KLARNA HIRED 700 AI AGENTS. THEN IT CALLED THE HUMANS BACK.

Andrea Navas

3 hours ago

4 min read

Quote graphic on a dark brown background reading “You can’t hand work to an AI employee if you can’t clearly describe the work.” Andrea Navas, Ignyte Studio.

If your feed looks anything like mine right now, you've seen it. "I run my business with 50 AI employees." "Here's how my 100 AI agents handle everything while I sleep." Screenshots of dashboards full of little bots with names and job titles.

And honestly, I get the appeal. I really do. When you're a founder carrying the whole business in your head, the idea of handing a big chunk of it to a team that never sleeps and never needs to be trained sounds like exactly what you've been waiting for.

But I keep having the same reaction every time one of these posts comes up. This feels familiar. We've seen this movie before, with the project management app that was going to fix everything, the automation platform, the all-in-one CRM. Founders get excited, set it up fast, and six months later they have one more thing to manage.

Here is the part everyone is omitting: You can't hand work to an AI employee if you can't clearly describe the work.

The big companies already ran this experiment

You might remember when Klarna made headlines in early 2024. The company said its new AI assistant was doing the equivalent work of 700 customer service agents and had handled about 2.3 million conversations in its first month. It froze hiring and leaned all the way in.

By 2025, Klarna was hiring humans again. CEO Sebastian Siemiatkowski told Bloomberg that the AI was cheaper, but the service it delivered was "lower quality," and the customers had noticed.

And Klarna isn't an outlier. CNBC has been tracking a whole wave of these reversals. Commonwealth Bank of Australia replaced more than 40 customer service staff with an AI voice bot and then walked it back. Ford has brought back roughly 300 experienced engineers to handle quality issues its AI systems couldn't catch on their own.

The survey numbers tell the same story:

32% of US hiring managers told Robert Half they eliminated a role mainly because of AI and later rehired for the same or a similar position.
Orgvue found that 39% of business leaders made people redundant because of AI, and 55% of those leaders admit they made the wrong call.
Forrester predicted that roughly half of AI-attributed layoffs would be reversed.

Sit with that for a second. These are companies with entire IT departments, documented processes, operations teams, and budgets most of us will never see, and they still got it wrong.

AI clearly handled part of those jobs well. What tripped them up was everything around the work: the judgment calls, the exceptions, the messy one-off situations, the context that lived in people's heads and never got written down anywhere.

Graphic titled “The big companies already ran this experiment.” 32% of US hiring managers cut a role because of AI, then rehired for it. 55% of leaders who made AI layoffs say they made the wrong call. Klarna said its AI could replace 700 support agents, then started hiring people again.

Now bring that down to a small business

If that's what happened inside a company with a whole operations team, think about what happens in a business where most of the operations live in one person's head. Yours.

And I want to say this gently, because there's nothing wrong with that. Most founders never had a reason to write any of it down. The business grew around you, and it worked because you were there holding it together, and that's an accomplishment.

But when you build AI employees on top of operations that were never mapped out, you're asking each agent to guess. One agent gets fuzzy instructions, two agents end up with overlapping jobs, and nobody's really checking the output, so small mistakes pass from one agent to the next until a client is the one who catches them.

And here's the piece that worries me most. You set out to get time back, and now you're managing 50 new team members who can't ask you a clarifying question. That's more decisions running through you, more things to supervise, more places for something to break. Which is the exact pattern you were trying to get out of in the first place.

I'm genuinely excited about AI employees

I want to be really clear about this, because I don't want anyone reading this as "AI is bad, stay away." I use AI every single day in my business and with my clients. For small businesses, and especially for solopreneurs, AI employees might be the biggest opportunity we've had in a long time. You can get the capacity of a small team without the payroll.

I've just seen enough builds to know where they work. AI employees do well when:

They have one clear job, with a defined start and finish
The process they're running already works when a human does it
Someone knows what "done well" looks like and checks for it
Their work connects cleanly to whatever happens next in the business

Every one of those depends on having clarity on how your business actually operates.

A few questions before you build your first (or fiftieth) AI employee

Before you sign up for that next platform, try answering these for the role you want to hand off:

Could you walk someone through how this work gets done today, step by step?
Do you know what a great result looks like? Could you write it down as a checklist?
Where does this work come from, and where does it go when it's finished?
Who's reviewing the output, and how often?
If this agent stopped working tomorrow, would you know exactly what it had been doing?

If some of those are hard to answer right now, that's useful information. It just tells you where to start.

Build the engine first, then hire for it

The founders who are going to win with AI employees are the ones who get their operations out of their heads first. Workflows mapped, roles defined, and tools connected. Once that foundation is there, AI has an obvious place to plug in, and every agent you build gets a real job description instead of a hopeful guess.

That's the order that actually gives you capacity back, and it's the work I love doing with founders.

If you're curious where AI could genuinely help in your business (and where it would just add noise), book a Free Operations Audit. We'll look at how your business runs today and where the real opportunities are.

Sources

Comments