Microsoft expands Copilot agents

Plus: Google's new market for AI agent creators, million-agent social simulations, and more

Author

Michael Cunningham
November 21, 2024

Welcome back to Building AI Agents, your biweekly guide to everything new in the AI agent field!

Some AI apocalypse scenarios are scarier than Terminator

In today’s issue…

  • Microsoft’s new agent feature bonanza

  • Google’s marketplace for agent builders

  • A universal protocol for agent communications

  • Simulating social networks with a million agents

…and more

📰 NEWS

Source: Microsoft

At its Ignite 365 conference, the company announced major expansions of its agentic AI service Copilot, including agents integrated into Teams, SharePoint, and more, third-party agents from providers like ServiceNow and Workday, Copilot Actions to allow Microsoft 365 users to automate everyday tasks, and a control system for IT teams to manage agents company-wide. Some of these are immediately available to users, while others will be rolled out in the coming months.

To help businesses automate processes at an enterprise level, Microsoft also released Azure AI Agent Service, which allows developers to build agents with broad access to company data and applications and deploy them in Azure.

Google Cloud released a marketplace called AI Agent Space, where agent builders can receive guidance and support from the company’s engineers on building and deploying agents, as well as assistance in marketing the resulting products.

The CRO giant launched a suite of tools to help Agentforce engineers safely test their creations and monitor them once deployed, which will be generally available in early December.

Microsoft AI CEO Mustafa Suleyman hinted that the company’s researchers had developed prototypes with nearly infinite memory, just days before Google’s Gemini rolled out a memory feature similar to that used by ChatGPT, indicating that major AI providers see memory as the next frontier in capabilities.

If you find Building AI Agents valuable, forward this email to a friend or colleague!

Refer one person and we’ll send you a set of weekly bonus links to additional AI agent news for a month! See here for a sample

🤝 WITH AI TOOL REPORT

There’s a reason 400,000 professionals read this daily.

Join The AI Report, trusted by 400,000+ professionals at Google, Microsoft, and OpenAI. Get daily insights, tools, and strategies to master practical AI skills that drive results.

🛠️ USEFUL STUFF

Source: LangChain

In an effort to standardize communication between agents built in different frameworks, LangChain has released a common protocol intended to be compatible with both its technology and its competitors’.

Scrapybara is a YC-backed startup which offers virtual desktop environments, accessible via an API, for AI agents to automate workflows without requiring users to give them control over their physical computers.

Another new YC company, Abundant, reverses the usual agents-assisting-humans paradigm by providing an API which AI agents can call for remote human assistance when they hit difficult edge cases.

The company released AI Shell, a tool which adds agentic assistance to the Windows terminal to facilitate coding, debugging, and automation.

AWS, LlamaIndex, and Redis are teaming up to host a webinar on agent building using the three companies’ tech.

💡 ANALYSIS

Source: Substack

A long-form overview of all things agents, including architecture, memory, reasoning, and planning, frameworks, evaluations and benchmarking, applications, and more.

This piece argues that Microsoft has built a sizable lead in the agent space over competitors such as Salesforce, powered by its view of agents as part of an interconnected system across the whole enterprise, rather than as isolated individual contributors.

Microsoft’s Stephen Kaufman describes the processes of working with a customer to architect an agentic system that generates Python code to solve a business use case.

A piece from Microsoft providing an intro to agents and describing what the company’s newly released features—see above—mean for the world of work, including for non-developers.

🧪 RESEARCH

Source: ArXiv

The authors of this paper simulate the dynamics of online social networks, including misinformation spread and group polarization using up to 1 million AI agents.

Thanks for reading! Until next time, keep learning and building!

What did you think of today's issue?

Login or Subscribe to participate in polls.

If you have any specific feedback, just reply to this email—we’d love to hear from you

Follow us on X (Twitter), LinkedIn, and Instagram