The short answer
AI transformation is the shift from people using AI tools on their own to an organization that has changed how its work gets done. In a 2024 BCG survey of 1,000 executives, about 70 percent of the challenges in implementing AI came from people and processes. I’d run it as a people program with one named owner, in four phases over 12 to 18 months, and judge it by how work and results change.
See the four-phase roadmap- What changes Workflows, roles, skills, leadership, rules and measures.
- Who owns it The executive team, with one named lead and HR involved from day one.
- How long In my planning, a first cycle of 12 to 18 months, in four phases.
What is AI transformation?
AI transformation is a planned change in how an organization works, where AI becomes part of the workflows, roles and decisions that produce results. It includes the technology, but most of the work is redesigning tasks, building skills and changing how leaders and managers lead.
AI adoption is people using AI in their daily tasks, and you can have plenty of it with little transformation. Eight in ten respondents in McKinsey’s 2026 survey say AI has improved their own productivity, but only 37 percent see any effect on their organization’s EBIT (McKinsey). If adoption is still your main problem, start with my guide to AI adoption that sticks.
| AI adoption | AI transformation | |
|---|---|---|
| The question | Do people use AI in their work? | Has the way we work changed, and does it show in results? |
| Unit of change | A person or a team, task by task | A workflow, a role, a function |
| Led by | Team leads, champions and L&D | The executive team, with HR, IT and the business |
| Sign it’s working | Steady use without reminders | Redesigned workflows, updated roles, measurable results |
Why do AI transformations fail on the people side?
AI programs often start with the technology, because it’s the part you can buy. And buying is a lot easier than asking a team to change how it handles a complaint. In my reading of these studies, six patterns hold AI transformations back. Here is the fix I’d start with for each.
- The budget goes to licenses. AI leaders in BCG’s 2024 research put 10 percent of their resources into algorithms, 20 percent into technology and data and 70 percent into people and processes (BCG). Check your plan against that split before you buy more seats.
- Tools without a plan. Name three to five workflows that matter, say why you chose them and who owns each one. Most US employees say their organization hasn’t communicated a clear plan for AI (Gallup). Below is a way to pick them.
- AI added on top of old workflows. Redesign the workflow first, then add the tool. Nearly three-quarters of McKinsey’s high performers have fundamentally redesigned workflows, against a quarter of other respondents.
- Managers left out. Train managers before their teams. Gallup surveyed 23,717 US employees in February 2026. Where AI is available, frequent use is 78 percent among employees who strongly agree that their manager actively supports AI, and 44 percent among the rest (Gallup).
- Leaders who aren’t aligned. Only 26 percent of the AI users in Microsoft’s 2026 Work Trend Index say their leadership is clearly and consistently aligned on AI (Microsoft). Agree on the priorities together, then tell the organization the same thing.
- Nobody measures what changed. Set a baseline before you change anything. McKinsey’s high performers are twice as likely as others to have defined processes for measuring the impact of AI.
If I had to pick one, I’d start with the managers. They decide what the team’s week looks like, and a tool nobody’s manager asks about stays a side project. When a team still stalls, the ADKAR model and the change curve help you see where each person is stuck.
What is an AI operating model?
An AI operating model is the way an organization decides, builds, runs and governs its AI work. It defines who sets priorities, who owns each AI-supported workflow, who builds or buys the tools, who sets the rules and how results get reviewed.
| Shape | How it works | Fits when | Watch out for |
|---|---|---|---|
| Central team | One AI team sets priorities, picks tools, builds and trains | You are early, small or heavily regulated | A queue where the business waits instead of learning |
| Hub and spoke | A small central team sets standards, and each function has named owners and champions | Most mid-sized and large organizations after the first pilots | Spokes without time or mandate |
| Federated | Functions own their AI work under light shared standards | Mature organizations with strong functions and good data | Duplicate tools, uneven rules, little shared learning |
I recommend hub and spoke for most organizations. Start central, then move ownership out to the functions once they have named owners with time set aside. If you need engineers for integrations or data platforms, compare the best AI consulting firms.
Who owns what in an AI transformation?
| Role | Owns | Common mistake |
|---|---|---|
| Executive team | Priorities, funding, the decision on freed time, the quarterly review | Handing it all to IT |
| Transformation lead | The roadmap and coordination across functions | A title without budget or time |
| HR | Role redesign, skills, manager support, employee involvement | Brought in after the decisions |
| IT, security and legal | Platforms, data protection, rules for high-risk uses | Choosing tools before asking what the work needs |
| Workflow owners | Redesigning their workflow, and its results | Reporting usage instead of outcomes |
| Managers and champions | Team habits, coaching, peer support | Doing it on top of a full job |
| Employees and their representatives | Input on how work changes, consultation where the law requires it | Informed at the end instead of involved early |
HR owns the parts that decide whether the change sticks. In the AI & HR Barometern 2026 from humAIn Academy, 329 HR professionals in Sweden answered between 8 June and 24 September 2026. They chose to take part and are likely more interested in AI than HR overall. Only 4 percent of them say HR has the main responsibility for AI, against 33 percent for the leadership team and 26 percent for IT (State of AI in HR). I’d put HR next to the transformation lead from week one.
How do you redesign work and roles?
AI workforce transformation happens one workflow at a time, such as handling a complaint, closing the books or hiring someone.
How do you pick the first workflows?
Score each candidate from 1 to 3 on five questions and start with the highest totals.
- Does it take a lot of time or happen every week?
- Can you measure its lead time, quality or cost today?
- Is there a named owner who wants to change it?
- Will a person catch an AI mistake before it reaches a customer or a decision about an employee?
- Do the people doing it already have the data and an approved tool?
Include at least one workflow that serves customers or brings in revenue. BCG’s 2024 survey found that AI leaders pursued about half as many AI opportunities as other companies, and that 62 percent of the value from AI came from core functions such as operations, sales and marketing, and R&D. I’d rather see three workflows redesigned properly than twenty pilots nobody follows up. Leave decisions about people, such as ranking applicants, for later.
How do you redesign one workflow?
- Map the workflow, not the tool. Write down its tasks together with the people who do them.
- Mark each task. An agent runs several steps and a person approves, AI does it and a person checks, AI assists, or a person does it alone. Note what happens if AI gets it wrong.
- Decide where the freed time goes. More customer time, shorter lead times, growth without new hires or fewer roles. Decide before you measure, and tell the team.
- Update the roles. Change the role description, the skills needed, who reviews AI output and what good performance looks like.
- Run it, then review. Check quality, time, errors and how the team experiences the new way of working. Keep, change or stop. See AI pilots that don’t die.
Use a worksheet like this for each workflow. The rows show a hiring workflow as an example.
| Task | AI’s role | Human check | Skill needed | Freed time goes to |
|---|---|---|---|---|
| Write the job ad | Drafts from the role brief | Hiring manager edits and checks for biased wording | Writing a clear brief | Calls with strong candidates |
| Summarize interview notes | Drafts | Interviewer checks against own notes. Use an approved tool, and the summary must not score or rank candidates | Checking a summary for accuracy | Quicker hiring decisions |
| Rank applicants | None for now | A person decides. High-risk under the EU AI Act, so decide separately | Structured assessment | Not applicable |
What about jobs and headcount?
People will ask, so answer first. In McKinsey’s 2026 survey, 39 percent of respondents expect AI to reduce their organization’s headcount in the coming year. Over the past year, though, only 14 percent of respondents at organizations using AI report that AI contributed to a smaller workforce, less than half the share who had expected it. In my view, silence does more damage than a hard message. If roles will go, involve HR and employee representatives early, and keep the entry-level work that teaches people the job. Before anyone decides on numbers, plan roles and headcount task by task, as I describe in strategic workforce planning when AI changes the work. For the conversation in each team, my guide to AI anxiety at work has talking points for managers and the promises to avoid.
A message in four parts works for an all-hands or a team meeting. Only promise what you can keep.
- What we have decided. “From [month], we use AI in [workflow]. [Role] will spend less time on [task] and more on [task].”
- What we haven’t decided. “We haven’t decided whether [team] changes in size. We will tell you by [date].”
- What you can count on. “Before your role changes, you will hear it from your manager, with HR, and you will get training for the new tasks.”
- How you take part. “[Name] redesigns the workflow together with the team. Questions and concerns go to [channel or person].”
What do leaders and managers need to do differently?
Leaders set the direction and managers turn it into the team’s week. Both need to use the tools themselves. My guide to AI for leaders goes deeper on the leadership side.
The executive team
- Use AI in your own work and share what worked.
- Agree on three to five priorities and drop the rest.
- Protect learning time and put a number of hours per month on it.
- Put AI results in the normal business review.
- Decide where freed time goes, and tell people.
Every manager
- Ask about AI use in one-on-ones and team meetings.
- Agree team rules on data and checking output.
- Redesign one workflow with the team each quarter.
- Share what failed as well as what worked.
- Pass risks and concerns to the transformation lead.
Only 36 percent of US employees in organizations that integrate AI strongly agree that their manager supports their team’s use of it (Gallup, May 2026). Closing that gap is what my AI training for managers and leaders is for. At Mediemyndigheten, the Swedish Agency for the Media, managers had two sessions of their own, and the leadership team did not take a separate course. Its members joined the employee groups my humAIn Academy colleagues and I trained, heard the concerns first-hand and used them to shape strategy and policy.
What skills and rules does an AI transformation need?
Skills come in three layers.
| Who | What they need | How to build it |
|---|---|---|
| Everyone | What the approved tools do well and badly, which data never goes in, how to check output | One hour of foundations, then one real task from their own week |
| Each role | How AI changes their own tasks | Training on the redesigned workflow, run with the manager |
| Owners, champions and the lead | Redesigning workflows and measuring results | Hands-on work on the priority workflows, with time set aside |
Since the omnibus amendment entered into force on 27 July 2026, Article 4 of the AI Act asks providers and deployers to take measures that support their staff’s AI literacy, without requiring a specific level (Regulation (EU) 2026/1744). My AI literacy playbook sets out the order.
What are the minimum rules before you scale?
- Approved tools and a quick way to request new ones
- One page on data that never goes in
- Which outputs a person must check
- A register of AI uses that affect decisions about people
- One named person for questions and incidents
AI used to screen or evaluate candidates, to make decisions on promotion or termination, to allocate tasks based on behavior or personal traits, or to monitor performance is listed as high-risk in Annex III of the AI Act. After the omnibus amendment, those rules apply from 2 December 2027. Who signs off on those uses is a governance question, and my AI governance framework sets out the roles, risk tiers and committee charter for it. Bring employees in early. Only 16 percent of the Swedish HR respondents in the barometer say employees or union representatives have been actively involved in AI questions. If people already use unapproved tools, start with shadow AI.
How do you measure an AI transformation?
Measure on four levels and review them every quarter. License counts show that people could use AI, and nothing more.
| Level | What it shows | Example measures |
|---|---|---|
| 1. Use | People use AI in real work | Share of teams using AI weekly in agreed workflows |
| 2. Work | The workflow changed | Lead time, rework, quality after review, hours freed |
| 3. Business | Results moved | Cost per case, sales cycle length, customer satisfaction |
| 4. People | People can and want to work this way | A short quarterly pulse on confidence with the tools, manager support and workload |
Set the baseline before you change a workflow. For level 1, I use breadth, depth and momentum from the AI adoption guide. Seven in ten of the Swedish HR respondents in the barometer say they don’t measure the effect of AI in HR, and 5 percent have defined KPIs.
What does an AI transformation roadmap look like?
This is the order I recommend for the organization as a whole. The timing is a planning example, and a large organization will have workflows in different phases at once. Inside each workflow, the five-phase change framework in my AI adoption guide still applies.
- Direction, the first 4 to 6 weeks. The executive team uses the tools on its own work, picks three to five priority workflows, names a transformation lead and an owner per workflow, writes one page of rules and sets baselines. A workshop for the executive team is a practical way to start. Done when the executive team has agreed on the one-page plan below and shared it.
- Prove, months 2 to 4. Redesign two or three workflows with the teams that do the work, and train those teams and their managers on real tasks. My guide to your organization’s first 90 days with AI covers this phase over 13 weeks. Done when each workflow has a keep, change or stop decision based on data.
- Scale, months 4 to 9. Spread what worked to similar teams, give champions time and update role descriptions, onboarding and performance criteria. Done when the priority workflows run in most target teams and you report level 2 measures.
- Redesign and run, months 9 to 18. Budgets, hiring profiles and career paths follow the new work, and AI results become a standing item in business reviews. Done when the next set of workflows has started.
A one-page AI transformation plan
Copy these headings and fill them in together as an executive team.
- Why now. Two or three sentences on what AI should change for customers, costs or growth.
- Priority workflows. Three to five, each with an owner, today’s baseline and a target.
- Freed time. What happens to the time AI frees up in each workflow, in plain words.
- People. Which roles change first, when managers get trained and how employees and their representatives take part.
- Rules. Approved tools, data that never goes in, which outputs a person checks and which uses need a separate decision.
- Measures. One measure per level, for use, work, business and people.
- Lead and review. The transformation lead’s name and the date of the first quarterly review.
Where are you now? A ten-question self-check
Answer all ten in order.
- Can every executive name your top AI priorities?
- Does one named person lead the transformation, with time and budget?
- Does each priority workflow have an owner in the business?
- Do people know which tools and data are approved?
- Have you told employees what happens to the time AI frees up?
- Did managers get training before their teams?
- Have you redesigned at least one workflow, beyond adding a tool?
- Do you measure lead time, quality or cost, beyond usage?
- Have you updated the role descriptions for the roles that changed?
- Is AI a standing item in your business review?
Your first no shows which phase you are in. A no on questions 1 to 5 means phase 1, on 6 to 8 phase 2, on 9 phase 3 and on 10 phase 4. The other no answers go on your list of next steps. Ten yes answers mean it’s time to pick the next set of workflows.
Frequently asked questions
How is AI transformation different from digital transformation?
Digital transformation moved processes and data onto digital systems. AI transformation changes who, or what, does the work inside those processes, so roles, skills and decision rights change too. Digital work is often the starting point, because AI needs data it can reach and systems it can connect to.
Why do AI transformations fail?
Mostly for people and process reasons. In a 2024 BCG survey of 1,000 executives, about 70 percent of the challenges in implementing AI came from people and processes. Common patterns are money spent on licenses instead of people, tools without a plan, AI added to old workflows, managers left out, unaligned leaders and no measure of what changed.
Who should lead an AI transformation, IT or HR?
The executive team owns it and names one transformation lead. IT runs platforms and security, but the parts that decide whether the change sticks, such as role redesign, skills and manager support, sit with HR and the business. Put HR next to the lead from the first week.
What should an AI transformation strategy and roadmap include?
A one-page plan with why now, three to five priority workflows with owners, baselines and targets, what happens to freed time, the people plan, the rules, one measure per level and a named lead with a quarterly review date. The roadmap then shows which phase each workflow is in.
Sources
- McKinsey, The state of AI in 2026, on the road to ROI (August 2026)
- BCG, AI adoption in 2024, 74% of companies struggle to achieve and scale value (October 2024)
- Gallup, Global indicator, artificial intelligence (data from May and August 2026)
- Gallup, AI in the workplace, what separates adopters and holdouts (April 2026)
- Microsoft, 2026 Work Trend Index (May 2026)
- EU AI Act, Regulation (EU) 2024/1689
- Regulation (EU) 2026/1744, amending the AI Act
- AI Act Annex III, high-risk AI systems
- State of AI in HR (AI & HR Barometern 2026, humAIn Academy)
Planning an AI transformation?
I work with executive teams, managers and HR on the people side, from a first leadership workshop to AI change programs that cover the first 3 to 12 months of the roadmap. The leadership teams I’ve worked with give an average rating of 4.97 out of 5, measured across 13 groups.