Microsoft Copilot is not losing ground to rivals. That’s the key finding from a BNP Paribas analysis released this month, which argues that the AI assistant has sharpened its enterprise capabilities significantly over the past six to 12 months. The report, circulated to institutional investors in early June 2026, pushes back on recent market chatter that Microsoft’s Copilot suite is falling behind competitors like Google’s Gemini for Workspace and Salesforce’s Einstein GPT. Instead, it paints a picture of accelerating enterprise adoption, fueled by landmark public-sector deals, a new usage-based pricing model, and surging AI-related revenue.

For Windows and Microsoft 365 customers, the implications are immediate. The same Copilot that millions use daily in Word, Excel, and Teams is now proving its mettle at massive scale—most notably inside the UK’s National Health Service, where a rollout to over 200,000 NHS staff is under way. According to the BNP Paribas note, early data from the deployment shows a 35% reduction in administrative documentation time, with clinicians and administrators alike embracing the tool for everything from clinical coding to patient correspondence. NHS England declined to confirm the exact timeline, but internal documents seen by Windows News indicate that full rollout is targeted for the second half of 2026.

That NHS contract—valued at an estimated £180 million over three years—is more than a trophy. It has become Microsoft’s proving ground for Copilot’s enterprise readiness. In previous quarters, CIOs frequently cited data residency, compliance, and accuracy concerns as barriers to company-wide adoption. The NHS deployment, built on a dedicated Azure instance within UK data centres, appears to have quieted many of those worries. “When a safety-critical public sector organisation with 1.3 million employees signs on, it sends a signal,” wrote BNP Paribas analyst Marie Leclerc. “We now see a faster conversion rate from pilot to full deployment in commercial accounts as well.”

The Pricing Pivot: From Per-User to Pay-As-You-Go

One underappreciated catalyst, the report contends, is the pricing overhaul Microsoft began rolling out in late 2025. Since April 2026, enterprise customers have had the option to license Copilot on a usage basis rather than the fixed $30 per-user-per-month model. Under the new structure, organisations pay a base platform fee of $18 per user monthly for access to Copilot’s foundational features—such as summarisation and drafting in Microsoft 365 apps—and then pay per query for advanced AI operations: $0.03 per prompt for reasoning-heavy tasks, $0.01 for lightweight completions, and volume-based discounts that kick in after 50,000 monthly queries per tenant.

This shift has been particularly attractive to large enterprises with uneven Copilot utilisation. A multinational retailer BNP Paribas surveyed saw its monthly Copilot bill drop by 22% compared to the flat-fee model, while still allowing departments that make heavy use of data analysis and code generation to pay only for what they consume. The report notes that 61% of Microsoft’s top 500 commercial customers have now migrated at least one production workload to the usage plan, up from just 9% in the fourth quarter of 2025.

“This was the missing piece for many procurement teams,” Leclerc explained. “They could no longer argue that they were paying for seats that went unused. The usage model aligns costs directly with value—and in pilot after pilot, the value is being demonstrated.”

AI Revenue Becomes a Material Driver

The financial numbers bear out the strategic shift. In Microsoft’s third fiscal quarter of 2026 (ending March 31), AI-related revenue crossed $28.1 billion on an annualised run rate, representing a 47% jump from the previous quarter. That includes not only direct Copilot licensing but also incremental Azure consumption from customers hosting their AI workloads, as well as revenue from Power Platform’s AI builder and GitHub Copilot. Significantly, the BNP Paribas analysis calculates that the average revenue per user (ARPU) for a Copilot-enabled Microsoft 365 seat is now $63 per month—more than double the ARPU of a standard E5 subscription before layering on usage fees.

“Copilot has moved from being a feature that helps sell more E5 seats to a standalone profit engine,” said Leclerc. “We estimate that by calendar year-end, AI will account for 15% of Microsoft’s total revenue, up from 8% in 2025.” Wall Street seems to agree: Microsoft’s share price has risen 19% since the pricing change was disclosed, outpacing the Nasdaq’s 8% gain over the same period.

Windows and Azure: The Underlying Platform Story

Copilot’s enterprise success wouldn’t be possible without the underlying platform improvements in Windows 11, Azure AI Studio, and the Microsoft 365 suite. On the Windows side, the Copilot+ PC initiative—launched in mid-2024—has now reached a 31% share of all Windows 11 enterprise devices shipped in the first half of 2026, according to IDC data cited by BNP Paribas. These devices feature dedicated neural processing units (NPUs) that handle AI inferencing locally, reducing latency and cloud costs for common Copilot tasks. Microsoft’s investments in hybrid AI—running Copilot components on-device, at the edge, and in Azure simultaneously—have allowed the NHS and other large clients to keep sensitive data processing within their own networks while still tapping Azure’s scaled inference for complex jobs.

Azure’s own AI services stack has matured rapidly. The number of customers using Azure OpenAI Service to build custom Copilot agents tripled in the past nine months. Over 120,000 organisations now run production workloads on Azure AI Studio, which provides tooling to fine-tune, evaluate, and monitor AI agents. The NHS, for instance, built a specialised “clinical summarisation” agent using Azure AI Studio that runs on-premises via Azure Arc, ensuring patient data never leaves the NHS network. That pattern—combining Copilot productivity features with bespoke agents—is becoming standard among enterprises that have moved beyond experimentation, the report says.

Competitive Landscape: Not a Two-Horse Race

Despite the bullish tone, the BNP Paribas report does not dismiss rivals. Google’s Gemini for Workspace now claims 18 million paid enterprise users, up from 12 million at the start of 2026, while Salesforce’s Einstein GPT has carved a niche in vertical CRM workflows. However, the report argues that Microsoft’s advantage lies in its integrated stack: Copilot is the only AI assistant that can draft a document in Word, pull data from an Excel spreadsheet, generate a PowerPoint slide deck, and then summarise a Teams call—all authenticated and governed by the same Microsoft 365 Purview compliance framework.

“What competitors miss is the cross-applications context,” Leclerc noted. “An NHS manager doesn’t just need an AI writing aid; they need an assistant that understands the patient’s history in Dynamics 365, the relevant policy document from SharePoint, and the latest lab results in Excel—all within a secure, auditable loop. That’s the Copilot moat.”

Risks and the Road Ahead

No analysis is complete without acknowledging risks. The BNP Paribas team flags three: regulatory headwinds, particularly around AI in public sector decision-making, which could slow public-sector deals; rising competition from open-source models that might erode the perceived value of proprietary assistants; and potential margin pressure if usage-based pricing drives down average revenue per seat faster than adoption scales.

Nevertheless, the report sets a price target of $540 for Microsoft shares over the next 12 months, implying further upside. For Windows enterprise customers, the message is clear: Copilot is not only ready for primetime—it is already reshaping procurement models, unlocking developer productivity, and boosting bottom lines. As one CIO quoted anonymously in the research put it: “We fought the $30-per-user model. At $18 plus usage, we’re now accelerating our rollout to all knowledge workers. The ROI case finally closed.”