Part 3 of 3: Building Tech Teams in the Era of AI
Our core engineering team at A17 has been together since 2020. Seven or eight people who’ve stayed through scaling pains, market shifts, and roughly fourteen existential crises. In this industry, that’s basically a geological era. People ask me how we did it, expecting some elaborate retention framework. The honest answer is simpler and more uncomfortable: we got lucky with timing, and we happen to work on things that are genuinely interesting.
That second part –the “interesting work” part– used to be a soft perk. Nice to have. Now it’s the single most important retention lever you’ve got, and if you’re not using it, someone else will.
The math changed and your engineers noticed#
Here’s what happened. AI tools made your best engineers significantly more productive. Across our team and peers at other companies, we saw velocity gains between 10% and 70%, with a median around 40%. Three people now deliver what used to take five. Sprints that used to max out at 10 story points consistently hit 15.
Your engineers did this math before you did. And they arrived at a perfectly logical conclusion: if I’m producing 40% more value, I should be earning at least, maybe, 40% more money?
They’re not wrong. And the market agrees. Ravio’s data shows that AI-first startups at Series A and B operate with 34% leaner teams –but pay 36% more per individual contributor. Fewer people, higher output, higher pay. That’s the new equation. Your competitors aren’t building bigger teams. They’re building smaller ones and paying each person significantly more.
The uncomfortable truth for bootstrapped companies like mine (no venture money, no big investment rounds) is that you can’t always match these numbers. We always tried to compensate with everything else — interesting projects, autonomy, flexibility. And honestly, it worked. But pretending the salary pressure doesn’t exist would be naive. The engineers who’ve successfully adopted AI for their work are getting actively more expensive, and the companies that don’t have a champion yet are willing to overpay to get one from you.
AI FOMO is real, and it works both ways#
Retaining tech talent in the AI era isn’t only about money, though. There’s something else happening that’s harder to put in a spreadsheet.
Your engineers are watching the industry shift in real time. They know that the skills becoming valuable right now —- AI-assisted development, prompt engineering, building AI-native workflows —- are the same skills that will determine their career trajectory for the next decade. Nearly 7 in 10 workers say they want to grow their AI skills specifically to stay marketable. This isn’t casual interest. It’s career anxiety dressed up as professional development.
And here’s the retention angle most founders miss: Betterworks found that nearly 4 out of 5 AI power users –your best, most productive engineers– are actively looking for other jobs if they feel constrained in their AI usage. Let that sink in. Your most valuable people, the ones driving the productivity gains you’re benefiting from, are the first ones out the door if you don’t give them room to grow.
I’ve seen this pattern up close. Engineers who are already senior, already well-paid, who are 35 or 40 years old — they understand better than anyone that seniority alone doesn’t guarantee relevance anymore. Everything changes fast. They need practice, not just permission. They need recognition that their new skills matter. They need an employer who gets this.
If you give your engineering team this opportunity, they stay. Maybe forever. If you don’t develop them in this direction, don’t adopt AI into your processes, don’t even give them sandbox time to experiment.. they start looking. Not at pet projects on weekends — at actual jobs, with employers who’ll let them grow.
Upskilling is cheaper than replacing#
Only about 16% of workers have high AI readiness, according to Forrester. That number is projected to reach just 25% by end of 2026. The gap between what companies say they offer and what employees actually get is brutal –44% of employers claim they provide AI training, but only 33% of employees confirm they’ve received it.
Companies that delay upskilling face 26% higher turnover. I want to repeat that because it’s probably the most actionable number in this entire series. Twenty-six percent. That’s not a rounding error, that’s a quarter of your team walking out the door because you didn’t invest in their growth.
So what actually works? In our case, it wasn’t formal training programs or expensive workshops. It was messier than that, and more organic.
We took the time saved by AI-assisted development —- the efficiency gains we talked about in Part 2 — and instead of immediately filling it with more sprint tickets, we spent some of it on play. Internal hackathons where teams of three would build personal AI assistants on OpenRouter. Competitions for who can automate the most tedious part of their workflow. Show-and-tell sessions where someone demonstrates a new technique they figured out last week.
One hackathon started because our Gen Z engineers just… announced it. “Hey, look at this cool thing. You can build an assistant that learns your codebase and talks to you through Telegram. Want to try? Teams of three. Tokens are on the company. Let’s go.” And people showed up, because it wasn’t mandatory —- it was interesting. The champions paired up with the curious, and even some skeptics came along to make sure nobody built anything too ridiculous.
No fancy framework. No LMS. No mandatory AI certification. Just space to play and people excited enough to fill it.
Hire a few, grow the rest#
Here’s the economic reality of retaining tech talent when AI skills are in demand. The engineers who already have strong AI adoption experience –who’ve led teams through this transition, who have real case studies— they’re in maximum demand right now. There are very few of them. They are very expensive.
Hiring a whole team of these people is something only well-funded companies can afford. For the rest of us, the strategy is different: hire one or two champions, and let them pull everyone else up.
Find a couple of engineers who can show your existing team what good AI-assisted work looks like in practice. Give them interesting challenges. Let them lead by example rather than by mandate. Internal upskilling is vastly cheaper than external hiring, and it has a compounding effect —- every person who levels up becomes someone who can level up the next person.
This is also, by the way, a retention strategy for the champions themselves. Smart engineers don’t just want to use AI. They want to teach, mentor, and shape how their team works. Give them that opportunity and they have a reason to stay that goes beyond compensation.
The new team economics#
Let me be blunt about where this is heading. Entry-level tech hiring has dropped significantly — Ravio’s 2025 data shows AI-first companies just aren’t building the same junior-heavy teams anymore. What’s emerging instead is the “Super IC” model: experienced professionals who combine strategic thinking with hands-on, AI-augmented execution. One person doing what used to require a small team.
Midjourney, Replit, ElevenLabs — these companies scaled to millions in revenue with fewer than 30 employees. That’s not a quirk. That’s the direction.
For retention, this means something specific: the people who stay with you and grow into these super-IC roles are extraordinarily valuable. Losing one isn’t like losing one of five interchangeable team members. It’s losing a node that your entire operation depends on.. And they know it.
The old retention playbook — annual reviews, standard raises, maybe a team offsite — wasn’t built for this reality. When one engineer with AI skills can genuinely replace two or three without them, the power dynamic shifts. Your retention strategy needs to reflect that shift: real investment in growth, meaningful work, competitive compensation, and the autonomy to keep evolving.
What this adds up to#
Retaining tech talent in the AI era comes down to understanding that the game changed and acting accordingly. Your best engineers are more productive, more valuable, and more aware of their market worth than ever before. They want interesting work where they can develop AI skills. They want employers who invest in their growth, not just their output. And yes, they want to be paid fairly for the disproportionate value they now create.
You don’t need a massive budget to do this. You need to care about it. Give people space to experiment. Let your champions teach. Reinvest efficiency gains into growth rather than just more tickets. And have honest conversations about compensation before your competitors have them first.
None of this is rocket science. But doing it consistently, when there are always urgent things competing for your attention and budget? That’s the hard part. That’s always been the hard part.