Agents are everywhere (AI),
Claude agents are SO amazing! equally OpenClaw is really cool, both of these “cool” technologies are being used in businesses all over the world right now. Ask the next time you are engaged with one of your clients or they will say to you “we are going to sign up for Claude as our Agentic AI strategy.
This conversation is the one I have heard from almost every partner over the past couple of weeks, so if you are looking for a stock tip buy shares in Anthropic (this is not financial advice!)
However, this adoption is massively scary, or at least should be for any business providing technology services to businesses! because people in most cases don’t think about what happens when they give an LLM access to their data, ALL OF THEIR DATA……. just yesterday I was speaking with an MSP who’s customer Enabled Claude, configured the Graph API to connect to their companies data – everything, every sensitive document, every detail including sensitive information related to all of their customer base (I will not expand on the details as it would be easy to identify) and now were in a panic asking their MSP how to secure the data, it is too late! Any documents, any powerpoint, any agents, anything at all that they create from now on could very well have information within the content that is sensitive and not for sharing.
And now with Agents, all of this sensitive data can be shared wherever and however because agents just “get stuff done” in whatever way they ’think’ is best.
Or OpenClaw, another MSP speaking with me yesterday just helped reduce a $100k per month Token usage many fold down (again not using details for protection of the guilty), these costs are staggering – they accounted for the amount they were spending, but the reduction after an expert had a look and reviewed the workflows just goes to show that expertise is needed!
So I have some advice here, for the generic use case which applies to most Small to mid sized businesses in Australia (and the world), use Microsoft Copilot, and maybe even choose “Claude” as your LLM model (yes you can choose Claude inside Copilot) because it is grounded in your data and respects your permissions by Default, yes you still need a data discovery assessment to understand what might be sensitive but out of the box your data is more useful when you using an LLM for 80% of a business use case.
But I keep hearing from customers that Microsoft Copilot isn’t good enough?
Absolute rubbish(with a qualification) for the majority of corporate use cases because Copilot Cowork is about to be released, there is an early adoptor program you can try and sign up for today called the Frontier program. Cowork is Microsoft’s Agent answer to Openclaw and Claude Agents and it is excellent, runs on Anthropics Claude by default and completely changes the landscape, I have been using it for weeks and it is a game changer in my workflow.
Lastly as a managed service provider, you can not monetise Claude but you can monetise Microsoft Copilot – so the choice is easy, but we have to be on the front foot with our business customers because they aren’t waiting, they are buying licenses and creating great big data security exploits without your advice or support.
Reach out if I can help, here
