Quick answer
You prepare a company for AI through people, not license purchases: start with trust (why we use AI and what will change), basic AI fluency for every employee, a few small pilots with measurable results, and leaders who use the tools themselves. In McKinsey's survey, 88% of organizations regularly use AI in at least one business function, yet nearly two-thirds have not started scaling it (McKinsey, 2025). The difference comes from the organization, not from access to the technology.
Why do AI initiatives stall after the pilot?
Most often because the organization is not ready for the change: goals are vague, communication is thin, and people do not know how AI fits into their daily work. The technology usually works. The trouble starts when a pilot has to become a habit for the whole team.
AI adoption is the state in which people deliberately and regularly use AI in their workflows, know when to trust it, and can judge its output. That is different from deployment, which only makes a tool available.
The data backs this up. In Anthropic's report on AI agents, 39% of organizations named change management needs as a major challenge to scaling, alongside integration and data quality. Small and mid-sized businesses are significantly more likely than large enterprises to struggle with the human side of adoption, including employee resistance and training needs (51%) (Anthropic, 2026).
That is why the first step to AI success is not code. It is communication.
Where do you start? Trust before tools
Start by explaining the purpose. AI changes how decisions are made and who makes them, so the conversation has to begin with "why," not with a feature list. Tie AI initiatives to goals people genuinely care about: faster processes, fewer repetitive tasks, and more time for work that takes real thinking.
Trust is not a given. In the global study by KPMG and the University of Melbourne, only 46% of people are willing to trust AI systems (KPMG, 2025).
You do not build trust with announcements. You build it into how the solution works:
- Sources with every answer. In our agentic knowledge base, every answer cites the source passages, and the system checks that the answer is grounded in them before returning it. Employees can see where a result came from and verify it.
- A person approves sensitive actions. In our AI agents, the outreach agent drafts messages, but nothing goes out without human approval.
- Clear data boundaries. When client data must not leave the company, we run models on our own GPUs without sending data to external APIs.
When people see AI as a partner that makes their work easier rather than a black box, adoption follows.
What AI skills do employees actually need?
Every employee needs a basic understanding of how AI fits into their work, but nobody has to become a data scientist. AI fluency (the EU AI Act calls it AI literacy) is the ability to use AI deliberately: knowing its capabilities, its limits, and its risks.
In practice it comes down to four skills:
- What AI can and cannot do. Strengths (summarizing, search, drafts, classification) and weaknesses (facts without sources, calculations, anything that needs context the model does not have).
- How to ask good questions. Framing a problem so AI can solve it: context, an example, the expected format.
- Comfort with data-driven decisions. Being willing to take a result that challenges intuition seriously and check it instead of dismissing it.
- How to spot when AI gets it wrong. Because it will.
That last skill is the weakest today. In the same KPMG study, 66% of employees who use AI at work said they rely on AI output without evaluating its accuracy, and 56% said they had made mistakes in their work because of AI (KPMG, 2025). Training lags behind usage: in Microsoft's 2024 survey, only 39% of people who use AI at work had received AI training from their company (Microsoft Work Trend Index, 2024).
Learning works best when it is practical: short workshops built on your own examples, regular office hours for questions, and explicit permission to experiment. Confidence grows when people feel safe to try.
What does the EU AI Act say about AI literacy?
If you operate in the EU, Article 4 of the AI Act has applied since February 2, 2025, to both providers and deployers of AI systems, which in practice means most companies that use AI tools at work (Regulation (EU) 2024/1689). The original text required them to ensure, to their best extent, a "sufficient level" of AI literacy among staff.
Regulation (EU) 2026/1744 (the Digital Omnibus on AI), in force since July 27, 2026, softened that wording: organizations must take measures that support the development of AI literacy, without having to guarantee a specific level for each person (Regulation (EU) 2026/1744). The duty did not disappear. It changed from an obligation of result to an obligation of effort.
What that means in practice, according to the European Commission's guidance (European Commission, Q&A):
- measures are tailored to roles, existing knowledge, the context, and the risk of the systems in use,
- no certificates are required, internal records of training and initiatives are enough,
- simply handing out a tool's instructions for use is usually not sufficient.
A literacy program with a training log is good practice anywhere, and for EU operations it is also the simplest evidence of due diligence.
How do you start small and show results fast?
Pick small, practical use cases that solve an everyday problem instead of launching a large transformation program. Big AI initiatives often stall because the payoff arrives too late for anyone to believe in it.
Good first use cases: automating reports, faster search across company documents, summarizing customer feedback, reading forms.
Share results in a simple "it used to take X, now it takes Y" sentence. For example: the weekly report used to be assembled by hand from several spreadsheets, now it builds itself and the team only checks it. A concrete comparison like that convinces people more than any slide about AI's potential.
Quick wins create advocates, spark curiosity, and build momentum faster than top-down mandates. For what decides whether a pilot reaches production, see our article on why AI pilots never reach production.
What is the leader's new role in AI adoption?
Leaders move from approving projects to enabling teams to experiment and learn. It is one of the clearest differences between companies that get value from AI and everyone else.
McKinsey identifies a small group of "AI high performers" (about 6% of respondents) that attribute more than 5% of EBIT to AI. Respondents at these companies are three times more likely to strongly agree that senior leaders demonstrate ownership of and commitment to AI initiatives, and much more likely to say those leaders actively drive adoption, including role modeling the use of AI (McKinsey, 2025).
Key leadership behaviors:
- encouraging collaboration across teams,
- rewarding curiosity and learning, not just results,
- communicating goals, progress, and limitations openly,
- using AI visibly and sharing what you learned.
We do this ourselves. We run pimento from a repository where projects, meeting notes, and procedures live as files, and an AI agent acting as a COO assistant takes notes, turns decisions into tasks, and prepares weekly summaries.
A role split for an adoption program can look like this:
| Role | Responsible for | Common mistake |
|---|---|---|
| Executive sponsor | Business goal, budget, personal example of use | Handing AI entirely to IT |
| Team managers | Choosing pilot tasks, time to learn, measuring impact | Expecting results without time to learn |
| AI champions in teams | Helping peers, collecting problems and ideas | A side role with no time in the calendar |
| IT and security | Approved tools, data access, logging | Bans with no alternative |
| HR, legal, privacy | AI literacy program, training records, data protection | A one-off training "for compliance" |
How do you overcome resistance to AI?
Resistance comes from uncertainty, so the best remedy is clarity. People want to know:
- which tasks will change,
- what new opportunities may appear,
- how success will be measured,
- what support is available.
A lack of clear rules does not stop people from using AI. It just pushes usage underground. In Microsoft's 2024 survey, 75% of knowledge workers used generative AI at work, and 78% of those users brought their own tools (Microsoft Work Trend Index, 2024). This is called shadow AI: using AI tools outside the organization's knowledge and control, often with data that should never leave the company.
Transparency builds trust. Show people they have a role in AI-assisted work and that the company supports their growth.
How do you make experimentation feel safe?
Set clear boundaries and give people room to try things inside them. Innovation thrives where teams can test and learn without fearing consequences for a failed experiment.
- Privacy and ethics boundaries. Write down on one page which data must never go into AI tools (personal data, trade secrets) and which tools are approved.
- An AI sandbox. A separate environment with approved tools and test data where people can experiment without risk to customers or production systems.
- Lessons from experiments. Recognize the ones that failed too, as long as the team knows why.
- Experimentation as part of the job. Learning time on the calendar, not after hours.
If you are deciding whether to build your own solution or start with an off-the-shelf tool, read our practical guide to building vs. buying AI.
How do you measure AI adoption?
Measure habits and process change, not just ROI. Financial returns arrive later than changes in how people work, and without adoption metrics you cannot tell whether that change happened at all.
Pay attention to where value comes from: according to McKinsey, high performers are nearly three times more likely to fundamentally redesign workflows when deploying AI, instead of adding a tool to an old process (McKinsey, 2025).
| Area | Example metric | How to measure |
|---|---|---|
| Usage | Share of people using the tool weekly | Tool logs, active vs. assigned licenses |
| Process impact | Task time before and after, amount of rework | Measure a sample of tasks before and after the pilot |
| Skills | Share of trained staff, ability to verify outputs | Training log, a short hands-on check |
| Trust and satisfaction | Usefulness rating, number of reported AI errors | Short quarterly survey, a reporting channel |
| Security | Use of unapproved tools, data incidents | Tool policy, incident review |
| Business outcome | Cost or revenue in the process | Comparison with the baseline after 3-6 months |
Keep governance light but consistent: a regular review of what works, a community of practice where people share examples, and updated rules as tools and needs evolve.
People first, technology second
AI transformation is not about replacing people with machines. It is about giving teams better tools, clearer insights, and more room for meaningful work. The organizations that succeed are not the ones with the biggest budgets or the most sophisticated algorithms. They are the ones where curiosity, collaboration, and clarity drive everyday decisions.
Start small, encourage experimentation, and keep the focus on people. Technology only delivers real value when your culture is ready to use it.
Sources
- McKinsey: The state of AI in 2025: Agents, innovation, and transformation (PDF, November 2025)
- Anthropic: The 2026 State of AI Agents Report
- KPMG and the University of Melbourne: Trust, attitudes and use of artificial intelligence: A global study 2025
- Microsoft Work Trend Index 2024: AI at work is here. Now comes the hard part
- Regulation (EU) 2024/1689 (AI Act)
- Regulation (EU) 2026/1744 (Digital Omnibus on AI)
- European Commission: AI literacy, questions and answers
