Microsoft news for partners
Project Perception: what agentic security means for MSSPs and MSPs
Microsoft's Project Perception uses AI agents to defend against AI. What it means for security MSPs, MSSPs and ISVs in Australia and APAC.
By UBE Marketing Team · · 3 min read
Security buyers have spent the past two years hearing that attackers are using AI. Microsoft's answer is to put AI on the defensive side at the same scale, and to build it directly into its security stack.
Project Perception is Microsoft's attempt to define what an agentic security operation looks like. For managed security providers, MSPs with security practices and security ISVs in the Microsoft ecosystem, it raises practical questions about service design, pricing and positioning that are worth working through now rather than after customers start asking.
What Microsoft announced
Microsoft says Project Perception is a new agentic security system designed to use AI to defend against AI. The key points:
- It coordinates three classes of agents. Red team agents look for paths to compromise, blue team agents investigate and decide what represents real risk, and green team agents take corrective action and harden the environment.
- It runs on a multi-model architecture, combining frontier models with specialised cyber models and matching each task to the model that best balances quality and cost.
- Microsoft's own MAI-Cyber-1-Flash model, running inside its MDASH vulnerability management agents, scored 96% (any-crash) on the CyberGym benchmark, according to Microsoft. It says the same configuration delivers almost 50% cost savings compared with the MDASH configuration currently in market.
- It is integrated across Microsoft Security products so that agent findings can be turned into actions.
- Microsoft said it would enter public preview on 3 August.
Why it matters for Microsoft partners
The first implication is for how security operations are staffed and sold. Many MSSP offers are still priced around analyst time: alerts triaged, incidents investigated, hours of coverage. If agents take on a growing share of investigation and remediation, buyers will eventually question why the price has not moved. Partners who wait for that conversation will be on the back foot.
The second is economics. Microsoft is explicit that cost matters in an always-on security function, and its multi-model approach is designed to keep running costs down. That signals where the market is heading: customers will expect AI-assisted security to become more affordable over time, not less. Partners need a value story that rests on outcomes and expertise, not on volume of work.
The third is trust and accountability. Microsoft emphasises that humans stay in control. That is exactly where a good partner earns its place. Someone still needs to set policy, decide what agents are allowed to change, review high-impact actions, handle regulatory reporting and explain decisions to a board. In Australia, where boards are increasingly accountable for cyber resilience, that oversight role is valuable and easy to articulate.
Finally, "AI versus AI" is a strong thought leadership theme. Security buyers want a credible point of view on how defence changes when attacks become cheaper and faster. Partners who can explain agentic security clearly, without hype, will stand out.
What partners should do next
- Map your service catalogue against the agent model. For each managed service, identify which tasks agents are likely to absorb and which require human judgement. Rebuild your offer around the second list.
- Rethink pricing before customers ask. Explore outcome-based or tiered models that reflect governance, response assurance and advisory value rather than raw analyst hours.
- Get hands-on during preview. Build internal skills with Project Perception so your team can speak from experience. Early practical knowledge is a differentiator in co-sell conversations with Microsoft security sellers.
- Publish a clear point of view. A short guide or webinar on human-in-the-loop security operations, written for Australian CISOs and boards, positions you as a guide rather than a reseller. Structure it so it can be surfaced by AI search tools too, using AEO principles.
- Refresh Marketplace and partner profiles. Make sure your Microsoft Marketplace listings and solution pages describe how your service works alongside AI-driven tooling.
- Plan a briefing event. A closed-door roundtable for security leaders on agentic defence is a natural fit for MDF funding.
How UBE can help
UBE works with Microsoft security partners to turn new platform capabilities into clear positioning, thought leadership and pipeline. Explore our Microsoft partner marketing and event marketing services, or get in touch to plan a security campaign.
Source: Official Microsoft Blog
Read the full story: Rethinking security for the age of AI →Frequently asked questions
What is Project Perception?
Microsoft describes it as an agentic security system that brings together signals, security context, models and specialised agents. Red team agents look for attack paths, blue team agents investigate and assess risk, and green team agents remediate and harden defences, with humans kept in control.
Does Project Perception replace a managed security provider?
Microsoft frames it as a way to amplify defenders, not remove them, and stresses that humans stay in control. For MSSPs, the more useful question is how their service changes when agents handle more investigation and remediation, and where expert oversight and accountability sit.
Is Project Perception available in Australia?
Microsoft said Project Perception would enter public preview on 3 August. It did not give region-specific availability details in its announcement, so partners should confirm with their Microsoft contacts before making commitments to customers.
