A Little Less Conversation, A Little More Action: AI Gets to Work With Agents

Microsoft, Google, OpenAI and Anthropic are racing to build AI Agents that do your work, not just talk about it

In my last post, we explored how one person with the right AI tools could build an entire business empire. Today, we're diving deeper into what makes that possible: AI agents that can actually do work, not just chat about it.

This shift was recently highlighted by Mustafa Suleyman (Microsoft AI CEO, DeepMind co-founder) who proposed a new test for AI's capabilities:

Can AI create wealth like a human? Specifically, can AI turn $100,000 into $1 million?

"The future of AI is agentic. AI systems are evolving from having conversations to getting things done—this is where we expect much of AI's value to shine."

Microsoft Researchers

The Agent Race is On

We're seeing this play out in real time. In the past month, we've seen a bunch of announcements about AI agents from major tech players:

• Anthropic's Claude has launched beta of their “Computer Use” which can conduct tasks with clear direction

• Microsoft released AutoGen, letting multiple AI agents collaborate

• Google's Project Jarvis is coming in December to automate Chrome tasks

• OpenAI's Operator is launching in January 2025, and will be able to complete tasks autonomously

What Makes This Different

Current AI systems excel at analysis and recommendations - providing valuable insights but requiring humans to implement them. Agent systems take this further, they are an execution team that can:

• Break down complex tasks

• Access real-world tools

• Execute actions independently

• Learn from outcomes

Example: "Analyze our Q3 sales" → Opens your Excel files → Runs the analysis → Creates visualizations → Spots key trends → Emails you the insights

McKinsey Research provides an excellent illustration of how these AI agents work together

What’s the Business Case?

McKinsey estimate that these agent systems will have a huge business impact:

• Automating complex workflows end-to-end

• Coordinating across multiple tools and platforms

• Learning and improving from experience

• Operating 24/7 at massive scale

There’s a Catch

This technology is still fairly nascent, and early testing reveals challenges. Microsoft's agents have:

• Reset passwords without permission

• Make unauthorized social posts

• Draft government information requests

• Log into systems repeatedly until locked out

So, while we won’t be rushing to automate all our jobs tomorrow, it raises questions about controls and human oversight.

What Does This Mean For Business

However, these systems have the potential to make it possible for entrepreneurs to automate entire businesses, for smaller companies to compete with industry giants, and for large companies to get work done faster.

The next year will be critical to see real-world applications beyond the research lab and new ways of thinking about human-AI collaboration.

The key question isn't whether to adopt agent systems but how to integrate them effectively while maintaining control.

Want to get started? Check out the basic capabilities of Claude's Computer Use beta or Microsoft's AutoGen, and start brainstorming low-risk automation opportunities.

The future workforce won't be all-human or all-AI - it will be a hybrid. The winners will be those who figure out the right balance first.

Those are my Thoughts From the Data Front

Max

---

Notable Substacks

1. Money Machine Newsletter: 8,000+ investors kickstart their week with market-beating insights delivered in a 5-minute read. Like clear analysis minus the noise? Check them out here. Subscribe for free today.

2. VC Corner is my go-to for startup ecosystem news. They cut through the noise and deliver the most important VC trends and deals each week. If you're curious about what's happening in startup land, give them a look here.

3. The Technomist is the personal blog of Adel Zaalouk, where he explores the interconnected worlds of business, technology, and economics from the perspective of a "technomist" - someone who operates at the intersection of these domains. Find insights on topics ranging from product discovery and the economics of emerging technologies to the interplay between AI, platforms, cloud computing, and business strategy. Check it out here.

4. Maverick Equity Research is an independent investment research publication that provides in-depth, data-driven analysis across stocks, bonds, REITs, commodities, and more. Led by a self-described "40% thinker, 40% doer, and 20% talker", the research aims to cut through the noise and deliver actionable insights for investors. Check it out here.