Hiring and team building in the age of AI

At times, companies come to us with hiring requests that worry me. They have a senior engineer looking after architecture, they’ve moved heavily into AI, and now they want juniors who can prompt it to write code. AI will handle the difficult work, so apparently the people using it don’t need much experience.

That to me is a failed proposition. Someone still needs to understand what AI is doing, including when it comes up with fantasies. Which it does.

I use AI extensively, and I love what it makes possible. A smaller team can do work that previously required more people. But you need people who can figure things out, question the output, and take ownership. That combination makes the hiring pool more specific and international hiring more important. Why restrict yourself to one location when the world is available to you?

Experience matters when the answer looks right

I remember a company where I’d worked for about four years. A new VP of engineering wanted a junior engineer to make a change. The engineer had only been there a few months. To both of them, it looked simple.

We went into a conference room and walked through it. The first change would affect another part of the system. Adjusting that would affect something else. By the time we followed the dependencies, we were talking about rewriting the search engine and the reservation engine.

From the outside, the existing design probably looked ridiculous. Honestly, it probably was. But there was a reason it worked that way.

“I didn’t know.”

You didn’t know, and you didn’t ask. You assumed.

That’s the experience I want involved when my team uses AI. Someone who knows why an apparently simple change needs another look. Access to a tool doesn’t give you four years of product history or tribal knowledge.

The next generation of senior engineers

Of course, some companies have gone the other way. They’re keeping their seniors and stopping junior hiring because AI can do the junior work.

I keep asking the same question: what happens when that senior engineer retires? Who replaces them? Look five years ahead. Who are you training now?

Meanwhile, juniors struggling to find work ask us, “What do we need to learn?” They expect a tool or a programming language. My answer is: “What you need to learn is how to solve problems, how to get to the bottom of the problem, how to figure things out.”

Those soft skills deserve more attention. Ask questions. Investigate. Own the problem until it’s solved. Engineering fundamentals still matter, but knowing another language won’t help much if you accept an answer without checking it.

Companies have responsibility here too. Do you really think a company hires a college student because of how much they’ll contribute during two months of summer? No. They’re training someone they may hire after graduation. That investment still needs to happen. AI doesn’t train your future senior engineers for you.

Protect the knowledge inside your team

Part of that training is learning what experienced colleagues know. They remember the bug fix from three years ago that broke something else. They know why the obvious solution wasn’t used.

Is it documented? No. Should it be? Yeah, but it’s not.

That’s tribal knowledge. Before removing someone from a team, think about what leaves with them. AI can help document decisions and why they were made. I want people using it for that. But somebody has to supply the history in the first place, and juniors need the chance to learn it.

Why international hiring matters more

These are the people you need to find and retain: engineers with technical understanding, curiosity, adaptability, and ownership. Finding that combination is harder than finding someone who knows a language. Restricting the search to one location makes it harder again.

I get messages from agencies saying they only hire from Argentina, Brazil, or Poland because that’s where the best people are. They all have great talent. What the message also tells me is that the agency can’t hire elsewhere.

At Mirigos, we search across regions. A great Argentinian developer could be living in Portugal. A Ukrainian engineer could be in Spain. I want the right person, wherever they happen to live.

How you hire them matters too. With outsourcing, another company manages the work and the people. That can leave the product knowledge you need sitting outside your team. You can hire internationally yourself, but then recruiting, payroll, and local employment requirements are yours to figure out.

Team augmentation is the third option, and it’s what we do. You interview, choose, and manage the engineer. We handle recruiting, payroll, HR, compliance, and local employment. Our local recruiters know the markets and compensation. The engineer works directly as part of your team, building knowledge of your product.

Interview people the way they will work

To find those people, look at how they solve problems. We ask engineers to walk us through production code they wrote with AI. What worked? What went wrong? What did they have to change?

Too often, someone gets an AI response and says, “This is it, I’m done.” Well, did you read it? Did you review it? Does it even match your expectations?

Years ago, I interviewed for a VP of engineering role managing about 70 people. They had me write a SQL query on a whiteboard. I made a syntax mistake, and one interviewer made a big deal out of it.

I asked what would happen if I entered it into the actual tool. It would flag the syntax error. Exactly. So I’d correct it. What are you testing? “Are you testing if my memory is perfect? It’s not.”

Give people a problem and let them solve it the way they would at work. “Google it, AI it, whatever you want to do.” Then ask them to explain their decisions. If I take my car to a mechanic, do I expect them never to look in the manual? It’s not a university exam. It’s a job.

My hiring principle has always been to hire smart people who can learn stuff. I grew up in the Soviet Union. If your radio broke, you couldn’t just buy another one or walk into a store for the part. You figured it out. When someone asks what I know, my answer is, “What do you need me to know?” Give me three days and I’ll figure it out.

That mindset matters when the tools keep changing.

Measure the work it takes to get a usable result

We use both ChatGPT and Claude. One week it’s, “ChatGPT, you gotta be fired because you lost your brain.” The next week it’s, “Claude, what are you smoking?” You have to keep checking what works.

I’ve spent longer on some tasks with AI than I would have spent manually. “No, no, no, that’s not what I meant. Also, I forgot to tell you this.” You look at the clock, and two hours have gone. That time counts too.

So measure the whole job, including review, debugging, and fixing the output. If you now have two engineers maintaining AI code that was supposed to replace one engineer, what exactly did you save?

I have clients spending more on AI than the cost of hiring an engineer in Latin America. Look at the work you actually get for that money. And don’t build a hiring plan assuming today’s prices will last forever. I expect the economics to change. Where will they land? I don’t know. Nobody does.