Microsoft has announced that Microsoft 365 E7 will be available from 1st May. On the surface, it’s a new bundle, but in practice, it’s a signal of where things are heading.
At the centre of this shift is the idea of the Frontier Firm.
Microsoft uses this term for organisations moving beyond AI as a helpful assistant, and towards AI as something that actively gets work done. Not in pockets or pilots, but across the business. It’s a useful way to frame what many teams are already working towards, whether they call it that or not.
E7 is built with that model in mind.
What is Microsoft 365 E7?
Microsoft 365 E7 is positioned as the Frontier Suite. A single, integrated platform for organisations that want AI to operate across the whole business, not just in isolated use cases.
It brings together:
- Microsoft 365 E5 – Enterprise-grade productivity, security, identity and compliance
- Microsoft 365 Copilot – AI embedded directly into everyday work experiences
- Microsoft Entra Suite – Advanced identity, access and governance controls
- Microsoft Agent 365 – Centralised visibility and control for AI agents
What does it mean for you?
Most organisations have taken sensible first steps with AI.
Small, contained use cases:
- summarising emails
- drafting documents
- speeding research
The shift that E7 is trying to enable is from “AI that helps” to “AI that does.”
That means agents.
What’s an agent?
Agents don’t just suggest, they act. They trigger workflows, move information between systems, and make decisions. This is powerful and it changes your risk profile. When work is delegated to agents, the questions you and your leadership team need to answer stop being technical and start being operational.
- Are we comfortable letting AI act on our behalf?
- Do we know where our boundaries are?
- Can we prove control when someone asks?
The challenge is to get confidence at scale, the ability to let AI run without constantly worrying what it might break. This has a few important implications which have fed into Microsoft’s positioning of Agent 365 as the agent “control plane.”
If you’re exploring how AI agents could operate safely inside your Microsoft environment, CWSI’s consulting team can help define governance, identity controls and operating models for AI adoption. Find out more here.
Why AI Agents Change the Enterprise Risk Model
Data exposure becomes a leadership issue
Agents don’t just read data. They combine it, reason over it, and act on it. Without strong guardrails:
- sensitive information can be accessed out of context
- outputs can breach internal policy or regulation
- data can be reused in ways no one intended
The organisations that scale AI safely are not the ones that trust agents blindly. They’re the ones that treat data boundaries as enablers of speed, not blockers, so agents can move quickly inside well defined limits.
Visibility becomes more important than prevention
Traditional security assumes humans are the primary actors. Agentic AI breaks that assumption. The first question leaders ask shouldn’t be “can we stop this?” it should be “do we even know this is happening?”
In practice, the organisations that scale agentic AI tend to prioritise runtime visibility.
- What are they doing right now?
- Are they behaving as expected?
- Can we pause or stop them if needed?
Good visibility doesn’t mean things don’t go wrong, but it means the organisation can see, respond, and improve quickly — especially when backed by 24/7 monitoring and response capabilities.
Accountability doesn’t disappear, it gets blurred
When an agent approves something, escalates an issue, or updates a record, who is accountable? The human who asked for it? The team that built it? The business function that benefits from it?
In regulated or risk sensitive environments, “the AI did it” is not an acceptable answer. Frontier organisations design accountability up front: clear ownership, clear authority, and the ability to intervene when needed. A helpful test is simple: can you explain who owns an agent, what it’s allowed to do, and how you would intervene if something isn’t right?
Compliance can’t lag behind innovation
AI adoption is moving faster than policy cycles, audit frameworks, and risk committees are used to. That creates tension:
- business teams want momentum
- compliance teams want assurance
When this doesn’t land well, it’s usually because governance is treated as a phase gate. The organisations that make progress tend to treat governance as part of the operating model designed once, then applied consistently as AI usage grows. This is the difference between pilots that stall and AI that delivers value.
Where should you start?
Strip away the marketing language, and Frontier Firms share a few observable traits – often aligned to a broader security and compliance solutions strategy across the organisation.
- They can list their agents, their owners, and their purpose
- Agents operate with least privilege access, just like people
- Actions and outputs are auditable by default
- Humans stay in the loop for exceptions and high-risk decisions
- Security and compliance teams have real-time visibility, not after-the-fact reports
None of this is glamorous. All of it is what makes adoption sustainable. Frontier transformation isn’t about being first. It’s about being deliberate.
What can you do – quick wins?
If you’re exploring what E7 means for your organisation, start with two lists: the decisions you’re happy to delegate and the data you’re not prepared to expose. Then pressure test your operating model against them; identity, data boundaries, visibility, and response. The technology will move quickly either way; confidence comes from what you can evidence.
Here are five practical recommendations that tend to create early momentum without introducing unnecessary friction:
Quick win 1:
Be explicit about who can create and run agents (and where). Start with a small creator group and expand deliberately — it helps avoid duplicated effort and makes ownership clearer from day one.
Quick win 2:
Treat agents like identities — least privilege, scoped access, and a clear lifecycle (owner, review cadence, expiry/retirement).
Quick win 3:
Fix obvious data oversharing. If broad access is the default today, E7 will amplify it; tightening permissions is one of the fastest safety improvements you can make.
Quick win 4:
Make activity observable – log it, retain it, and make “what happened?” answerable without a forensic exercise.
Quick win 5:
Keep humans in the loop for high-impact actions (payments, external sharing, HR decisions, customer-impacting comms) until you have evidence and confidence.
Get those foundations right and E7 becomes less about managing risk, and more about running AI safely at scale.
A final thought
E7 won’t make you a Frontier Firm on its own. But it does make one thing unavoidable: AI is no longer just helping people work. It’s starting to do the work. The organisations that succeed in this next phase won’t be the ones with the smartest agents. They’ll be the ones that took the time to design trust, accountability, and control into how those agents operate. That’s not a technology problem. It’s a leadership one.
About the Author
Mark Mitchell – CTO, CWSI
Mark joined CWSI in 2018 to build the Microsoft practice and now serves as Group CTO, leading teams delivering security services across identity, compliance and wider cybersecurity controls.
Exploring Microsoft 365 E7?
If you’re looking at how AI agents, Copilot and Microsoft 365 E7 might change the way your organisation operates, the first step is understanding whether you’re ready.
CWSI’s AI Readiness Assessment helps you:
- assess your current Microsoft environment
- identify risks around data access and governance
- define safe boundaries for AI and Copilot usage
- build a clear, secure path to adoption
