In July 2026, the city of Raisio, Finland, quietly wrapped up a Microsoft 365 Copilot training program for nearly 100 employees that most organizations would do well to copy. Rather than simply flipping on licenses and hoping for productivity gains, Raisio spent months building role-specific workshops, co-creating AI usage guidelines, and tightening data governance—and the payoff was a workforce that actually understood both the power and the limits of generative AI.

Raisio’s Copilot Experiment in Detail

Raisio, a municipality of about 25,000 people, launched its Copilot initiative in autumn 2025 as part of its 2026 city strategy, which aimed to turn data into a daily asset for decision-making and public service. The city already owned Microsoft 365 Copilot licenses, but instead of a blanket deployment, it partnered with Sogeti, Capgemini’s technology arm, to design a phased learning path.

The program covered a wide range of municipal functions—education, infrastructure, HR, communications—and was structured into three parts: an introductory training for all participants, hands-on workshops built around actual job tasks (like drafting reports, summarizing citizen feedback, or searching across internal systems), and a forward-looking module on process automation.

Nearly 100 staff took part. Pre- and post-project surveys, cited in a Capgemini case study, showed a measurable reduction in employee hesitation toward AI and a clearer grasp of what Copilot could—and could not—do. Participants reported using Copilot for producing documents faster, summarizing long email threads, pulling information from across SharePoint and OneDrive, and managing communications.

But equally important: Raisio did not release hard productivity metrics, cost savings, or adoption rates after training. This was a deliberate people-first approach, not a numbers-first sprint. “We wanted to create a safe and supportive learning environment that acknowledged the wide range of roles and allowed everyone to learn at their own pace,” said Eero Rostiala, Raisio’s CIO, in the case study.

Why This Matters for Your Microsoft 365 Tenant

For everyday Microsoft 365 users

Copilot isn’t magic. The Raisio project underscores that the tool works best when you understand its boundaries. Participants learned that AI can draft and summarize, but it doesn’t replace judgment. If your organization rolls out Copilot with a similar training-first approach, you’ll likely spend less time wrestling with the tool and more time putting it to work on tasks that genuinely save effort—like condensing a 20-page council report into a one-page briefing note.

For power users and IT pros

The real story here is governance. Microsoft 365 Copilot draws from your organization’s existing data in Microsoft Graph—SharePoint sites, Teams chats, OneDrive files, Exchange emails. If those repositories have lax permissions or stale access controls, Copilot will happily surface sensitive information to anyone who asks the right question. Raisio and Sogeti didn’t just train employees on prompts; they co-created AI usage guidelines that covered privacy, security, and appropriate use. That step should be mandatory for any Copilot rollout, and it’s a job that falls squarely on IT admins, compliance officers, and security teams.

For administrators

Licensing Copilot is the easy part. Before you assign a single license, audit your tenant. Identify SharePoint sites with overly broad permissions, check sensitivity labels, review retention policies. Then, build a training curriculum that maps to the real workflows of different departments. Raisio’s approach—assessing each group’s needs before designing workshops—is far more effective than a generic “How to Use Copilot” webinar. Finally, make self-study materials available and establish a feedback loop; continuous learning is the only way AI adoption sticks.

The Backstory: From Hype to Hard Lessons

In early 2025, Microsoft began aggressively pushing Microsoft 365 Copilot into enterprises, touting it as a revolutionary “AI assistant” for work. Major system integrators like Infosys, TCS, and Wipro announced plans to scale Copilot to over 300,000 employees by 2026, as reported by Microsoft News. But amidst the hype, a quieter counter-narrative emerged: many early adopters found that employee trust, data hygiene, and task-specific relevance were bigger adoption barriers than the technology itself.

Raisio’s experience fits into a broader trend where smaller, risk-averse organizations—especially in the public sector—are choosing cautious, training-heavy rollouts. The city’s mayor, Eero Vainio, likened the moment to the early Industrial Revolution, when workers feared the spinning jenny. “The organizations and employees who approach AI with curiosity will be the ones who maintain their relevance,” he said.

And there’s a practical reason for this caution. In a Microsoft 365 tenant, Copilot’s conversational interface can turn old oversharing into an instant data breach. A poorly configured SharePoint library that everyone in the organization could already access becomes trivially discoverable when an employee asks Copilot “show me the confidential budget proposals.” By making governance a core part of the training, Raisio addressed the root cause rather than the symptom.

A Six-Step Plan for Your Own Deployment

If your organization is considering—or already deploying—Microsoft 365 Copilot, here are concrete steps you can steal from Raisio’s playbook:

  1. Start with a permissions cleanup. Before enabling Copilot, conduct a full audit of your Microsoft 365 data estate. Close open sharing links, review external guest access, and tighten site permissions. Treat this as a prerequisite, not an afterthought.

  2. Create an AI acceptable-use policy. Draft clear guidelines that address what employees can and cannot do with Copilot, how data is handled, and what constitutes responsible use. Involve legal, HR, and compliance teams. The Raisio-Sogeti policy was co-created alongside training, ensuring it felt practical rather than theoretical.

  3. Design role-specific training. Don’t settle for a one-size-fits-all session. Map Copilot’s capabilities to the actual tasks of different teams—HR might use it for summarizing candidate feedback; public works could draft maintenance reports from spreadsheet data. Hold hands-on workshops where employees use their own files and scenarios.

  4. Pilot before scaling. Raisio started with nearly 100 employees across multiple functions, far from an enterprise-wide switch-on. Pick a representative group, gather feedback, and adjust your training and governance before expanding.

  5. Measure sentiment and behavior, not just speed. The city’s pre- and post-surveys tracked hesitation and understanding, not just “emails written per hour.” Repeat such surveys at intervals to gauge whether AI is actually becoming a trusted tool or a source of anxiety.

  6. Provide ongoing self-study resources. The training didn’t end with the live sessions; Raisio made all materials available for continuous learning. Maintain a repository of use-case examples, how-to guides, and policy updates.

What Comes Next

Raisio is continuing its AI work within its 2026 strategy, and other municipalities are watching. As more public-sector bodies follow suit, expect to see a growing demand for packaged “training plus governance” consulting services—and perhaps more features from Microsoft to help admins manage AI exposure. But the core lesson is unlikely to change: successful AI adoption is about culture and caution, not speed. For IT pros, the most important Copilot-related task in 2026 may not be deploying the tool, but cleaning up the data it can see.