Microsoft CEO: AI agents will end SaaS

Plus: OpenAI o1 (finally) goes agentic, Salesforce announces Agentforce 2.0, and more

Author

Michael Cunningham
December 19, 2024

In partnership with

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

Life is tough these days if you’re a person who can instantly write perfectly coded React apps. So many cases of mistaken identity.

In today’s issue…

  • API access for o1 arrives with agentic upgrades

  • Salesforce’s Agentforce 2.0 is coming

  • IBM’s new open-source model for enterprise agents

  • Satya Nadella on the future of agents and SaaS

…and more

📰 NEWS

Source: OpenAI

As part of its 12 Days of OpenAI, the company rolled out API access for its reasoning model o1 and—perhaps in response to critiques from a certain AI agent newsletter—enhanced its agentic capabilities with function calling, structured outputs, vision capabilities, and more.

The new version of the CRM giant’s enterprise agent platform, arriving in February, will bring a host of new features, including a natural language-driven Agent Builder and integration with Slack. Salesforce is also planning to hire over 2,000 people to sell the technology to business customers.

Automated Design of Agentic Systems—previously covered in Building AI Agents—which made waves in August as an important step for self-building AI, received the Outstanding Paper Award at a NeurIPS workshop on agentic AI.

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

🤝 WITH HUBSPOT

Want to get the most out of ChatGPT?

ChatGPT is a superpower if you know how to use it correctly.

Discover how HubSpot's guide to AI can elevate both your productivity and creativity to get more things done.

Learn to automate tasks, enhance decision-making, and foster innovation with the power of AI.

🛠️ USEFUL STUFF

Source: Wikipedia

Granite 3.1 8B Instruct is the latest in IBM’s Granite series of open-source LLMs specifically focused on agentic enterprise workflows, which outperforms other models in its weight class on a variety of benchmarks.

No-code enterprise AI agent startup Wrkdn provides a list of 20 critical business challenges that their technology—or any agent framework—can address.

CodeSandbox SDK solves the issue of agents potentially writing and running dangerous code by allowing users to spin up isolated virtual machines to run them in.

💡 ANALYSIS

Microsoft CEO Satya Nadella | Source: The New York Times

A clip from an interview with the Microsoft CEO, in which he claims that the rise of agentic AI will spell the end of the current software-as-a-service (SaaS) ecosystem

This piece, based on a survey by Trai.ai, points to the excitement professionals feel about agents, with one in four believing most of their companies’ core businesses will run on them—but also the trepidation created by legacy tech stacks unable to support agentic automation.

Wayne Hamadi of AgentOps narrates the origin of the agent field with the creation of AutoGPT in March 2023, and gives an in-depth example of the kind of business use-case agents are perfectly suited to.

The author of this article describes the shift that has occurred in the agent field from reinforcement learning towards LLMs, and the broad range of complexity they cover, from simple for-loops to sprawling multi-agent systems.

The WEF and consulting firm Capgemini provide a deep-dive on agents: what they are, the problems they promise to solve, and new ones they could potentially create.

🧪 RESEARCH

Source: arXiv

With growing interest in using LLMs to automate scientific research, this paper highlights the potential applications of AI science agents, as well as the challenges that must be overcome along the way.

True engineering disciplines, the authors of this paper argue, require the precise specification of each components’ behavior, something that the agent field currently lacks. They review the progress that is being made, and argue for more rigorous specification statements for LLM-based systems.

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