Digital Insurance published an August 11 interview in which Manulife technology executive Aaron Minkovich described the agreement as the foundation for an “AI-powered organization,” citing a roughly 30% productivity increase among developers using assisted and autonomous capabilities. The agreement itself was announced by Manulife and Microsoft on July 22, 2026, and Dow Jones also reported the core terms: five years, adoption of Frontier Suite, deployment of Agent 365, and Copilot expansion beyond 30,000 workers.
The meaningful operational change is that Manulife is buying the governance layer around AI agents at the same time it expands the end-user AI layer. That is the part many Copilot rollouts have lacked. A company can distribute Copilot licenses quickly; controlling which agents can access which data, act under which identities, and generate which audit trail is a longer project.
Agent 365 turns AI use into an identity and governance project
Microsoft launched Microsoft 365 E7 and Agent 365 into general availability on May 1, 2026. E7 bundles Microsoft 365 E5, Microsoft 365 Copilot, Microsoft Entra Suite, and Agent 365, with Defender, Intune, and Purview capabilities intended to extend existing identity, endpoint, data-protection, and compliance controls to AI use.
Microsoft describes Agent 365 as a control plane for observing, managing, governing, and securing agents. In practical terms, that makes it an administrative answer to the question many large organizations now face: which AI agents exist, who owns them, what services and data can they access, and how can security teams revoke or constrain those permissions?
Manulife’s own July 22 release says its global enterprise AI platform is still in pilot, even as it already has agentic tools and AI workflows operating in parts of the business. The company says Azure and Microsoft Foundry will support everything from citizen-development tools to advanced model fine-tuning and monitoring, while Agent 365 will become an enterprise-wide registry and management interface for agents.
That wording is important. Manulife is not announcing that Agent 365 has automatically secured every AI workflow it already runs. It is announcing the platform it intends to use to bring those workflows under a common governance model. The hard work is likely to be the inventory, ownership assignment, access design, policy mapping, logging, and migration decisions needed to make the registry represent reality.
Microsoft’s documentation also puts limits around the product’s reach. At general availability, Agent 365 is licensed per user: agents managed or owned by a licensed user are covered by that user’s Agent 365 or E7 license. That is different from treating every autonomous workflow as a standalone managed asset regardless of who created it. For an insurer with employee-built automations, developer-built agents, and third-party services, clean ownership records will be a prerequisite for dependable controls.
Microsoft also changed Agent 365 licensing rules on June 1, requiring Microsoft 365 E5 for new enterprise purchases of the standalone product. E7 sidesteps that prerequisite because E5 is included. Manulife’s decision to use the larger E7 bundle therefore buys more than Copilot access; it standardizes the foundational security and identity stack Microsoft now treats as necessary for its agent-control platform.
The 30,000-user figure does not mean every Manulife worker is moving to Copilot
The 30,000-plus Copilot deployment is a large number, but it should not be read as a first-time enterprise AI rollout. Manulife reported in its 2025 annual report that its proprietary generative AI assistant, ChatMFC, had reached 100% of its global workforce, including contingent workers. The Microsoft agreement layers a commercial productivity suite onto an organization that has already spent years putting a company-specific AI assistant in employees’ hands.
That distinction changes the adoption challenge. Manulife is not starting from zero on AI literacy, but it now has to determine where Microsoft 365 Copilot overlaps with ChatMFC, where the two tools serve different purposes, and where staff should use one rather than the other. The company has not disclosed how many employees will retain access to both tools, whether ChatMFC will be consolidated into the Microsoft stack, or how information boundaries will differ between the products.
Those omissions matter for desktop administrators and security teams. Two approved assistants can create two sets of data connectors, two development paths, two support models, and two answers to what employees should use for a given task. A workforce can be technically “covered” by AI while still receiving inconsistent guidance about where sensitive work belongs.
Manulife says its Copilot rollout will include training, adoption programs, governance frameworks, and responsible-use guidance. It also says employees already use Copilot for meeting preparation, document summaries, communications drafting, analysis, and knowledge discovery. Those are familiar low-friction use cases, but the company’s stated ambitions go further: underwriting, adviser support, operational automation, and increasingly complex agentic workflows.
The security bar rises sharply when an AI tool moves from summarizing a meeting to preparing preliminary underwriting assessments or processing insurance servicing documents. Manulife’s public materials say GenAI systems support more than 110 million calls annually in North America and are being extended into Asia. They also cite an AI underwriting tool at John Hancock that can produce a preliminary assessment in under 15 minutes. These are business-process claims, not merely employee-productivity claims, and they are why auditability and human escalation cannot be treated as back-office features.
The ROI claim is substantial, but the contract price is undisclosed
Manulife says it generated approximately C$300 million in enterprise AI value by the end of 2025 and expects to exceed C$1 billion by 2027. Its 2025 annual report supports the broader direction: as of December 31, 2025, the company reported 91 AI use cases in production and 121 in development. That is a sizable implementation pipeline for a financial-services company, and the target is part of Manulife’s publicly stated corporate strategy rather than a number introduced only for this Microsoft deal.
The company attributes some value to specific operational gains. In the Digital Insurance interview, Manulife said more than 80% of developers use GitHub Copilot and that developer productivity rose about 30%. It also said AI removed more than 400,000 manual intake and indexing tasks from U.S. claims and servicing operations. Manulife and Microsoft separately said the company rebuilt its mortgage-renewal application in weeks using its expanding AI-enabled development capabilities.
But the C$1 billion figure is an enterprise value target, not a disclosed, audited revenue line or a published net-profit forecast. Manulife has not provided a public breakdown showing how much is expected from reduced labor time, avoided technology spending, higher sales, improved retention, faster underwriting decisions, or lower claims-handling cost. Nor has it disclosed the contract value, number of E7 licenses, number of Agent 365 licenses, Azure consumption commitment, or a timeline by region and business unit.
The list-price math shows why those missing figures are material. Microsoft lists E7 at US$99 per user per month. If all 30,000 Copilot users also received E7 at list price for the full five years, the bundle alone would imply roughly US$178.2 million before discounts, regional terms, consumption charges, or taxes. That is only an illustrative ceiling, not the cost of Manulife’s agreement: the companies have not said that all 30,000 employees receive E7, and enterprise contracts are frequently priced below retail. Still, it demonstrates that the C$1 billion value objective must cover more than text-generation time savings to justify the scale of the commitment.
Manulife’s public record carries a personnel discrepancy
There is one detail in the August 11 Digital Insurance interview that does not line up with Manulife’s own current public record. The interview identifies Aaron Minkovich as Manulife’s chief technology officer. Manulife’s July 22 Microsoft announcement instead names Shamus Weiland as Chief Technology & Operations Officer, and Manulife’s May 21 executive leadership announcement said Weiland’s expanded role would take effect July 1.
Manulife’s public leadership page had previously described Weiland as Global Chief Information Officer, while the company’s 2026 leadership changes elevated Jodie Wallis to Global Chief AI Officer and moved Weiland into the combined technology-and-operations position. No Manulife announcement located during reporting explains Minkovich’s title or indicates that he replaced Weiland. That does not alter the underlying agreement, but it does make the interview’s executive attribution unclear.
For a company presenting governance as central to its AI program, accountability should be equally clear. The officials responsible for technology architecture, AI policy, model risk, operations, and customer outcomes need defined public roles when systems are being pushed deeper into insurance decisions.
The next milestone is controlled deployment, not more licenses
Manulife’s annual report is unusually direct about the risks it sees in its own AI plans. It warns that AI can produce inaccurate, incomplete, or ineffective results; that unexpected outcomes can cause operational and reputational harm; that technology may be embedded in processes poorly suited to digitization; and that vendor capabilities may fall short of what the company needs. The report says Manulife tests for hallucinations, monitors outcomes, and uses AI councils and governance processes to manage the risk.
That candor supplies the real test of the Microsoft expansion. Copilot licenses and a five-year contract establish capacity, while E7 and Agent 365 provide tools for control. Neither substitutes for deciding which customer data may be exposed to which agent, ensuring preliminary underwriting remains genuinely preliminary, recording consequential decisions, and maintaining a workable route for employees and customers to challenge bad outputs.
Manulife has committed to a specific scale: more than 30,000 Copilot users, a company-wide agent registry, and C$1 billion-plus in enterprise AI value by 2027. Its success will be measured less by how many workers can prompt Copilot than by whether its governance program can keep pace with the agents that are allowed to act on the answers.