Change Management And AI Literacy

Why Change Management Comes First

In week 7 you mapped a process and decided where AI helps. A map changes nothing until people use the new way of working. Many AI projects deliver a working tool and then see low use, because staff do not trust it, do not know how to use it, or fear for their jobs. Change management is the planned effort to move people from the old way to the new way. It deserves the same attention as the technical build.

Leaders and employees see the same project differently. A leader sees cost, speed and competitiveness. An employee sees a tool that might change their tasks, their status or their job security. Both views are reasonable. Good change work states both and answers both.

The Adoption Curve

Everett Rogers described how new ideas spread in his book Diffusion of Innovations. He sorted people by how early they adopt: innovators, early adopters, the early majority, the late majority and laggards. The first two groups are small. The two majority groups are large, and a tool succeeds in a company only when the early majority joins.

Early adopters try things because they are curious. The early majority waits for proof that a tool works for people like them. This tells you what to do: give early adopters room to experiment, then publish their results in plain language, with real examples from your own teams. A story from a colleague in the next office convinces better than a vendor demo.

The curve is a guide, not a law. Within one company, people can be early adopters of drafting tools and late adopters of anything that touches their core expertise. Track adoption by team and by use case, not by a single company number.

AI Literacy And Training Plans

AI literacy is the ability to understand what AI tools do, use them well, check their output and know when not to use them. The EU AI Act includes a duty for providers and deployers to take measures so that their staff have a sufficient level of AI literacy. Treat it as a standing programme, with a record of who completed what.

One course for everyone fails, because roles differ. Plan three levels. Level one is for all staff: how models work at a high level, what they get wrong, what data may never be entered, and how to check output. Level two is for daily users: prompt writing, using approved tools, and writing a review step into their own work. Level three is for builders and owners: evaluation, monitoring, security and the rules from weeks 5 and 6.

• Level 1, all staff: basics, limits, data rules, checking output (about 2 hours)
• Level 2, daily users: practical use inside their own workflow (about 6 hours plus practice)
• Level 3, builders and owners: evaluation, monitoring, governance (about 2 days)
• Refresh every six months, because tools change

Role Redesign

AI changes tasks before it changes jobs. Take a role and list its tasks. Mark each task as handled mostly by AI, shared between AI and the person, or kept by the person. A support agent may spend less time typing replies and more time on difficult cases and on checking drafts. A financial analyst may spend less time collecting data and more time asking why numbers moved.

Write the new role down: the tasks, the skills, the quality standard and the career path. If the draft shows that a role shrinks, say so early and offer a route: retraining, a move to a growing team, or a changed scope. Silence breeds rumour, and rumour is more damaging than a hard fact stated with a plan. Involve HR, the works council or employee representatives where your country or contract requires it, and take local labour law advice.

The Champions Network

A champion is a respected colleague in a team who learns the tools early, helps neighbours, and reports problems back to the project. Champions are the most reliable way to reach the early majority, because people ask the person at the next desk before they open a manual.

Choose champions for credibility, not for seniority or technical skill. Aim for roughly one champion for every twenty to thirty staff in a team, then adjust to what you observe. Give them protected time, for example a few hours a week, a direct channel to the project team, early access to new features and visible recognition. Without protected time a champions network turns into unpaid extra work, and it fades within months.

• Duties: demonstrate, answer questions, collect feedback, share good prompts
• Support: weekly office hour, shared channel, training for champions
• Recognition: named in team updates, counted in performance goals
• Rotation: new champions join each quarter so the load is shared

Understanding And Handling Resistance

Resistance is information. It usually has a reason: fear of job loss, doubt about quality, extra work during the transition, loss of status for experts, or a past tool that failed. Ask people what worries them, listen without defending, and sort the concerns into those you can fix, those you can explain and those that are valid and need a decision.

Do not claim that no jobs will change if some will. Do not promise that AI output is always right. Say what you know: what the tool does, what it gets wrong, who checks it, and what the company commits to. Prosci's ADKAR model gives a useful checklist for each person: Awareness of the need, Desire to take part, Knowledge of how, Ability to do it in practice, and Reinforcement so that it lasts. Find the first step in that list where a person is stuck and help with that step.

Measuring Adoption Honestly

Count more than logins. Useful measures include the share of eligible staff who use the tool each week, the share of tasks where the tool is used when it could be, quality scores from sampled outputs, time saved reported against a baseline, and staff confidence in a short survey. Run the same survey before the pilot and after, and report negative findings to the steering group along with the positive ones.

Watch for hidden problems. Staff may use unapproved tools if the approved one is slow or weak. Some may use the tool and not check its output. Others may stop using it after one bad answer. Each of these is a training or design issue and not a character flaw. Fix the cause, and make it safe to report errors without blame.

Leaders carry a special duty in this work. Staff watch what leaders do more than what they say. When an executive uses the approved tool openly, shares a mistake the tool made, and asks a team how the change affects them, trust grows. When leaders announce a programme and then stay away from it, staff conclude that it is optional. Ask each sponsor to commit to a few visible actions in the first ninety days, such as attending a training session and joining one champions meeting.

Key Terms

Change management

The planned effort to move people and teams from an old way of working to a new one.

Adoption curve

A model of how a new tool spreads from a few early users to the majority.

AI literacy

The ability to understand AI tools, use them well, check their output and know their limits.

Role redesign

Rewriting a job's tasks, skills and standards after AI takes over or shares some tasks.

Champion

A respected team member who learns tools early and helps colleagues use them.

Resistance

A person's reluctance to change, which usually points to a real concern.

ADKAR

Prosci's model of five steps for individual change: awareness, desire, knowledge, ability, reinforcement.

Works council

An employee body that must be informed or consulted on workplace changes in many European countries.

Shadow AI

Use of unapproved AI tools by staff outside company control.

Exercise: Draft A Literacy And Champions Plan

1. Pick one team of 20 to 60 people and list its roles.
2. Assign each role to literacy level 1, 2 or 3 and set hours for each.
3. Pick the process from week 7 and write the new task split for one role.
4. Name two possible champions and write what protected time you would give them.
5. List the three strongest objections you expect and write an honest answer to each.
6. Choose four adoption measures and a date to review them.

Further Reading

• Diffusion of Innovations, Everett M. Rogers, Free Press
• Leading Change, John P. Kotter, Harvard Business Review Press
• ADKAR: A Model for Change in Business, Government and Our Community, Jeff Hiatt, Prosci
• Regulation (EU) 2024/1689, the Artificial Intelligence Act, article on AI literacy, Official Journal of the European Union