The AI Talent War: Companies Aren’t Playing Moneyball

After building some of the earliest AI teams at Google and Meta, here’s the undervalued talent most companies are missing

The AI talent war is real, but most companies aren’t playing Moneyball. They’re desperately hiring AI engineers, ML researchers, and data scientists when what they actually need is something much rarer: people who can turn technical capabilities into products that millions will use.

Just like Billy Beane found undervalued players who could actually win games, there’s undervalued AI talent that can actually ship products. The companies that figure this out first will have a massive competitive advantage.

Having built some of the earliest AI teams at both Google Research and Meta’s FAIR, I’ve seen this talent opportunity up close. Companies think they need more AI technical expertise, but what they really need is people who can bridge technical capabilities with real user needs: the undervalued players who can actually win the game.

The Hidden Value in AI Talent

Here’s what I see happening: companies are loading up on technical talent while completely missing the product, design, and user experience skills that make AI actually useful. They’re hiring brilliant people who can build sophisticated models but have never shipped a consumer product or understood what users actually need.

This creates a massive opportunity for companies that recognize the undervalued skills. While everyone else is bidding up AI PhDs, smart companies are finding people who can actually turn AI research into products people love.

• Working with non-deterministic systems where you can’t predict exactly what the user will get
• Building trust without predictability — users need to feel confident even when AI responses vary
• Designing feedback loops that make the AI better while feeling natural to users
• Understanding technical constraints that change monthly as models improve

When I started at Google Research in 2017, we didn’t have “AI product managers” or “AI designers.” We had curious people willing to experiment with technology that barely worked. The people who succeeded weren’t the ones with the deepest technical AI knowledge. They were the ones who could prototype rapidly, think systematically about user needs, and communicate across disciplines.

At Meta, I saw this pattern again: the most successful AI products came from cross-functional teams that included people who understood both what the technology could do and what users actually needed.

The Technical Depth + Product Instincts Combination

The most valuable AI talent has a rare combination: deep enough technical understanding to know what’s possible, combined with product instincts to know what’s useful.

At Meta, I worked with researchers who could explain transformer architectures in detail but couldn’t design a simple user flow. I also worked with traditional product managers who created beautiful roadmaps for AI features that were technically impossible to build. The magic happened when we found people who could bridge both worlds.

What technical depth actually means in AI: • Understanding how models are trained and what affects their outputs • Knowing the difference between what’s a model limitation vs. an engineering problem • Recognizing when AI uncertainty is a bug vs. a feature • Seeing patterns in AI behavior that inform product decisions

What product instincts look like in AI: • Knowing when to hide AI complexity vs. when to expose it • Understanding which AI capabilities solve real user problems vs. which are just impressive demos • Designing for the 80% case while gracefully handling edge cases • Building products that get better as they scale, not worse

The people who have both are incredibly rare. They’re usually people who started in traditional tech roles and adapted to AI, not people who started with AI and learned product skills.

Building Teams That Navigate AI Uncertainty

Traditional product development assumes predictable outcomes. AI product development requires comfortable uncertainty. Most companies try to apply traditional processes to AI and wonder why everything takes longer than expected.

Here’s what actually works:

1. Prototype Everything, Ship Nothing (At First)

At Google Research, we built hundreds of prototypes knowing 99% would never ship. This isn’t wasteful — it’s the only way to understand what AI can do. Companies that try to go straight from concept to product miss the crucial exploration phase.

2. Build Learning Loops, Not Feature Lists

Traditional roadmaps don’t work for AI. Instead of committing to specific features, commit to learning objectives. “Can we make image editing feel magical?” is better than “Build an AI photo editor with these 10 features.”

3. Hire for Curiosity, Not Experience

The AI landscape changes so fast that yesterday’s expertise is often today’s outdated knowledge. I’d rather hire someone who’s genuinely curious about AI capabilities than someone who claims to be an “AI expert.”

4. Create Cross-Functional Innovation Teams

AI products require constant collaboration between researchers, engineers, designers, and ethicists. Traditional handoffs don’t work when everyone is figuring it out together.

The Roles Companies Actually Need

Instead of hiring more AI engineers, companies should look for:

AI Product Strategists: People who can identify which AI capabilities solve real business problems. They need enough technical depth to separate hype from reality, plus enough product sense to build sustainable businesses.

AI Experience Designers: People who can design for non-deterministic systems while maintaining user trust. They need to understand both human psychology and AI behavior patterns.

AI Research Translators: People who can take cutting-edge research and turn it into practical product features. They need to speak both “researcher” and “product manager.”

AI Ethics Integrators: People who can embed safety and fairness considerations into the product development process without slowing everything down.

Cross-functional AI Leaders: People who can build and lead diverse teams that include researchers, engineers, designers, and ethicists working together on uncertain problems.

What This Means for Your Hiring

Stop looking for “AI expertise” and start looking for these traits:

• Systems thinking — Can they see how different parts of a complex system interact? • Rapid learning — How quickly do they adapt to new information? • Comfort with ambiguity — Do they need perfect requirements, or can they work with directional goals? • Cross-functional communication — Can they translate between technical and business teams? • User empathy — Do they really understand what users need vs. what they say they want?

Red flags in AI talent: • Claims to be an “AI expert” (the field moves too fast for permanent expertise) • Focuses only on the technology without understanding user needs • Wants to build AI for AI’s sake rather than solving real problems • Can’t explain complex technical concepts in simple terms

The Future of AI Talent

The AI talent war will intensify, but it won’t be won by companies with the biggest recruiting budgets. It will be won by companies that understand what they actually need and create environments where curious, adaptable people can do their best work.

The most successful AI companies won’t just hire AI talent, they’ll develop it. They’ll take great product people and teach them about AI, rather than taking AI researchers and hoping they learn product skills.

As AI becomes more integrated into every product, the distinction between “AI talent” and “product talent” will disappear. The winners will be the people who can navigate uncertainty, think systematically, and build products that actually help humans.

What’s your experience hiring for AI roles? Are you seeing the talent mismatch I’ve described? I’d love to hear your perspective in the comments.

If you’re building AI products and struggling with talent strategy, let’s connect. I advise companies on building AI product teams that can navigate uncertainty while shipping real products.