AI agents are here. Do we still need this many managers?


Party of five

Last year, I looked at the growth signals from 180 different startups, and I found that companies with teams of about 5 employees per manager perform best. If you're going to err on one side or the other, it is better to have bigger teams than smaller ones.

Today, most companies have too many managers, especially in technology companies. But how will this change as agentic AI enters the workforce? How many managers are enough, and how many are too many?

What is agentic AI?

Agentic AI is designed to act autonomously to complete tasks without continuous human oversight. It is typically focused on completing a domain-specific task. For instance, agentic AI might independently respond to customer help queries or order product inventory based on recent buyer demand.

Last year, I shared that more than 75% of the AI startups I saw were explicitly pitching job replacement in their fundraising decks. The majority of these were building some kind of agentic AI.

Not all workers are employees

The agentic AI transition is not only happening, but it is happening quite rapidly for certain roles. In many organizations, humans will work alongside AI agents for tasks as soon as this year. But AI agents stop short of being traditional employees.

The traditional employer-employee relationship is like this: In exchange for completing certain tasks for their employer on a defined schedule, employees receive financial compensation, health benefits and time off, and legal worker protections that are grounded in a shared social understanding of fairness and human rights.

The promise of agentic AI is that it can complete its tasks without receiving any of these things (notwithstanding the fees that companies will pay to AI vendors for the use of their products). In other words, AI agents may be workers, but they aren't traditional employees in any other way.

This has me thinking about what kind of management AI agent workers need, and what this means for traditional people managers.

Low oversight doesn't mean no oversight

Look, no one is deploying AI agents without some oversight. Yes, AI agents will complete discrete, individual tasks autonomously. However, agent performance will be reviewed with the same metrics that managers should already be using to assess employee performance today. For instance:

  • AI agents will answer customer help questions autonomously. A customer support leader will monitor NPS on these interactions.
  • AI agents will make automatic inventory adjustments. A supervising buyer will review the agent's inventory forecast accuracy.
  • AI agents will autonomously run payroll tasks. An accountant will review tax statements before filing to make sure they are accurate.

In all of these cases, the AI agent functions autonomously to complete individual tasks. The supervising human then uses metrics to assess agent performance. Based on that performance, the supervising human can refine its AI agents by supplementing with new training data or prompt guidance.

An AI supervisor? Is that like a people manager?

In our prior data, we saw that human teams work best when managers have teams of about five people. Five is big enough to provide a range of skills and perspectives and build a team culture around, but small enough that the manager can engage meaningfully with every individual.

To my knowledge, no one has collected data yet about how AI agent workers will change this dynamic. I'm working with teams to collect data about this and will report back later this year.

What I can tell you is this:

1/ Each AI task will be associated with a performance metric. AI may perform the task, but a human supervisor will be accountable for the performance metric. They will use and tune AI agents as necessary to hit the metric.

2/ Great managers already use clear metrics to assess team and individual performance. However, most managers are not great and do not have these skills today. Managers who are already good at this are best positioned to make the jump to overseeing AI agents.

3/ As AI agents come online, it's hard to see organizations needing as many people managers. Not because human teams will get bigger; I don't think this will happen. But because there will be fewer humans to manage overall.

The bottom line: A lot of people are trying to sell you courses and workshops on how to be ready for AI in your career. Most of this material is trash.

In my view, the most useful thing you can do to prepare yourself is to get more metrics-driven in your area of expertise.

Thanks for reading!

Kieran


I wrote down three of my tried-and-true prompts for team meetings that drive team connection and performance. No big budget required!

If you liked this article, help me keep nerd processor broadly available.

My latest data stories | Tell your own Viral Data Stories | nerdprocessor.com

kieran@nerdprocessor.com
Unsubscribe · Preferences

nerd processor

Every week, I write a deep dive into some aspect of AI, startups, and teams. Enter your email address to subscribe below!

Read more from nerd processor

The most magical place on earth I had 90% of a nerd processor draft finished last week, on a piece called How to break your team spirit in three acts. But then I took our 15yo and her best friend to Disneyland for a few days over spring break, and as I sat down to finish the newsletter, my heart honestly wasn't in it anymore. Maybe another time. But right now as I write this, I'm sitting by the pool with a mocktail while my kid and her friend are going on roller coasters in the park, and I'm...

Nine-year itch Since I graduated from college, I have changed careers in a meaningful way every nine years. I was an academic until I craved something more applied. I was a product and engineering leader in BigTech until a reorg caused me to think hard about what I wanted. I was a startup CEO until I stepped down for a break last year. I have generally enjoyed my jobs and I didn't plan any of these moves. In all three cases, I didn't know what I wanted when I made the change. At the end of...

The image generation Two weeks ago, I asked ChatGPT's new image generator to use everything it knows about me to generate an image of me at work. It gets a lot of details right! And ONE ENORMOUS FUNDAMENTAL PART extremely wrong. ChatGPT's image of me at work After I wrote about this on LinkedIn, over 1,500 people sent me their own images. Some people used something other than ChatGPT's new image generator for their images, and others didn't send me the system's explanation of its assumptions....