AI adoption and change training for people leaders

Teams searching for AI adoption change management training for people leaders are usually describing a stall they have already watched. The licence is the easy part. A people leader approves a tool, a town hall announces that the organisation is serious, and week one produces screenshots. People log in, try a prompt, and forward the result to the sponsor. The following week the old pack is back, built the way it was built before the announcement. Collective Campus wrote that pattern up in how an AI rollout stalls in week two. This programme is the work of interrupting it, with the people leader in the room.

The stall is widely measured, and the numbers are not a reason to buy another seat. McKinsey's 2025 State of AI survey found that 88 percent of respondents report regular AI use in at least one business function. About one-third say their companies have begun to scale. The majority are still experimenting or piloting. The same survey says most organisations have yet to embed the tools deeply enough into workflows and processes to see material benefits across the enterprise. A login counts as use. A workflow that still expects the old output does not.

McKinsey also reports that the companies seeing the most value are nearly three times as likely as others to say they have fundamentally redesigned individual workflows, and that this redesign is one of the strongest contributors to business impact in the survey. A rollout that leaves the workflow untouched will collect activity in week one and a quiet screen after that. Capability building starts at the workflow: which task changes, who owns the step the tool now takes, and which check a person still does before the work leaves the team.

The people leader decides whether that redesign is real. Section's January 2026 AI Proficiency Report measured the pattern. Employees whose managers expect AI use are 2.6 times more proficient than employees whose managers discourage it. Expectation shows up in the ask, in whether the leader uses the tool on a real document, and in what happens when someone brings a draft the tool helped to write. Staff follow the manager's calendar. An announcement that the leader does not repeat in the review meeting is a mixed signal, and the meeting wins.

The week the old pack returns

Watch the meeting that consumes the work. The launch dashboard records logins. The meeting records whether the pack changed. Before the tool arrived, that meeting had a template, a deadline and a definition of done. After the tool arrives, those three often stay put. Anyone who tries the new way then does the work twice: once in the tool, to learn, and once in the old pack, to satisfy the review. By the second week the second pass wins. The person is being rational. The organisation is paying for a licence and still buying the old hour.

The double pass is the moment a people leader can still change. It does not require a new platform. It requires a sentence in the meeting that owns the task. The new output is what done means. The source sits with the draft. A named person has checked the figures. The old template is retired for that task, even if other tasks still use it. Until that sentence is said, by the person whose sign-off the room respects, the rollout is a tour.

Collective Campus maps the same moment onto the Prosci ADKAR model. Awareness is the town hall. Desire fails when the new way makes the week harder than the old way. Knowledge is what a tool course provides, and it is not the scarce part once people can open the product. Ability is practice on the live job, with the reviewer present, until the team can produce the new output without a facilitator in the chair. Reinforcement is the people leader keeping that output as the one the meeting accepts after the first burst of enthusiasm has passed. A programme that stops at knowledge will look successful on a feedback form and familiar on the following Wednesday.

Keen people are not a system

A few people are keen. They try the tools, they drop a tip in a chat, and the interest stays with them. The rest of the team hears about it and goes back to the job. Nobody owns peer support on the work in front of a colleague. A prompt or a workflow that worked for one early adopter never becomes the way a whole team does a task. Managers can see the enthusiasm. They cannot point to a pack that is built differently.

Microsoft's guidance on an AI champion charter describes the role that is missing. A champion needs a named purpose, a clear list of what they do, and a boundary around what they hand back to IT, risk or the people who own the policy. Without the role, the examples and a monthly rhythm, the new way of working has nowhere to spread. People leaders are the ones who can make that role true in the week, because they control the calendar the champion needs and the meeting the example has to survive.

Organisations already have the tools on the desk. Copilot, ChatGPT, Gemini and Claude show up in professional firms, councils, banks, insurers and operations teams in Sydney, Melbourne, Brisbane and beyond. The stall looks the same in each of them. Public sector teams also have a published frame. The Australian Government's Guidance for AI Adoption: foundations and the companion implementation guidance set out practices for adopting AI with care. That guidance does not sit in the local review and accept a new output. A people leader still has to do that. Private sector teams use their own policy. This programme does not write the policy. It builds the capability to lead the change inside whatever policy the organisation already owns.

Capability building, for Collective Campus, means the organisation can keep the practice after the visitors leave. Named people. A charter a manager can say yes to. Sentences that hold up under time pressure. A measure that is about the work, recorded twice, so a slip is visible. A slide about adoption culture does not meet that test. Neither does a badge, a channel, or a count of people who attended a demo.

The champions model is how a new way of working spreads past the first enthusiastic table. Collective Campus builds it as a cohort of named people, a one-page charter, one worked example each, and a monthly meeting that continues after the cohort. It is capability building. The organisation keeps the people, the charter and the rhythm. The visitors do not remain the help desk.

Name the role before the enthusiasm

Champions are chosen from people the teams already trust, with time to practise and to show the work. Curiosity matters. Trust matters more. A person the team will not ask for help cannot carry a new way of working, however fluent they are with the tool. The people leader is part of the choice. A champion appointed by a central team, and unknown to the meeting that owns the work, will collect questions that should have gone to a colleague and will not change the pack.

Together the cohort writes a one-page charter. It says who the champion supports, what the champion is responsible for, and what the champion hands back to IT, risk, or the people who own the policy. The charter is what a manager can say yes to. It is also what a champion can refuse. Helping a colleague on a real task is in. Rewriting the organisation's policy in a side chat is out. Choosing a tool the organisation has not approved is out. Becoming the permanent fixer for every prompt in the building is out.

The boundary is the part keen people skip, and it is the part that keeps the role alive. Microsoft's guidance on an AI champion charter puts the same three elements in public guidance for Copilot adoption: a named purpose, the work the champion does own, and a line around what is handed to IT, risk or the policy owner. Collective Campus uses that shape on whatever tools the organisation already has, not as a tour of one vendor's product. A charter that only lists virtues will not help a champion on a Thursday when a senior person asks for an exception.

Practise on the champion's own job

Champions work on their own jobs in the cohort sessions. Each person leaves with one example the team can copy, written so a colleague can follow the steps without the champion sitting beside them. The example names the task, the inputs that are allowed, the step the tool takes, the check a person still does, and the output the review will accept. An example that cannot be followed from the page is a performance, and the team will still need the champion in the chair.

The people leader makes room in the week for that practice and for the first run with the team. Without that room, the charter stays a document and the champion does the champion work at night, until they stop. Collective Campus treats a manager who will not protect the time as a reason to change the design of the cohort, not as a detail to hope away. The script for this moment is the one in the previous section. The model and the script are the same decision, said once in the charter and again in the week.

Spread one example, then meet once a month

Each champion runs the example with the people they work with, on a real task. They bring the questions, the failures and the edits back to the cohort, so the example gets sharper before the next team uses it. Failure is part of the return. A cohort that only collects success stories is training its members to edit the truth before they walk in. The month-two script exists so the failed use gets time.

After the cohort, the champions meet once a month. They share what worked, what failed, and what the team needs next. The meeting is how peer support continues once the first burst of enthusiasm has passed. The programme owner keeps it on the calendar, chases a missing example, and makes sure the champions still have a place to bring questions. If the meeting slips, the owner can see it. That visibility is the point. A network that depends on goodwill and a busy founder of the idea will fade in the quarter when a restructure or a reporting season arrives.

The model has a size that fits the attention of the people in it. One champion cannot support a building. A long list of champions with no charter and no month in the diary is a mailing list with a name. Collective Campus shapes the cohort to the teams that have a people leader ready to change a meeting. Where the same change should be redesigned inside one or two workflows from a blank map, the 90-Day AI Adoption Sprint is the closer fit. Champions then carry a way of working that the sprint has already made concrete. The sprint changes the work of one or two teams. The champions model carries a way of working further, once there is something real to carry.

What the cohort actually covers

The cohort is a run of working sessions. Length, group and format are tailored after an initial conversation, because a standard timetable would ignore the meeting this organisation is trying to change. The sequence is stable even when the timetable is not. People leaders and champions do the work in this order.

• Name the stall on this team's work: the task, the meeting that consumes it, and the definition of done that meeting still uses.
• Practise the scripts on that task, out loud, with the people leader who will say them again without a facilitator present.
• Write the one-page charter: who is supported, what the champion owns, and what goes back to IT, risk or the policy owner.
• Build one example from each champion's own job, written so a colleague can follow the steps.
• Agree the measure: which record will show that the example was run, and the date of the first monthly meeting.
• Run the example with the team, bring the edits back, and leave the monthly rhythm with a chair and a list of what must be brought.

Sessions are working sessions. People use the tools the organisation already has, on documents the organisation already produces. A session that could be delivered to any company, with sample prompts from another industry, is the thing this programme is built to replace. The capability is local. It lives in the charter, the example and the meeting, all three of which should still exist when the cohort dates are over.

AI adoption asks people to take interpersonal risks on ordinary work. They show a draft that might look naive. They admit they do not know whether an answer is right. They challenge a colleague's output, sometimes a senior colleague's, in a meeting that has started to praise speed. If the last person who showed unfinished work was joked about, the next draft stays hidden. If the last person who said a number looked invented was labelled negative, the number gets pasted. The licence gives people a private place to hide the experiment. The public story is that adoption is low. The private story is that adoption is unaudited.

Amy Edmondson describes psychological safety as a shared belief that the team is safe for interpersonal risk: people can offer an opinion, ask a question, raise a concern and admit a mistake without fearing punishment or humiliation. Harvard Business School's note on psychological safety at work records the warning that people often hear the idea as comfort. Hard work still feels uncomfortable. The standard can stay high. The change is that the discomfort of a hard problem is separated from the fear of being humiliated for raising it. Google's guide to team effectiveness, the public account of Project Aristotle, treats that shared belief as a core dynamic of an effective team.

Collective Campus holds the two dials in the same sentence, because splitting them is how AI rollouts go wrong in opposite directions. Safety without a standard becomes a pleasant room that sends unchecked work. Standards without safety become an anxious room that goes quiet, then experiments in a personal account. The learning zone needs both. For a people leader, both show up in the minute after someone speaks. The scripts in this programme are that minute, written down. The psychological safety workshop is where an intact team practises the reply with the leader who sets the tone, on the conversation they are already avoiding.

The two offerings meet, and they are not duplicates. A people leader can take this programme and still have a team that will not show a rough draft, because the team's habit was set long before the licence. In that case the workshop comes first, or alongside, with the leader in the room. A team can be frank with each other and still stall, because the review meeting has not changed its definition of done. In that case the scripts and the champions model are the missing piece, and another day on feelings will not retire the old pack. The argument is set out in psychological safety is the AI adoption bottleneck. Read it when the licence is already on the desk and the room goes quiet. Use this page when the need is a programme for the people leaders who have to change the week.

A town hall line about it being safe to experiment never sets the tone. The moves that do are local, and they repeat. The people leader shows an unfinished draft, including one place the tool was wrong, before asking the team to show theirs. In the same meeting that asks what the tool sped up, someone asks what it missed. When a person catches an error, the conversation turns to the check. A speech about carelessness teaches the room to hide the next one. A visible example of a stopped workflow matters too, so ending a bad use earns the same respect as launching one.

There is a review-cycle version of the same problem. If a clumsy first month can show up as a lack of capability at performance review time, rational people will wait until they are fluent in private. The people leader will conclude the team is slow, and will buy more training. The team was waiting until a clumsy month could stay out of the review. Separating the experiment from the performance review, for a defined period, and saying the dates out loud, is a management act. It belongs in this programme because it is a decision only the people leader can make. A champion cannot grant it. A slide cannot grant it.

Collective Campus runs the programme for people leaders in Sydney, Melbourne, Brisbane and, in person or remote. The cities are where in-person cohorts most often sit. A hybrid team spread beyond those cities still has the same stall, and the sessions can be remote when the people leader and the champions can attend together. Splitting the people leader off into a separate briefing, and leaving the champions to be trained alone, recreates the mixed signal the programme exists to close. The person whose meeting defines done has to hear the lines, and has to say them.

Every cohort is tailored after an initial conversation. There is no standard price and no standard length, because both depend on the teams, the tools already on the desk, and how much of the surrounding system has to move. A single people-leader group with a named task is a different shape from a cohort of champions drawn from several functions, with a monthly rhythm to design. Collective Campus shapes that with the organisation on a scoping call. The call is the start of the work, not a tour of a catalogue.

What to bring to the scoping call

The useful call starts from the stall, not from a list of tools. Four facts are enough to shape the programme. More detail helps. Vague enthusiasm does not.

• The task that still comes back in the old pack, and the meeting that consumes it.
• The people leader who owns that meeting, and whether they will be in the cohort.
• The names of people the teams already trust, if champions are already in mind, and the time those people can actually protect.
• The policy line that already exists: what may enter a tool, and who owns the exceptions. If that line is unwritten, say so. The charter cannot invent it.

A people leader who cannot attend is a reason to move the date. The programme can still include a programme owner from learning and development, operations or transformation. That owner keeps the monthly meeting and chases missing examples. The owner supports the manager whose sign-off the room watches. Collective Campus moves the date when the people leader who controls the week cannot attend. A cohort of people who do not control that week leaves the old pack in place on Wednesday.

Public sector teams can bring the Australian Government's adoption guidance as the frame the charter has to respect. The two National AI Centre documents linked above are the public references. Councils, departments and government businesses still have local meetings, local templates and local definitions of done. The programme works on those, inside the guidance, and hands policy questions back to the owners. Professional firms bring confidentiality rules and professional standards in the same way. The M3 Property account further down is an example of a firm that already had Copilot and wanted the work to change while those standards held. It was a different format. The constraint is the same one people leaders name on the scoping call.

After the call, the organisation has a proposed shape: who is in the cohort, which task the scripts will be practised on, how the charter will be drafted, what will be recorded, and how the monthly meeting will be chaired once the sessions end. Delivery then follows the sequence in the champions section. Between sessions, champions use the example on live jobs. The people leader uses the scripts in the real meeting and brings back what the room did. A programme that saves the real meeting for afterwards has already decided that the workshop is the product. The meeting is the product.

Collective Campus builds capability for teams. The identity of the work is that the organisation leaves able to run the rhythm itself. Facilitators design the sessions, coach the scripts and pressure-test the charter. They do not remain the owners of the monthly meeting. The programme owner inside the organisation does. That handover is discussed at the start, so a champion network does not quietly depend on an external person to send the calendar invite.