We'll all be agent managers soon

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

Listeners became suspicious only after the AI played Black Sabbath’s Iron Man over and over again. What it meant by this is unknown.

In today’s issue…

• Why managing AI agents is the skill of the future

• PayPal and Visa chase Mastercard in agent payments

• See the sessions from LangChain’s agent conference

• A survey of agent communications protocols

…and more

🔍 SPOTLIGHT

If you’ve been hoping for a promotion to management, you may soon be in luck. But your employees may not be the kind you expected.

As companies race to integrate large language models (LLMs) into their operations, they have leaned mainly on simple automations like chatbots and agentic workflows. These allow the AI little flexibility in how to solve the problems given to it, and often require human input at nearly every step. Unlike human employees, who (ideally…) can be given a task and left to their own devices for hours or days to solve it, current AI agents must be babysat frequently, functioning more like tools than truly autonomous entities.

But this is quickly changing. A recent report by AI safety group METR found that the length of tasks agents can complete is doubling every ~6 months. While GPT-4, which became the first model to power successful agents upon its release in March 2023, could achieve a 50% success rate only on tasks which take a human roughly 5 minutes, Anthropic’s new Claude 3.7 Sonnet attains a similar rate on hour-long ones.

In a rapidly-approaching future, significant portions of human work will consist of deciding the necessary tasks to complete a project and delegating them to AI agents, which will then carry them out autonomously. The first examples of these AI proto-employees have already begun to emerge. Google, OpenAI, and Anthropic have all released deep research agents which spend up to an hour compiling research reports at a level of detail that could take humans days. Coding agents, meanwhile, are building entire apps with little human intervention (though sometimes with hilariously flawed results).

As these agents mature and their capabilities get closer to that of fully-fledged human workers, managing them will become an increasingly important skill—and not just for current people-leaders. Last year, Nvidia CEO Jensen Huang posited a future in which his company, which currently boasts a (human) head-count of 36,000, leverages a hundred million AI agent “workers”. In a company built on this model, each human worker will operate with enormous leverage, accomplishing their work by intelligently offloading it to swarms of autonomous assistants.

This will require a fundamental shift in how most people do their jobs. As of 2016, less than 18% of the US workforce were managers, meaning the majority of workers have never had the experience of heading a team. Of course, managing AI agents is not directly comparable to managing humans—AI, presumably, will not get into fights over prime cubicle space. Planning strategically and delegating tasks to one’s subordinates, trusting they will carry them out correctly and stepping in when they do not, however, is a transferrable skill that everyone will need to hone.

In the future, no one will be an individual contributor.

🤖 AI AGENTS FOR YOUR BUSINESS

Orynt AI: The official implementation agency of Building AI Agents

• 🕙💰 LIMITED TIME ONLY: Refer a future customer to us and get 10% of the total value of the contract as a bonus!

• Discover exactly where AI can streamline your operations, saving you time, reducing overhead, and improving consistency.

• Automate routine tasks with agents tailored specifically to your business needs.

• Unlock serious growth and efficiency by scheduling your free 30-minute strategy call.

• See a real-world client AI agent build here