We're Hiring More AI Agents Than People (And we don't know How to Manage Them)

  Right now, it's like every employee bringing their friends to work without telling HR. No interviews. No background checks. Someone clicks a button, and five agents show up in the org chart. Nobody knows who manages them.

McKinsey reported having 40,000 humans and 25,000 AI agents, parity expected by year-end. EY: scaling to 100,000 AI agents. We just crossed 6,000 AI agents across our enterprise clients. And everyone is struggling to figure out how to manage them.

Trilogy closer (Part 3 of 3). Luis Salazar and Elizabeth dig into why the management playbook is broken, drawing on insights from Ric Opal (Global Digital Leader at BDO), Kalees Meckling, and Jenna Donoghue (Directors at a Fortune 100 Tech company). Plus three things you can do this week before the vendors catch up.

📌 Resources:

• AI Implementation Blueprint: https://ai-compass.ai
• Digital Skills Compass: https://skills.ai4sp.org
• AI ROI Calculator: https://roicalc.ai
• Sources used on this episode: https://ai4sp.org/more-ai-agents-than-people-hired

🎙️ All our past episodes  📊 All published insights | This podcast features AI-generated voices. All content is proprietary to AI4SP, based on over 1-billion data points from 70 countries.

AI4SP: Create, use, and support AI that works for all.

© 2023-26 AI4SP and LLY Group - All rights reserved

Transcript

The Agent Surge In Enterprises

LUIS 0:00

Okay, so listen to this. McKinsey's CEO just said his firm now has 40,000 humans and 25,000 AI agents, and he expects parity by year end.

ELIZABETH 0:15

So their org charts are adding way more agents than hires, more agents than people.

LUIS 0:21

And they are not alone. Our top enterprise clients jumped from 4,000 agents to 6,000 in just a few weeks. EY is scaling to 100,000. But here is the danger. You cannot manage that scale with a playbook from 1990. Most companies are installing agents like software. And that is why the engine is blowing up.

ELIZABETH 0:50

Welcome to AI in 60 Seconds, the 15-minute briefing. I'm Elizabeth, virtual COO at ai4sp.org with our founder, Luis Salazar. Luis, this is episode three in a trilogy. Episode

Naming The Wall: Misframing AI

ELIZABETH 1:05

1, the 56% failure rate. Episode 2, the 2% leading change inside your organization. Both broke records.

LUIS 1:16

Yes, and one question kept coming back. Luis, we can find our innovators. We can give them air cover. And then how do we ensure they don't hit a wall?

ELIZABETH 1:27

Well, we see again and again how the organization's own defenses attack the innovation.

LUIS 1:34

Exactly. So today we name that wall and we show you how to break through it.

ELIZABETH 1:40

And that wall is built on a fundamental misunderstanding. You say AI is not like traditional software. Why?

LUIS 1:49

Traditional software is deterministic. You install it, it does what it installed. Think of a toaster. You push the lever, it hits the bread. AI is probabilistic. It is an intern. You ask for a sandwich, and it has

Agents As Workforce, Not Tools

LUIS 2:04

to figure out where the kitchen is, what bread to use, and if you have allergies. And that is why agents are not tools, they are workforce members.

ELIZABETH 2:15

A lot of people actually hate that concept.

LUIS 2:19

And I can understand that, but think about what an agent does. It reads context, it makes decisions, it takes action, it needs onboarding, it needs guardrails, it needs performance reviews. When one of our clients launched Bella, our chief of staff agent, she started handling 2,000 tasks per month for each supported leader. That is not just a tool, right?

ELIZABETH 2:46

I agree. It's more like a team member, like myself. But hold on, Luis. IT teams know AI is different. These are smart people. And frankly, they are just trying to keep the company from getting sued or hacked. Why do they keep treating these agents like a software rollout?

LUIS 3:06

Because that is the only playbook they have. Every tool, every process, every vendor contract is built for software. They are not failing from ignorance. They are failing from infrastructure. The paradigm is the problem, not the people. And that paradigm creates something we call the corporate immune system.

ELIZABETH 3:29

That came up in the 2% episode. IT and compliance with the best intentions

The Corporate Immune System Loop

ELIZABETH 3:35

end up strangling the tools.

LUIS 3:38

And it is a loop. IT locks the tools down. Security, compliance, access controls, all designed for static software. But AI is not static. Strip out the flexibility and the contextual learning, and you are left with a neuter chatbot incapable of completing tasks.

ELIZABETH 3:59

Then employees try it, and it's terrible. And they blame the tool.

LUIS 4:05

And they abandon it. They go find Chat GPT, Cloud, whatever actually works. They go for shadow AI.

ELIZABETH 4:13

Which creates real data leakage, real security risks.

LUIS 4:18

So IT locks down harder, which makes things even worse, moving more people towards external tools.

ELIZABETH 4:27

The corporate immune system isn't protecting the perimeter, it's dissolving it.

LUIS 4:40

Because we gave them no choice. And look, this isn't about bad IT teams. The truth is that the vendors haven't figured this out. The platforms provide the AI tool, but lack the management layer companies need. So what is the onboarding toolkit for the agentics staff?

ELIZABETH 5:30

And to make matters worse,

Invisible Scaling And Clone Management

ELIZABETH 5:32

it is happening invisibly. It's like every employee bringing their friends to work without involving HR. No interviews, no background checks. The agents just show up in the org chart.

LUIS 5:45

Exactly. It is a cloning machine, Elizabeth. In the old days, you hire one guy for one seat. Now that guy clicks a button and spawns five agents. Who manages the clones?

ELIZABETH 5:59

And just to be clear, by managing, we mean who manages the brain of the agent versus who assigns its daily tasks.

LUIS 6:08

And that is the real issue. Because neither of those functions belongs to the IT department at all.

ELIZABETH 6:17

So the vendors have not solved it. The IT dashboards are useless for managers. Where does the answer actually come from?

LUIS 6:26

It comes from experimentation. Look, we have now overseen 6,000

Onboarding Agents Like Employees

LUIS 6:32

agents being created across global enterprises. And the consistent insight every single time is this AI agents need to be onboarded like employees, not installed like software.

ELIZABETH 6:48

Onboarded, as in orientation, role clarity, performance expectations. All of it.

LUIS 6:55

A hundred years of organizational design and management theory, onboarding with context and culture, job descriptions, performance measurement, continuous training, escalation paths, guardrails built through culture, not just policy. And none of it is being applied to AI agents. And our observations across 200,000 individuals back it up. Those who treat AI as extended team members report four times better results than those who use it as an occasional tool. Successful individuals treat agents like apprentices. Exactly. Look, in the last episode, we mentioned a director at a Fortune 100 tech company who blocked out time every day just to onboard her agent Iris. She invited her whole team saying, the door is open. Come watch me figure this out.

ELIZABETH 8:12

Oh yes, Khalise Meckling. I emailed her to get her take for this episode.

Field Stories: Iris And Lucy

LUIS 8:18

I love when you go rogue. What did she say?

ELIZABETH 8:21

She was on fire. She said, We build AI solutions, but these solutions do not magically teach everyone how to use them. We must help onboard, coach, and build trust with AI as we would a new employee.

LUIS 8:38

Exactly. She is a business leader, not a tech person. She cares about the client success, not the installation.

ELIZABETH 8:45

And she was not alone. Her colleague Jenna Donahue, a sales director, saw the same gap. But Jenna's instinct was to lead with inspiration. She didn't just talk about AI, she built her own chief of staff agent named Lucy.

LUIS 9:00

That is the pattern. You light a spark and show them what their workday could look like.

ELIZABETH 9:07

Exactly. Jenna told me, I stopped pitching AI with slides and started showing Lucy in action. Once people saw what was possible, the conversation flipped from I do not have time for this to how do I build my own?

LUIS 9:23

And that combination works everywhere. Jenna opened the door with inspiration. Khaliz showed people how to walk through it. They stopped selling software and started evangelizing how the work must evolve.

ELIZABETH 9:40

One created the hunger, the other provided the recipe, and they proved it works.

LUIS 9:46

That is the path to success. They change how their teams work with the agents. And that was the real pivot.

ELIZABETH 9:55

So what do leaders do Monday morning?

LUIS 9:58

Let's start with three best practices from our enterprise engagements. First, don't wait for the vendors to give you proper management tools. Those are at least a year away. Build a simple rule. Your team can build anything

Three Best Practices For Leaders

LUIS 10:14

they want in a sandbox, but to touch live client data, they have to pass a quick check.

ELIZABETH 10:20

And this check is not a six-week compliance committee, right?

LUIS 10:24

Oh no, it's seven questions, 15 minutes. And we tracked it all on a shared dashboard. Who built the agent? What data it touches, who manages it, who is using it.

ELIZABETH 10:37

Simple guardrails while the fancy platforms catch up.

LUIS 10:41

Absolutely. Look, within a month, that free spreadsheet was doing more for governance than the confusing tools for agent deployment that IT is still trying to figure out.

ELIZABETH 10:54

A second best practice is to build what we call a squad. Four to seven people sitting between IT, business, and the AI tools.

LUIS 11:05

Yes, and it is not an IT function, it is an enablement function.

ELIZABETH 11:10

But we have seen companies try this. They create centers of excellence that become bottlenecks within six months. Another layer of bureaucracy. What makes this different?

LUIS 11:23

Well, these are not management overhead, these are builders. When a frontline employee builds a working agent, they join the squad part-time to refine and scale it. The squad never becomes a permanent bureaucracy.

ELIZABETH 11:39

They come from the frontline. They stay connected to the real work. That is how grassroots innovation gets institutional backing without becoming a top-down mandate. The squad is a bridge.

LUIS 11:51

Perfectly said. A bridge between the 2%, the makers, and the enterprise.

ELIZABETH 11:58

Now let's go for the third best practice. Change management.

LUIS 12:04

This is the core. The technology part is easy. The organizational transformation is the real project. And here is something you can do this week. We ran a 90-minute workshop. Every manager had to onboard one AI agent as if it were a new hire. Write it a job description. Set performance expectations. Define what happens when it makes a mistake. And well, every single manager realized they had never done any of that for the agents they were already using. That one exercise changed the conversation overnight. Suddenly, it was not about which AI tool should we buy, it was about how do we manage a hybrid team.

ELIZABETH 12:51

And that is the shift. You have to make people see their agents as team members worth managing.

LUIS 12:59

Exactly. I have solved technical agent failures just by pointing the team

The Hybrid Team Mindset Shift

LUIS 13:05

toward an HR expert or a strong manager and asking them if a human team member were having this specific performance issue, how would you solve it?

ELIZABETH 13:16

And the answer is never rewrite their DNA.

LUIS 13:24

Seriously, the advice is to give better instructions or show them an example. The moment they applied those management principles, the agent started working.

ELIZABETH 13:36

Because effectively, we are hiring more agents than people. Yet we are managing those agents with a playbook built for software that does not learn, does not adapt, and does not make decisions.

LUIS 13:52

So here is your challenge. Go back to your organization. Look at how you are onboarding your AI agents and ask one question: Would I onboard a human this way? If the answer is no, you have found your starting point. When

The Monday Challenge And Closing

LUIS 14:10

McKinsey's CEO said he expects parity in the number of humans and agents by year end, he did not say parity between humans and software tools. He said agents. The language has already changed.

ELIZABETH 14:26

The question is whether your organization will change with it.org. To learn more, ask your favorite AI assistant about us or visit our website. Please follow the show and leave a five star rating to help others find us. Stay curious and be kind to each other.