordination problem worth bringing people together to solve.
Use the model within your broader people and change approach. Keep the scope small enough to observe, and give continuing practice and maintenance an owner.
Decide whether the problem needs a coalition
Kotter's current methodology includes urgency, a guiding coalition, a shared vision, participation, removing barriers, short-term wins, continued effort and institutionalizing change. Its methodology overview distinguishes the original linear steps from later iterative accelerators designed for ongoing change. A rigid eight-stage launch checklist misses that evolution.
The guidance becomes useful when progress depends on decisions across boundaries:
- A shared reason. The sender wants a faster draft, while the receiving team needs fewer missing production details. Which problem will the trial address?
- Authority to remove an obstacle. A team cannot practise with customer briefs until someone settles permitted inputs and access.
- Support beyond the first attempt. Managers must make room for checking, corrections and feedback after the demonstration.
If one employee needs help with a particular task, a large coalition may add little. ADKAR can help identify individual support needs. If a necessary action is not happening, COM-B can help examine its conditions. Choose the framework for the decision in front of you.
There is also an evidence boundary. McDonnell and colleagues' 2026 systematic review examined 17 case studies of digital and virtual healthcare change. It found limitations in linear application, attention to people and processes, and monitoring. The case reports offered limited evidence of how the models themselves produced success beyond providing a roadmap. These heterogeneous healthcare cases do not establish that Kotter improves enterprise generative AI adoption.
Build the group around decisions it can settle
Suppose Alex Morgan in account service receives a customer's brochure brief. It specifies paper stock, quantity, finishing and a delivery date. Alex wants an approved assistant to turn those details into a draft job sheet, the instructions production uses to prepare the order.
Jamie Lee, the production planner, knows that a missing finishing instruction can delay the press schedule. Sam Taylor in IT knows which customer information the approved tool may receive. Their manager can arrange cover while they test and check the new process.
These people have different reasons to join. Select them because their knowledge or authority can resolve something that would otherwise stop the work.
| Participant | Decision needed | Evidence to bring |
|---|---|---|
Account service | Which customer details must be preserved, and how should an unclear request be raised? | A permitted brief and a previously clarified customer request. |
Production planner | What makes the sheet usable, and which omissions must stop the handoff? | An accepted job sheet and an example that needed correction. |
IT or information owner | Which inputs and tool settings are permitted? | The approved-use rules and the actual access route. |
Manager or sponsor | Who has time to practise, check and resolve exceptions? | The work schedule and authority to change competing commitments. |
A senior supporter can help secure decisions, but seniority alone does not supply knowledge of the handoff. Include the people who will receive and check the output. Give the group a specific unresolved question, such as whether an unclear finishing instruction must go back to account service before a job is scheduled.
The answer may be to retain the existing process for that case. Removing a barrier can mean changing the proposed use, rather than persuading people to accept it.
Make urgency about a problem people recognize
An approaching delivery deadline is a reason to improve coordination. It is also a reason to be careful about disrupting a working process. Explain what is going wrong, what the trial might improve and what remains open to challenge.
In a September 2026 Reddit discussion about two company-wide tool rollouts, a Reddit user at a media company described a previous switch that killed the old system overnight. It worked technically, but left frustration behind. They were asking whether a full Kotter process was proportionate for a scheduling platform followed by a content management platform.
The replies disagreed about how much structure was needed. Some supported Kotter as straightforward guidance; another stressed that the middle of change is messy and support must continue after go-live. This was a discussion of general technology change, with no reported outcome for the proposed rollout. Its useful question is whether the change process fits the disruption people will experience.
For the printer, discuss the cost of incomplete job sheets with account service and production. Ask what else those teams are changing at the same time. A credible reason to try the assistant does not require presenting refusal as a lack of commitment.
Define a win across the whole handoff
A quick draft is encouraging. It becomes a useful short-term win only if the finished work meets the receiving team's needs at an acceptable cost.
Agree on a comparison before the trial. Use comparable permitted briefs and the existing process, with the same required checks. Record:
- The finished output. Does the sheet preserve the customer's instructions and mark unresolved details clearly?
- The complete effort. Include drafting, checking, clarification and corrections on both sides of the handoff.
- The consequence. Can production use the sheet, or does it need to return to account service?
Keep unsuccessful and incomplete attempts in the result. A good demonstration can justify another bounded attempt; it cannot establish that every job should move to the new process. The guide to running a credible AI pilot explains how to choose a comparison without losing failed or missing work.
Give continuing change somewhere to live
In an April LinkedIn post, advisor and author Paul Gibbons described returning to IBM and finding a change model similar to one he had helped develop at PwC in the 1990s. He argued that episodic models have a narrower role when AI adoption is continuous and depends on behavior and trust. His post introduced his own proposed approach, so it is practitioner commentary rather than comparative evidence of model effectiveness.
The concern is useful even though current Kotter guidance explicitly allows iteration. The question for the printer is where continuing work belongs once the initial coalition has settled the trial's decisions.
Give recurring issues an ordinary route:
- A new kind of customer brief. Account service and the planner review whether it fits the supported use before treating it as routine.
- A tool or access change. IT checks permitted inputs and the workflow before staff resume the affected use.
- Repeated omissions or rising checking effort. The planner brings actual examples to the process owner, who can revise or pause the practice.
- A new colleague or absent reviewer. The manager arranges practice and a capable reviewer before relying on the new handoff.
Keep that responsibility in the team's work schedule and operating instructions. A volunteer champion may contribute feedback, but cannot be the only person keeping the process safe and useful.
Questions and answers
When is Kotter useful for an AI rollout?
Kotter is useful when an AI rollout needs coordinated decisions across teams, such as agreement on purpose, permitted inputs, output requirements and capacity for practice. Start with a defined workflow and a real obstacle. Build a group with the knowledge and authority to resolve it, then check whether its decisions help the next attempt. The coalition decision table shows how this works for a printing job handoff.
Do you need to follow all eight steps in order?
Kotter's original model described a linear sequence, while its current methodology describes iterative accelerators. For AI adoption, use the guidance to inspect what the change needs and revisit decisions as evidence changes. Do not treat a completed stage checklist as proof of useful adoption. In a job-sheet trial, discovering an unclear input may require revisiting scope or access before more training can help.
Can Kotter support continuous AI adoption?
Current Kotter guidance is designed to address ongoing change, but a framework does not assign your organization's continuing work. Name who maintains the workflow, reviews tool changes, supports new staff and responds to failures. Give those people time and authority to act. The continuing-change section identifies triggers worth putting into ordinary team responsibilities.
Who should join an AI adoption coalition?
Include people who understand the task and its consequences, people who receive or check the output, and owners who can settle access, capacity or process decisions. Membership should follow the unresolved work. For a printer's job-sheet trial, that means account service, production planning, the information owner and a manager able to protect checking time. A sponsor can help secure decisions; frontline knowledge remains necessary to judge whether the handoff works.
