From trial to company-wide adoption
The deployment progressed through three stages:
- July 2025: 7,500 trial licenses across departments.
- January 2026: Expansion to 30,000 licenses.
- April 2026: Approximately 100,000 licenses.
Fujitsu supported adoption with its Fujitra transformation network, hands-on training, videos, and use-case awards. Microsoft reports usage above 90% among employees with access in June 2026; some customer and security-restricted environments remained excluded.
For enterprise IT teams, the useful lesson is organizational rather than numerical. A cross-department trial can test whether an assistant helps distinct workflows, while training and internal knowledge sharing give employees something more useful than an activation email.
The reported usage figure nevertheless needs careful handling. The case study does not define its measurement window or activity threshold. It should not be interpreted as proof of daily use, consistent output quality, or financial return.
The service-desk result—and its boundary
Fujitsu’s clearest operational example is a service desk handling nearly 300,000 inquiries annually. AI agents reportedly automate roughly 75% of initial-response and triage work. That is not a claim that three-quarters of inquiries are resolved without people.
This distinction matters. Classifying a request and delivering an opening response are different from diagnosing a fault, authorizing a change, or confirming that an employee can work again.
For administrators evaluating comparable automation, a sensible measurement plan would separate:
- Intake performance: How quickly requests receive an initial response.
- Routing quality: Whether classification sends tickets to the correct team.
- Resolution quality: Whether issues are actually fixed.
- Human workload: Whether automation reduces effort or merely moves it into correction and review.
These are analytical recommendations, not additional results reported by Fujitsu. They help prevent an impressive automation percentage from becoming a catch-all productivity claim.
Copilot as an entry point, not the entire system
Fujitsu envisions Teams and Outlook as familiar entry points to specialized agents and ERP or CRM systems. Its Customer Zero strategy uses internal experience to inform customer services, combining external AI with technologies such as Kozuchi, Takane, and FUJITSU-MONAKA. The broader multi-agent environment remains an early-stage ambition.
The practical attraction is straightforward: employees could spend less time navigating separate systems. But interface simplicity should not be confused with operational simplicity. Each connected workflow still needs clear ownership, reliable information, and appropriate access controls.
What Microsoft 365 administrators should check
Microsoft Learn provides an important technical qualification: Copilot can reference only content the user is authorized to access, and SharePoint and OneDrive controls affect discovery without changing the user’s permissions.
That makes existing permissions a deployment concern, not yesterday’s housekeeping. An organization should not assume that introducing Copilot will repair overly broad access.
Microsoft’s governance guidance recommends identifying overshared, sensitive, ownerless, and inactive content using Microsoft Purview and SharePoint Advanced Management. It also describes removing excessive access, correcting permission inheritance, confirming ownership, and applying interim discovery or data-loss-prevention protections while remediation proceeds.
For an IT team drawing lessons from Fujitsu, the actionable contribution is to run two workstreams together: employee adoption and information governance. Training helps people ask useful questions; permission remediation helps ensure the answers draw on appropriately accessible material.
Adoption is not yet proven financial return
Fujitsu is still developing indicators connecting AI adoption to financial outcomes. Microsoft’s account supplies no quantified return on investment, and it is a vendor-published customer story rather than an independent evaluation.
The balanced conclusion is therefore narrower—and more useful—than “100,000 licenses prove AI transformation.” Fujitsu offers a concrete example of staged deployment, adoption support, and bounded process automation. Its next test is whether those activities produce sustained improvements in service quality, employee workload, and business results. A license count starts that conversation; it does not finish it.