Most enterprise AI programmes do not fail because the technology is weak. They fail because the organisation around the technology does not change with it. The firm you choose to manage that human change is one of the most important decisions in your whole AI strategy, and the market is crowded with firms that have simply added the word "AI" to a generic change deck.
Here is the answer in one paragraph. The best AI change management firms treat AI adoption as mostly a people-and-process change, not a technology install. When you evaluate one, look for seven things, and weight three most heavily: whether the firm redesigns how your people actually work, builds AI skills across every role, and transfers capability so your team can carry on without it. The full seven criteria, each with the questions to ask and the warning signs, are set out below.
If you lead HR, transformation, or an enterprise function, you are probably being asked a hard question in board meetings: what share of our people can actually use AI, and are we getting a return on what we spend on it? The honest answer for most large organisations today is "not much yet." In its March 2025 State of AI research, McKinsey found that more than 80 percent of companies were seeing no tangible impact on enterprise-level earnings from their use of generative AI (generative AI means tools that create text, code, or images, such as ChatGPT). The gap is rarely the model. It is adoption, workflow, skills, trust, and governance. In other words, it is change management.
This guide explains what an AI change management firm actually is, how that work differs from traditional change management, and the seven things enterprise buyers should look for before signing. Each criterion comes with the questions to ask and the warning signs to watch, so you can tell a genuine AI change partner from a repackaged one.
The short answer: what to look for in an AI change management firm
- Genuine AI-specific change expertise, not a generic change model with "AI" pasted on top.
- A method that treats AI as mostly people and process, and redesigns how work is done, not just which tool is installed.
- Real workforce and skills transformation, including role redesign and AI literacy for everyone, not only technical teams.
- Capability transfer that leaves your own people able to run without the consultant, rather than permanently dependent on them.
- Responsible AI governance built into the change, not bolted on after go-live.
- Measurement tied to business outcomes, not usage numbers that look good but prove nothing.
- Long-term adoption support and leadership alignment, because AI change never really has a finish line.
What is an AI change management firm?
An AI change management firm helps an organisation manage the human and operational side of adopting artificial intelligence: getting employees to use it well, redesigning roles and workflows around it, building the skills to sustain it, and putting the governance in place to do it responsibly. The goal is not to install software. It is to make sure the people, processes, and culture actually shift so the technology delivers value.
One word of caution before you shortlist anyone. The phrase "AI change management" is used in two very different ways, and firms rarely make the distinction clear.
Meaning 1: Using AI to run change management. Some vendors use "AI change management" to mean software that uses AI to support a change programme, for example tools that track sentiment, nudge users, or automate communications.
Meaning 2: Managing the organisational change of adopting AI. This is the enterprise problem most leaders actually have: helping thousands of people adapt how they work as AI enters the business. This guide is about meaning 2, which is the harder and higher-stakes discipline.
When you talk to a firm, ask which one they mean. A partner who leads with adoption dashboards may be selling meaning 1. A partner who talks about role redesign, capability building, and governance is working on meaning 2. The best firms can do both, but you should know which problem you are buying help for.
How is AI change management different from traditional change management?
Traditional change management grew up around projects with a clear start and end: a new enterprise resource planning system (ERP, the core software that runs finance, supply chain, and operations), an office move, a reorganisation. You prepared people, went live, embedded the change, and closed the programme. AI breaks that pattern in several ways, which is why a firm's traditional playbook is not enough on its own.
| Dimension | Traditional change management | AI change management |
|---|---|---|
| Timeline | One-off project with a go-live date and a close. | Continuous. Models, tools, and use cases keep changing, so there is no single finish line. |
| Where the effort goes | Communications, training, stakeholder buy-in. | The same, plus redesigning workflows and roles. BCG puts the split at roughly 10 percent algorithms, 20 percent technology and data, and 70 percent people and processes. |
| The core obstacle | Resistance and awareness. | Adoption and fit. Tools that do not match real workflows, plus gaps in skills and trust, stall adoption long before the technology is the limiting factor. |
| Skills | Train users on the new system. | Build lasting AI literacy and redesign what jobs consist of, because the work itself changes. |
| Risk and trust | Managed mostly at the edges. | Central. Bias, data privacy, job-loss fear, and accountability must be handled inside the change, not after it. |
| Measurement | Adoption and milestone completion. | Business outcomes and genuine proficiency, separated from vanity usage counts. |
Two findings make this concrete. McKinsey's 2025 research found that redesigning workflows has the biggest effect on whether a company sees earnings impact from generative AI, which is a change-and-operating-model task, not a technology task. And a 2025 study from MIT's Project NANDA, The GenAI Divide: State of AI in Business 2025, reported that about 95 percent of organisations were getting no measurable return on their generative AI spend, and concluded that the divide between winners and everyone else "does not seem to be driven by model quality" but by approach. Massivue has written about this pattern in more depth in why so many enterprise AI initiatives fail and how to manage change during enterprise AI transformation.
With that grounding, here are the seven criteria that separate a real AI change management firm from a generic one.
1. Genuine AI-specific change expertise, not a generic model with "AI" added
What it means. The firm can show, in plain terms, what it does differently for an AI programme compared with a standard software rollout or reorganisation. Change management as a discipline is valuable and well proven: Prosci's benchmarking across thousands of projects found that 88 percent of initiatives with excellent change management met or exceeded their objectives, against only 13 percent of those with poor change management. But AI adds problems that a classic model was never designed for: work that keeps changing as models improve, employees who fear being replaced rather than merely inconvenienced, and a constant governance and trust overhead.
Why it matters for AI. Many respected change firms and consultancies have layered AI onto an existing method. That can be a strength when the underlying method is strong, but only if the firm has genuinely adapted it. The risk is a partner who sells you the same programme they sold before generative AI existed, with new branding. The evidence says the AI-specific parts are exactly where value is won or lost, so probe for specifics rather than accepting "we do AI transformation" at face value.
Questions to ask:
- What do you do differently on an AI programme that you would not do on an ERP or restructuring programme?
- Can you describe an AI adoption engagement from the last 12 to 18 months, and what changed for the employees involved?
- How do you handle the fact that the tools and use cases will keep changing during and after our engagement?
- Can you give me a client reference from the last 12 to 18 months that I can speak to directly, ideally in our industry?
Warning signs:
- The proposal is a familiar change model with "AI" swapped in for the old system name.
- Every example is a pilot or a demo, never a sustained, organisation-wide adoption.
- The team cannot explain what makes AI change harder than past technology changes.
- Every reference is an anonymous case study, with no client you can actually contact.
2. A method that treats AI as mostly people and process, and redesigns the work
What it means. The firm spends most of its energy on how work is done and how the organisation is structured, not on the tool itself. Practically, that means mapping and redesigning workflows, clarifying decision rights, and reshaping the operating model (the way people, processes, and technology are organised to deliver work) so AI has somewhere useful to sit.
Why it matters for AI. This is one of the best-evidenced points in the field. BCG's widely cited guidance puts successful AI transformation at roughly 10 percent algorithms, 20 percent technology and data, and 70 percent people and processes. McKinsey's 2025 research found that, of all the moves a company can make, redesigning workflows has the biggest effect on whether generative AI shows up in earnings. Deloitte's 2026 enterprise AI research found that only about 30 percent of organisations were redesigning key processes around AI, and that many were using it only at a surface level with little change to how work actually happens. A firm that jumps straight to tools and skips the process work is optimising the 30 percent that is algorithms and technology and ignoring the 70 percent that is people and process.
Questions to ask:
- Walk me through how you would redesign one of our core workflows around AI, not just add a tool to it.
- How much of a typical engagement is process and operating-model work versus technology deployment?
- How do you decide which processes to change first?
Warning signs:
- The plan is organised around tools and licences rather than workflows and outcomes.
- Process redesign, role mapping, and operating-model work are missing or treated as an afterthought.
- Success is defined as "deployed," not "the work is now done differently and better."
Related reading: what an AI operating model is.
3. Real workforce and skills transformation, for everyone, not only technical teams
What it means. The firm assesses the skills your people have, builds AI literacy across roles, and redesigns what jobs consist of as tasks shift to or are shared with AI. AI literacy means the practical ability to use AI tools well and judge their output, not the ability to build models. This has to reach frontline and non-technical staff, not just data teams.
Why it matters for AI. Skills are now the binding constraint. The World Economic Forum's Future of Jobs Report 2025 estimated that nearly 40 percent of the skills workers use will change by 2030, and that 59 of every 100 workers will need reskilling or upskilling by then. McKinsey's 2025 workplace research found that 46 percent of leaders name skill gaps as a significant barrier to AI adoption. Generic training does not close that gap, because a lawyer, a warehouse supervisor, and a call-centre agent need very different things from AI. The firm should tailor learning to roles and tie it to the actual work.
Questions to ask:
- How do you assess our current AI skills, and how do you tailor learning by role and department?
- How do you redesign roles when AI takes over part of a job, and who decides what people do with the freed-up time?
- What proportion of the workforce, not just technical staff, does your programme reach?
Warning signs:
- One generic "AI awareness" course for everyone, with no role-based paths.
- Training stops at technical teams and never reaches the frontline.
- No plan for what changes in people's actual jobs, only what tool they will be shown.
Related reading: why generic AI training fails, department by department and a guide to AI upskilling programmes.
4. Capability transfer that leaves you independent, not dependent
What it means. The firm's aim is to make your own people able to run and evolve the change after the engagement ends, through knowledge transfer, coaching, and building internal capability, rather than keeping you on a permanent retainer for work you could own yourself.
Why it matters for AI. Because AI change is continuous, a model where the consultant holds all the knowledge quietly becomes a model where you can never stop paying them. That is bad economics and worse resilience. It also misses the point of the exercise: if the goal is an organisation that can keep adapting on its own, a partner who never builds your internal capability has not delivered it. This is where a firm's business model matters. Some are structured to bill more hours; others are structured to hand over. Ask directly how the engagement is designed to end, and what you will be able to do yourselves afterwards. This capability-first stance is central to how Massivue works, through a consult, upskill, and sustain model in which certified training is paired with the consulting so teams can keep going after the advisers leave.
Questions to ask:
- Who on our side will be able to run this after you leave, and how do you build them up to that point?
- What does knowledge transfer and handover actually look like in your engagements?
- How is your commercial model structured, and does it reward handover or ongoing dependency?
Warning signs:
- No defined handover, exit, or internal-capability plan.
- Key knowledge, prompts, or playbooks stay with the firm and are not documented for you.
- The commercial model only grows if you keep buying more of their time.
5. Responsible AI governance built into the change, not bolted on later
What it means. The firm treats governance, the policies, controls, and accountability for using AI safely and legally, as part of the change from day one. That includes data privacy, bias and fairness, human oversight, clear ownership of decisions, and alignment to recognised standards.
Why it matters for AI. Governance is not the brake on adoption; done well, it is what makes adoption possible, because people use tools they trust. Deloitte's 2025 research found that worries about complying with regulations were the single most-cited barrier to scaling generative AI, named by 38 percent of leaders. McKinsey found that a chief executive's oversight of AI governance was one of the factors most correlated with bottom-line impact. A capable firm can speak fluently about frameworks such as the NIST AI Risk Management Framework (a voluntary United States government framework built around four functions: govern, map, measure, and manage), the international management standard ISO/IEC 42001, and, for organisations touching Europe, the AI literacy duty in Article 4 of the European Union AI Act, which has applied since February 2025. For teams in Singapore and the wider region, the firm should also know the Model AI Governance Framework from the Infocomm Media Development Authority. Massivue's own view on why weak governance forces companies to pull agents back out of production is set out in its piece on agentic AI governance.
What to ask the firm to produce. Turn those frameworks into concrete deliverables. Ask the firm to show, for your use cases, a mapping of your AI systems to the NIST AI Risk Management Framework; a risk classification under the European Union AI Act where it applies to you; model cards (short factsheets that describe how each AI system works and where it should not be used); the results of bias and fairness testing; a data-protection and data-residency statement, for example alignment to Singapore's Personal Data Protection Act (PDPA); and a single named person accountable for AI decisions on your side. A firm that can name these deliverables is doing governance. A firm that can only name the frameworks is not yet.
Questions to ask:
- How do you build governance into the rollout, and which standards or frameworks do you work to?
- How do you handle data privacy, bias, and human oversight for the specific use cases we care about?
- Who becomes accountable for AI decisions inside our organisation once you have gone?
Warning signs:
- Governance is a late-stage workstream or a separate upsell, not part of the core plan.
- The team cannot name a single relevant framework or the rules that apply in your markets.
- Responsibility for AI risk is left vague, with no named owner on your side.
6. Measurement tied to business outcomes, not usage vanity metrics
What it means. Before the work starts, the firm agrees how success will be measured, and separates three different things: whether people are using the tools, whether they are genuinely proficient, and whether the business is better off. Numbers like logins and licences activated are "vanity metrics," easy to show and easy to inflate, but they do not prove value.
Why it matters for AI. The whole reason so many programmes disappoint is that spend rises while measurable impact does not. When most companies see no earnings impact from generative AI, as McKinsey found, a firm that measures only adoption is helping you tell a story that the profit and loss statement will later contradict. A serious partner defines value metrics up front, tied to the workflows being changed, and is honest that some benefits take time. Beware measures that reward the wrong behaviour, for example targets that push staff to log AI use they are not really making. Massivue explores this measurement problem in closing the AI return-on-investment gap.
The three layers to measure. A good partner tracks all three, because usage can look healthy while real value stays flat.
| Layer | Example metric | What good looks like |
|---|---|---|
| Usage | Share of the target users active in the tool each week. | Rising and then sustained, not a launch spike that fades. |
| Proficiency | Quality of AI-assisted work, checked by a human review. | Reliable results without heavy rework or correction. |
| Business value | Cycle time, quality, cost, or revenue on the changed workflow. | A measurable improvement against the pre-project baseline. |
Questions to ask:
- How will we measure success, and how do you separate usage from real proficiency and real business value?
- What outcome metrics do you tie to each workflow you change?
- What is a realistic timeline before we should expect measurable business impact, not just activity?
Warning signs:
- Success is reported only as adoption, logins, or licences activated.
- No baseline is taken before the work starts, so improvement cannot be proven.
- Metrics are chosen to flatter the programme rather than to inform your decisions.
7. Long-term adoption support and leadership alignment, not a launch and a goodbye
What it means. The firm helps align leaders behind the change and stays engaged past go-live to sustain adoption, manage resistance, and keep momentum as tools and use cases evolve. Leadership alignment means senior sponsors genuinely agree on the goals and back the change, not just approve a budget. This is not in tension with capability transfer (criterion 4): a good partner builds your independence and stays available while you grow into it, rather than vanishing at launch.
Why it matters for AI. People are tired. Gartner's research found that employees' willingness to support enterprise change fell from 74 percent in 2016 to 43 percent in 2022, and in 2025 Gartner reported that only about a third of leaders felt their last change achieved healthy adoption. Layer AI's constant change on top of that fatigue and the risk is a strong launch followed by quiet abandonment. Because AI has no true "go-live and done" moment, the support has to be ongoing and the leadership has to stay visibly committed. A firm that disappears after launch leaves you with the hardest part, sustaining the change, entirely on your own.
Questions to ask:
- What support do you provide after go-live, and for how long?
- How do you get and keep senior leaders genuinely aligned, not just signed off?
- How do you handle resistance and change fatigue when the novelty wears off?
Warning signs:
- The engagement ends at launch, with adoption treated as your problem from then on.
- No plan for sustaining momentum as tools and use cases keep changing.
- Leadership sponsorship is assumed rather than actively built and maintained.
Related reading: building an AI-ready culture.
The types of AI change management firms, and their trade-offs
No single kind of firm is best for everyone. The right choice depends on your size, your industry, how much internal capability you already have, and how much you value independence afterwards. Broadly, the market falls into five groups.
| Type of firm | Known for | Best when | Watch out for |
|---|---|---|---|
| Global strategy and consulting firms (e.g. McKinsey, BCG, Bain, Accenture, Deloitte, PwC, KPMG, EY) | Strategy, scale, and large end-to-end transformation. Several run their own AI build and upskilling arms. | You need board-level strategy and can fund a large programme. | Cost is high, and you can be left dependent; insist on a written capability-transfer and exit plan. |
| Dedicated change-management specialists (e.g. Prosci, Kotter, Changefirst) | Deep, proven change methods and certification. Prosci's ADKAR model is the most widely used. Some now offer AI-specific adaptations. | You want rigorous, proven change discipline. | Can be lighter on AI-specific and technical depth; check the AI version is a genuine adaptation, not a relabelled classic method. |
| AI-focused transformation boutiques | Hands-on AI deployment and adoption, often faster and more specialised than the giants. | You want deep AI specialisation and speed. | Can be strong on the technology but light on the people, governance, and measurement side; check all three. |
| Training-led and academy firms | Building AI literacy and capability at scale across the workforce. | Skills are your main gap. | Training alone will not redesign your processes or governance; it needs to sit inside a wider change plan. |
| Digital adoption platforms (e.g. WalkMe, Whatfix) | In-app guidance and adoption analytics. These are tools, not consultancies. | You need to guide and measure software adoption. | These are tools, not change partners; they do not redesign the work or manage the people side. |
Firms are named as examples of each category, not as endorsements or a ranking. Verify any provider's current capabilities and references directly before you engage them.
A useful lens is the trade-off between depth and independence. The largest firms bring scale and credibility but can be expensive and can leave you reliant on them. Specialists and boutiques bring focus but may not cover the whole picture on their own. A growing middle option combines consulting with a built-in training academy, so that strategy and capability building happen together and the client is meant to be self-sufficient at the end. This is the model Massivue uses, and it maps directly to criteria 3 and 4 above. Whichever type you lean towards, score them against the seven criteria rather than the brand name.
The seven-criteria partner scorecard
Turn the seven criteria into a one-page scorecard so you can compare firms consistently instead of on gut feel. Rate each shortlisted firm from 1 (weak) to 5 (strong) on every criterion, using the questions and warning signs in this guide as your evidence. Weight three criteria most heavily, criteria 2, 3, and 4, because redesigning the work, building real skills, and transferring capability are where the evidence says most of the value, and most of the failure, sits. This is a practical decision aid, not a scientifically validated instrument: use it to make your reasoning explicit and comparable, not to produce a falsely precise score.
| Criterion | Weight | What a strong (5) answer looks like |
|---|---|---|
| 1. AI-specific change expertise | Standard | Concrete recent examples, a contactable reference, and a clear account of what they do differently for AI. |
| 2. People-and-process method | High | Most of the work is workflow and operating-model redesign, not tool installation. |
| 3. Workforce and skills transformation | High | Role-based AI literacy for the whole workforce, plus role redesign. |
| 4. Capability transfer | High | A defined handover and a plan to make your people self-sufficient. |
| 5. Responsible AI governance | Standard | Governance built in from day one, with the concrete artifacts named above and your local rules covered. |
| 6. Outcome measurement | Standard | Baselines and value metrics agreed up front, separated from usage numbers. |
| 7. Long-term support and leadership alignment | Standard | Post-launch support and an active plan to keep leaders and staff on board. |
How to read the result. A strong candidate scores 4 or 5 on the three weighted criteria (2, 3, and 4) and no lower than 3 on the rest. Treat any score of 1 or 2 on governance (criterion 5) or measurement (criterion 6) as a reason to pause, however impressive the rest looks, because those are the areas that quietly sink AI programmes later. If two firms are close, let the weighted criteria and the quality of their references decide.
Where Massivue fits
Massivue is a Singapore-based AI transformation and consulting firm built around a "consulting plus academy" model. In practice that means strategic work is paired with certified training through Massivue Academy, so that teams gain the skills to keep going rather than depending on advisers indefinitely. Its enterprise transformation work follows a consult, upskill, and sustain sequence, its AI workforce transformation service focuses on building AI-fluent organisations, and its Protum framework, an AI operating model, sets out how people, processes, and technology fit together. That approach lines up with several of the criteria above, particularly workforce and skills transformation, capability transfer, and the people-and-process emphasis.
We have written this guide to be useful whoever you choose, and the criteria apply to any firm on your shortlist. If you want to see how these ideas play out in detail, the Massivue blog covers managing change during enterprise AI transformation and end-to-end change management at length, and the enterprise transformation service page explains how the consult, upskill, and sustain model works.
Conclusion: how to make the decision
Choosing an AI change management firm comes down to a single question behind all seven criteria: will this partner change how our organisation works and build our people up, or will it install technology and hand us a dependency? The research is consistent that the people and process side is where most of the value, and most of the failure, lives. So weight your decision there.
Shortlist two or three firms of different types. Put each through the scorecard, ask the specific questions in this guide, and insist on a recent, contactable reference where you can. Pay closest attention to criteria 2, 3, and 4, because a firm that redesigns the work, builds real skills across your workforce, and leaves you self-sufficient is the one most likely to still be paying off in two years. The best sign of all is a partner who is honest about what they cannot do, and whose goal is to make themselves unnecessary.
Frequently asked questions
What is AI change management?
AI change management is the discipline of managing the human and operational side of adopting artificial intelligence. It covers helping employees use AI well, redesigning roles and workflows around it, building the skills to sustain it, and putting responsible governance in place. The aim is to make sure people, processes, and culture actually shift, so the technology delivers value rather than sitting unused.
How is AI change management different from traditional change management?
Traditional change management handles one-off projects with a go-live date and a close. AI change management is continuous, because tools and use cases keep evolving. It puts far more weight on redesigning workflows and roles, treats governance and trust as central rather than peripheral, and focuses on closing a lasting skills gap. In short, it is less about a single launch and more about an organisation that keeps adapting.
What does an AI change management firm actually do?
A good one assesses how ready your organisation is, redesigns the workflows and operating model around AI, builds AI literacy across roles, embeds governance and risk controls, defines how success will be measured, and supports adoption after launch. Crucially, it should transfer capability to your own people so you can sustain the change once the engagement ends, rather than remaining dependent on outside advisers.
Why do most enterprise AI projects fail?
The evidence points to people and process, not technology. McKinsey found in 2025 that more than 80 percent of companies saw no tangible earnings impact from generative AI. MIT's Project NANDA reported that roughly 95 percent of organisations were getting no measurable return, and concluded the divide was driven by approach, not model quality. The common causes are weak adoption, unredesigned workflows, skills gaps, low trust, and no clear measurement.
What should enterprise leaders look for in an AI transformation partner?
Look for seven things: genuine AI-specific change expertise; a method that treats AI as mostly people and process; real workforce and skills transformation for all roles; capability transfer that leaves you independent; responsible AI governance built in from the start; measurement tied to business outcomes rather than usage; and long-term adoption support with genuine leadership alignment. Score each shortlisted firm against these rather than choosing on brand alone.
How do you measure AI adoption and its return?
Separate three things. Usage tells you whether people log in and click. Proficiency tells you whether they use AI well. Business value tells you whether outcomes such as cycle time, quality, cost, or revenue actually improved. Usage numbers are easy to inflate and prove little on their own, so agree baseline and value metrics before the work starts, and tie them to the specific workflows you are changing.
How much does AI change management cost?
It varies widely by scope, provider type, and organisation size, and firms rarely publish change-specific pricing. A small assessment or pilot costs far less than an enterprise-wide programme, and models range from fixed-fee sprints to monthly retainers. Any figure you see online is directional at best. The practical step is to get scoped, comparable quotes from two or three firms and judge them on value and capability transfer, not headline rate alone.
Who should own AI change management: HR, IT, or a dedicated function?
It should be shared, with clear accountability. HR usually owns skills, roles, and adoption, IT owns the tools and data, and a senior sponsor, often the chief executive or a transformation lead, owns the overall direction and governance. What matters is that ownership is explicit and that leaders are genuinely aligned, because AI change cuts across every function and stalls when no one is clearly accountable.
How do we handle employee resistance and fear of AI?
Diagnose the cause first, because job-loss fear and simple skill gaps need different responses. Fear is eased by honest communication about how roles will change and what people will do with freed-up time, plus visible leadership commitment. Skill gaps are closed with role-based training and support. Involving employees in redesigning their own work, rather than announcing changes to them, is one of the most reliable ways to build genuine buy-in.
What AI governance frameworks should an enterprise know about?
The main reference points are the NIST AI Risk Management Framework, a voluntary United States framework built around govern, map, measure, and manage; ISO/IEC 42001, the international AI management system standard published in 2023; and, for anyone operating in Europe, the European Union AI Act, whose AI literacy duty under Article 4 has applied since February 2025. In Singapore and the region, the Infocomm Media Development Authority's Model AI Governance Framework is the key local reference.
How long does AI change management take?
There is no true end date, which is part of what makes AI different. Early wins from a focused use case can appear within a few months, but organisation-wide adoption, role redesign, and cultural change unfold over one to three years and then continue as tools evolve. Treat any promise of a quick, one-off fix with caution, and plan for sustained support rather than a single launch.
Do we need a specialist AI firm, or can our existing change partner do it?
It depends on how much your current partner has genuinely adapted for AI. A strong change firm brings valuable discipline, but AI adds workflow redesign, technical depth, governance, and continuous change that a classic playbook was not built for. Ask them the questions in this guide. If they cannot explain what they do differently for AI, add a partner who can, or choose a firm that combines change expertise with real AI capability.
Sources
All statistics in this article were checked against the primary source. Sources are first-hand from the original publisher unless noted.
- Boston Consulting Group, The Leader's Guide to Transforming with AI (2024, updated 2025), for the 10-20-70 split of algorithms, technology and data, and people and processes.
- McKinsey & Company / QuantumBlack, The state of AI: how organizations are rewiring to capture value (March 2025), for the 80-percent EBIT finding, workflow redesign, and CEO oversight of AI governance.
- MIT Project NANDA, The GenAI Divide: State of AI in Business 2025 (2025), for the finding that about 95 percent of organisations saw no measurable return and that the divide is driven by approach, not model quality. Widely reported; primary document circulated as a research report.
- Deloitte, State of AI in the Enterprise 2026 and the earlier State of Generative AI in the Enterprise series, for process-redesign and regulatory-barrier figures.
- Prosci, The correlation between change management and project success, for the finding that 88 percent of initiatives with excellent change management met or exceeded objectives, against 13 percent with poor change management.
- World Economic Forum, Future of Jobs Report 2025 (January 2025), for skills change and reskilling figures.
- Gartner, via Harvard Business Review, Employees Are Losing Patience with Change Initiatives (2023), and Gartner's 2025 HR research, for change-fatigue figures. Data first-hand from Gartner; the HBR article is a second-hand outlet.
- McKinsey & Company, Superagency in the workplace (January 2025), for the 46-percent skill-gap barrier figure.
- NIST, AI Risk Management Framework 1.0 (2023); ISO, ISO/IEC 42001:2023; European Commission, EU AI Act; and Singapore IMDA, Model AI Governance Framework, for governance references.
This article is for general information and does not constitute legal or financial advice. Regulations and product features referenced here can change, so verify current requirements against the primary source before acting.