Microsoft took OpenClaw, wrapped it in enterprise security, and called it Scout

[Alfonso Maruccia]

In a nutshell: OpenClaw introduced a few innovative options for agentic AI solutions but was also plagued by some baffling security and reliability issues. Now, Microsoft Scout is allegedly ready to address those problems for Microsoft 365 users – if you're comfortable managing your digital life through an always-on AI agent, at least.

A few days after Google introduced its OpenClaw-based AI agent with Spark, Microsoft is now doing the same with Scout. The new AI tool is the first personal agent designed for the Microsoft 365 ecosystem, the Redmond-based company said. Scout is part of a new category of agents called Autopilots and can allegedly perform the tasks of a human assistant without needing to sleep, rest, or log off.

The Scout agent is deeply integrated into the Microsoft 365 environment, meaning it can access productivity apps, cloud storage, Teams, and more. It can analyze data contained in chats, calendar appointments, emails, and contacts, organizing a user's day or highlighting the most important events ahead of a new workday.

Microsoft confirms that Scout is based on OpenClaw, the same AI assistant that took the AI market by storm earlier this year. Unlike OpenClaw, however, the Microsoft agent has reportedly been built on enterprise-grade security principles and controls. Each Autopilot agent has its own Entra identity and can be configured to access only specific data or services within the Microsoft 365 cloud environment.

Redmond said that Scout can help employees and professionals organize and coordinate their workday, offering a more proactive approach to meeting scheduling. It can also generate relevant content or detect potential issues in decision-making before they affect workflow. In addition, the AI agent is expected to become more useful over time thanks to the contextual awareness provided by the Work IQ service.

Scout is available to enterprise organizations and to individual users enrolled in the Frontier program, although it is currently limited to a desktop app. Microsoft has also decided to contribute back to the open-source community by adding its own "policy conformance" layer to the OpenClaw project. This way, organizations that have already deployed the FOSS AI technology should benefit from a more secure agentic experience.

According to Scout Corporate Vice President Omar Shahine, the new desktop app is already popular among Microsoft developers. Around 3,000 Microsoft employees are currently using the Autopilot tool to manage work-related tasks, including attending meetings, handling paperwork, booking travel, and more. OpenClaw is an extremely fast-moving open-source project, Shahine said, which is why Microsoft decided to take the existing technology and build sandbox-style protections around it to prevent unpredictable behavior.

Permalink to story:

Microsoft took OpenClaw, wrapped it in enterprise security, and called it Scout

 

[ScottSoapbox]

It's funny that the PC effectively ended lower level secretaries (e.g. replacing it with email and the like), and now AI is basically being sold as a digital secretary. Which if it worked well and was cheaper than humans would be great. But so far it works mid and is quickly becoming more expense than just hiring an assistant (now that tokens aren't being as subsidized).

 

[opckieran]

I yearn for the day paperwork can finally be handed to an agent...

 

[VariableSpike]

It's funny that the PC effectively ended lower level secretaries (e.g. replacing it with email and the like), and now AI is basically being sold as a digital secretary. Which if it worked well and was cheaper than humans would be great. But so far it works mid and is quickly becoming more expense than just hiring an assistant (now that tokens aren't being as subsidized).

Exactly, what is the point in an AI "agent" that supposedly reads reports, summarises them and all when you have to check yourself through the report to make sure its not hallucinated nonsense or misinterpreted something or checj the emails it wrote you ao it doesn't send out something heinous or factually incorrect, so so pointless....

 

[OortCloud]

They will try to get you locked in to these (inherently unreliable and untrustworthy non-AI) LLM assistant ecosystems. They will scrub all the data you are constantly sending back to LLM-HQ, sell it to anybody who comes knocking and market you relentlessly. And as a final coup-de-grace, once you're reliant on them, they will start ratcheting the prices to try to recoup the massive unpaid investment and horrific energy usage these models require.

 

[Knot Schure]

They will try to get you locked in to these (inherently unreliable and untrustworthy non-AI) LLM assistant ecosystems. They will scrub all the data you are constantly sending back to LLM-HQ, sell it to anybody who comes knocking and market you relentlessly. And as a final coup-de-grace, once you're reliant on them, they will start ratcheting the prices to try to recoup the massive unpaid investment and horrific energy usage these models require.

Trust being the biggest issue here.

I've PAID for Outlook, but I use Thunderbird now, as I simply don't trust Micro$oft anymore.

My main machine now dual-boots Linux.

The relationship is irrepariable. They do too much the customer doesn't ask for - nor need, nor want. I actually perfer Windows, but I wish the Penguin-Huggers all the luck in the world.

 

[Vixiy]

So in other words scout is openclaw but with Microslop introduced security holes.

 

[OortCloud]

So in other words scout is openclaw but with Microslop introduced security holes.

In true Micoshambles fashion, I expect it takes screenshots of your desktop once a second, saves them in a easily accessible location and sends them to HQ for AI processing...
Only a company with absolutely zero idea/care about security would think this is in any way ok.