Change management for adopting AI

1. Why 70% of AI projects fail (and it isn’t the technology’s fault)

According to McKinsey research, about 70% of digitalisation programmes fail to reach their stated objectives. When we narrow the field to artificial intelligence projects, the percentage stays the same, if not worse. The surprising finding is that in the vast majority of cases the problem doesn’t lie in the technology: the algorithms work, the platforms are mature, the data exists. The real obstacle is human, organisational, cultural.

In Italian SMEs, this dynamic is even more marked. The business owner buys a promising AI solution, installs it, trains a few people and expects everything to change overnight. Six months later, the platform is under-used, the team has gone back to its old habits and the investment is filed away as a "failed experiment". It isn’t a problem of budget or technical skills. It is a problem of change management.

Change management for AI adoption is not an ancillary activity, a "nice to have" to add to the project if there is time left over. It is the project itself. Without a structured plan to manage organisational change, even the best AI technology will remain a foreign body inside the company, rejected by the organisational immune system.

In this guide we look in depth at a proven approach to change management for AI in Italian SMEs, based on the "Mindset First" framework that we put into practice every day at Athena AI. An approach that starts from people before technology, uses mixed business-technical teams and adopts an Agile methodology with Sprints to guarantee concrete, measurable results in a short time.

2. Cultural resistance to AI in Italian SMEs

Before talking about solutions, it is essential to understand the specific resistance that AI meets in the Italian business landscape. This isn’t generic resistance to change: artificial intelligence brings with it unique fears and perceptions that must be addressed with targeted tools.

The fear of replacement

It is the most visible and most discussed form of resistance. The employee who hears talk of "AI automation" in their department immediately thinks of losing their job. In Italy this fear is amplified by a media context that emphasises the most catastrophic forecasts and by a work culture in which professional identity is deeply tied to the role and to the skills acquired over years of experience. The skilled worker who took twenty years to master their trade sees AI as an existential threat, not an opportunity.

The generational gap

Italian SMEs often have three or four generations working side by side with radically different attitudes towards technology. The founder in their sixties who built the company "with their own hands" distrusts what they can’t touch or see. The fifty-year-old manager learned to use the ERP with difficulty and doesn’t want to start again from scratch. The thirty-year-old is enthusiastic but frustrated by the slowness of change. The Gen Z new hire takes for granted features that senior colleagues consider science fiction. This diversity is not a problem in itself, but it becomes explosive if it isn’t managed.

The myth of all-powerful (or useless) AI

A typically Italian phenomenon is the polarisation of expectations. On one side, those who believe AI solves any problem as if by magic: "Let’s put in AI and the problems disappear". On the other, those who dismiss it as a passing fad: "We’ve seen plenty of these, this one will pass too". Both attitudes are harmful. The first generates disappointment when results aren’t immediate. The second prevents you from even starting.

The "we’ve always done it this way" syndrome

In companies with a long tradition, established processes become almost sacred. The proposal to change a workflow that "has worked for thirty years" meets visceral resistance. It doesn’t matter whether that process is inefficient, redundant or expensive: it is known, predictable, reassuring. AI, by its nature, requires processes to be rethought, and this represents a direct challenge to the organisational comfort zone.

Distrust of data

Many Italian managers, especially in SMEs, make decisions based on experience and intuition. And they are often excellent decisions, refined by decades of practice. AI asks you to trust the data, to accept that an algorithm can identify patterns that the human eye doesn’t see. This doesn’t mean that intuition is wrong, but integrating human experience with algorithmic analysis requires a deep change of mindset.

3. The "Mindset First" framework: changing mindset before technology

The heart of the Athena AI methodology is the "Mindset First" principle: before introducing any AI technology, we work on the mindset of the people who will have to use it, manage it and live with it every day.

This doesn’t mean organising a couple of motivational workshops and then starting with the technical implementation. It means building a structured path in which each phase prepares the ground for the next, creating the cultural and organisational conditions for AI to be adopted in a natural and sustainable way.

The move from phase 1 to phase 3 is neither linear nor automatic. It takes time, patience and above all consistency. The most common mistake is to skip the awareness phase to "save time" and go straight to technical training. The result is predictable: people who know how to use the tool but don’t understand why they should, and so they don’t.

In a typical project, we devote the first 3-4 weeks exclusively to the awareness phase. It seems a lot, but it is an investment that pays back enormously in the following phases. A team that understands the "why" approaches the "how" with a completely different attitude.

4. The 4 pillars of change management for AI

Once the Mindset First framework is established as the guiding principle, change management for AI is structured around four operational pillars. Each one is indispensable: neglecting even one compromises the entire path.

The four pillars are not sequential phases but parallel dimensions that must be looked after simultaneously for the whole duration of the transformation path. In every Sprint of the project, we set aside explicit time for activities on each of the four pillars.

5. How to build a mixed AI team with an Agile approach

One of the most frequent mistakes in AI implementation is delegating the project entirely to the IT department or, worse still, to an external supplier without any internal involvement. The Athena AI approach is radically different: we build mixed teams that combine technical and business skills, internal and external, junior and senior.

The ideal composition of the mixed team

An effective mixed AI team in an Italian SME typically includes 5-8 people with complementary roles. The process owner (who knows the workflow to be automated in detail), one or two end users (who will use the tool every day), an internal IT contact (for integrations and infrastructure), a data analyst or controller (for data quality and metrics), the external AI consultant (for specialist skills and methodology). None of these roles is optional.

The value of the mixed team lies in the diversity of perspectives. The technician sees the possibilities of the algorithm. The end user sees the nuances of the real process that no document describes. The process owner sees the interdependencies with other departments. The controller sees the numbers. When these perspectives meet in a structured way, the solutions that emerge are infinitely better than those designed from a single point of view.

The Agile approach with Sprints

The mixed team works with an Agile methodology, organising the project into Sprints of 2-3 weeks. Each Sprint has clear objectives, defined deliverables and a review with the stakeholders. This approach is particularly effective in the AI context for several reasons.

First of all, it reduces risk. Instead of investing months in a solution that might not work, the team delivers working increments every 2-3 weeks. If something goes wrong, feedback arrives quickly and the correction is inexpensive. Secondly, it maintains momentum. Short Sprints with visible results keep motivation high and give participants a sense of concrete progress, countering the frustration that often accompanies long transformation projects.

The retrospective is the most underrated and most important moment of the Sprint. That is where latent resistance, communication problems and unexpressed fears come out. A good facilitator knows how to turn these moments into opportunities for growth for the whole team.

6. The role of leadership: sponsor, champion and ambassador

No change management project succeeds without the active support of leadership. But "active support" doesn’t simply mean approving the budget and asking for monthly updates. It means taking on three distinct and complementary roles which, in an SME, often converge on the same people.

The executive Sponsor

The sponsor is the member of top management who "puts their face" to the AI project. In the Italian SME this is almost always the business owner or the managing director. Their role is to state publicly and repeatedly the strategic importance of the project, remove the organisational obstacles the team encounters, allocate the necessary resources (not just budget, but above all the time of key people) and connect the AI project to the company’s long-term vision. An absent or barely visible sponsor sends a devastating message: "This project isn’t really important". And people act accordingly.

The AI Champions

AI Champions are key figures who act as a bridge between the project team and the organisation. They are people respected by their colleagues, curious about technology and with good interpersonal skills. They don’t necessarily have to be the most expert in technology: they have to be the most credible. In a manufacturing SME, the department head with thirty years of experience who says "I tried this AI tool and it really works" has a hundred times more impact than any presentation by the external consultant.

We typically identify 2-3 AI Champions in companies of up to 50 employees and 5-8 in larger companies. We train them in depth, involve them in designing the solutions and give them visibility and recognition for the role they play.

The Ambassadors

Ambassadors are the widespread network of change. They are the first enthusiastic users who, spontaneously or with a little encouragement, share their positive experience with their colleagues. Unlike Champions (who are appointed), Ambassadors emerge naturally during the project. Our task is to recognise them, value them and give them the tools to amplify their message. In a 100-person company, having 15-20 Ambassadors spread across all departments is the best guarantee of widespread adoption.

7. Managing the fear of "AI stealing jobs"

The issue deserves a section of its own because it is the deepest form of resistance and the hardest to manage. It isn’t enough to say "AI doesn’t replace people, it empowers them". It is a true statement, but too generic to reassure those who genuinely fear for their professional future.

The approach we take at Athena AI is radical transparency combined with concrete actions. Transparency means openly admitting that AI will change some roles and some tasks. Denying it would be dishonest and counterproductive: people aren’t stupid, and false reassurance destroys trust. But change doesn’t mean elimination: it means evolution.

In concrete terms, this translates into a reskilling and upskilling plan that accompanies every AI project. If automating reporting frees up 10 hours a week of the controller’s time, those 10 hours are reinvested in higher value-added activities: predictive analysis, internal consultancy to departments, development of new KPIs. The employee doesn’t lose their job: they change job, and change it for the better.

In Italian SMEs we have found that the most effective upskilling paths are those that combine formal training (courses, certifications) with on-the-job learning (shadowing, mentoring, pilot projects). Purely theoretical training doesn’t work: people need to see how the new skills apply to their specific context.

A telling case: in a manufacturing company in the Veneto region, the introduction of an AI-based predictive maintenance system initially terrified the maintenance technicians, who were convinced the machine would replace them. After a targeted training path, those same technicians became "predictive maintenance specialists", a role with higher skills, greater decision-making autonomy and higher pay. Today they are the most enthusiastic supporters of AI in the company.

8. Adoption metrics: how to measure the success of change

Change management is not an act of faith. It is a measurable process, and it must be measured rigorously so that it can be managed and improved Sprint after Sprint. But traditional project metrics (time, budget, scope) are not enough: you need metrics specific to adoption.

Quantitative metrics

Qualitative metrics

We recommend creating an adoption dashboard visible to the whole organisation, updated in real time. Transparency on the data creates accountability and celebrates progress, feeding a virtuous circle of motivation and adoption.

9. Case study: a typical path in a manufacturing SME

To make the principles described so far concrete, we reconstruct a typical change management path for AI adoption in an Italian manufacturing SME. The details are based on real experiences, aggregated and anonymised.

Sprint 0 (Weeks 1-3): Mindset First

The path begins with a phase of listening and awareness. We conduct individual interviews with 15 key people, from the production director to the line operator, to map perceptions of AI. The expected resistance emerges: fear for jobs (especially among the over-50s), scepticism among department heads ("we already have the MES that nobody uses"), enthusiasm but impatience among the young. We organise a half-day workshop with the whole company, in which we present AI in a concrete, demystified way, showing examples of similar companies that have adopted it successfully. The business owner opens the workshop with a personal talk explaining the strategic why of the project and commits publicly: no job will be cut because of AI.

Sprint 1 (Weeks 4-6): The mixed team and the first quick win

We form the mixed team: the production manager, a department head (chosen because sceptical but respected), a senior operator, the IT manager, the controller and our AI consultant. The first Sprint focuses on a high-impact, low-complexity quick win: a demand forecasting system based on historical order data. The sceptical department head is involved from day one in defining the requirements. At the Sprint review, when the system shows more accurate forecasts than his manual estimate, his attitude changes visibly. His comment at the retrospective: "I didn’t believe it, but the numbers speak for themselves".

Sprints 2-4 (Weeks 7-15): Expansion and training

The success of the quick win creates the momentum to expand the project. The sceptical department head becomes the first AI Champion and starts speaking positively about the experience with colleagues. In parallel, we launch the three-level training programme: a 2-hour session for all employees (AI literacy), an 8-hour course for the 20 direct users of the forecasting system, advanced training for the 3 AI Champions. Each Sprint adds features to the system and involves new users. The adoption metrics are shared in a dashboard displayed in production on a dedicated monitor.

Sprints 5-8 (Weeks 16-27): Predictive maintenance and scaling

The second AI project, predictive maintenance, is launched with a mixed team that now also includes two maintenance technicians. Thanks to the Mindset First path and the visible success of the first project, resistance is significantly lower. The technicians, initially the most frightened, realise that AI doesn’t replace their experience but empowers it: the system flags anomalies, but they are the ones who decide on the intervention. At the end of Sprint 8, the rate of unplanned machine downtime has fallen by 35%.

Results at 6 months

The most significant result, however, is not quantitative. It is the fact that, in the sixth month, three proposals for new AI applications came spontaneously from the operational teams, without any prompting from management or the consultants. The cultural change had taken place.

10. Conclusion: AI is a path, not a product

Change management for AI adoption is not a project with an end date. It is a path of continuous transformation that requires strategic vision, operational patience and a constant commitment to people.

Italian SMEs have an often underrated competitive advantage on this path: their small size allows more direct communication, more genuine relationships and a speed of change that large organisations can only envy. The business owner who speaks directly with their employees, who puts themselves on the line in person, who shows through their actions that change is a priority, has a transformative power that no structured change management programme can ever replicate.

The Mindset First framework, mixed teams, the Agile approach with Sprints, the four pillars of change management: they are powerful tools, but they are tools. What makes the difference is the genuine will to put people at the centre of change, before, during and after the technology.

If there is only one message to take home from this guide, it is this: invest in people at least as much as you invest in technology. Better still, invest in people first. The most sophisticated AI in the world is useless if nobody uses it. The simplest AI in the world changes everyday work, if everyone uses it.

The time to start is now. Not because the technology is ready (it has been for a while), but because your competitors are already building their AI-based competitive advantage. Every month of delay is a month of lost opportunities.