Why Every AI Agent Needs an Owner, Budget, and Off Switch

What happens when someone resigns? Their laptops get wiped, badges are off, but their agents keep running.

They're still spending money, still touching data, still reaching out, doing things.

Who's their manager now? Meaning who is the manager of these left agents that were created by a former employee? This is the invisible headcount that I wanted to talk about today, and this is becoming a big, bigger and bigger topic because now the huge companies out there have decided to do something about it.

Microsoft built a home for agents that employees create.

Each agent gets its own identity, their own rules for what it's allowed to do, what it's not allowed to do, and monitoring.

Basically, the IT controls you put on an employee's laptop, you apply to agents.

How interesting is that? Google added a feature in Gemini Enterprise so that when an employee leaves, an admin can hand that person's agent to someone else.

Anthropic launched a marketplace where companies can buy agents and software from vendors using money they've already committed to spending.

So these guys all know what's happening, right? New Relic published something just the other day showing that one in four AI agents run with no monitoring.

Those of us that have managed people for many years, it's probably still the case about employees as well, but this is really interesting.

So a lot of, lot of work and money is being spent on this, this topic that honestly not a lot of us are thinking about.

In fact, I'm almost sure nine out of ten of us aren't, aren't doing anything about these, right? This is the swap nobody's managing, you know? And let's talk about what's hap- why this is even relevant, guys.

Tech cuts this year are approximately one hundred and sixty-five thousand.

Last year, same time, about one hundred and eight.

So it's actually increasing even though we're hearing about it less and less.

Disney just did one recently.

September cut, cuts are up seventy-seven percent over August.

Think about that.

All of a sudden everybody decided, "Hey, September's a great month to do layoffs." Oracle is cutting people to pay for AI chips and data centers.

You know the results? Only thirty-seven percent are seeing any profit impact from AI.

It's flat year over year, so no growth, which is surprising, right, 'cause AI is getting better and better and better.

We're learning how to manage it better and better.

We're starting to create regulations around it, but apparently not enough.

We cut the part of our budgets that had governance and grew the one that didn't.

Isn't that ironic? From my seat, who-- you know, when someone's asking us to hire a CIO or a CTO or CSO, you know, the first question I ask is, you know, "How many agents are running in your organization and who owns each one?" And typically candidates can't answer this question because it's not something that they've thought about, but maybe you should start asking these questions.