High EQ separates great AI-fluent candidates from good

The most durable skills are the most human ones. 

Demis Hassabis is the CEO and co-founder of Google’s DeepMind Technologies. He runs one of the most advanced AI labs on earth. And when someone that.. that.. ahem.. deep into AI tells you which human skills will matter as the technology gets more capable, it's worth a pause.

He was asked this question during a recent Stanford interview. In his answer, he named taste, design sensibility, original thinking, and the ability to synthesize different subjects as the most critical skills going forward in this era. 

In both science and the arts, he argues, the people who pair creativity with technical tools are the ones who pull ahead. He went further to say that what separates great scientists from merely good ones isn't technical capability, it's creativity. And the biggest differentiator of all, whatever your job, is the ability to connect with another human.

EQ may end up mattering more than IQ

Senior recruiter Hugo Bouré argued that "AI won't take our jobs, it will kill the parts of them that should never have existed. Strip out the admin that grew around the role like scar tissue", he wrote, and what's left is "either real skill or empty space." 

The real skill of recruiting (and likewise in many other jobs) lies in soft skills. Skills like judgment, communication, design, and strategic thinking. 

The market is converging on a single idea: the most durable skills are the most human ones. 

And while naming those skills is easy. Hiring for them is not.

Why naming skills isn't measuring them

Hassabis describes the human qualities that survive automation but he doesn't tell you how to spot them in a candidate or demonstrate them as an applicant. That gap is exactly where hiring is starting to struggle. And it’s happening just as AI fluency has gone from a nice-to-have to a baseline expectation for nearly every role.

Our State of Hiring for AI Fluency report surveyed around 2,000 hiring experts, and the findings are uncomfortable. 95% of organizations now list AI fluency as a hiring requirement, and 71% have formally defined what it means. Yet 59% admit they've made a bad AI hire anyway.

The infrastructure is in place. The signal is broken.

Taste, judgment, and creativity don't show up in a tool list, and they rarely get a chance to shine in a thirty-minute chat.

So if AI fluency is now non-negotiable, the candidate who stands out isn't the one who simply uses AI. It's the one who brings high EQ to it: judgment, communication, and human sense applied across every part of the work. 

Our report breaks that work into five pillars that make a candidate AI-fluent. Here's what high EQ looks like inside each one.

Pillar 1: Applied AI use and workflows

This is the pillar people assume is purely technical. It isn't. Anyone can prompt a model. The EQ that matters here is judgment: knowing when to automate, when to augment, and when to step back from AI entirely. That last call, when not to use the tool, is where taste lives, and no model makes it for you.

It looks different in every role. A marketer can generate fifty ad variations in seconds, so the skill isn't generating them, it's the taste to know which three are worth running and the judgment to spot the one that's technically clever but off-brand. A support lead can auto-resolve routine tickets, and the judgment is recognizing the angry customer who needs a person, not a faster bot.

In each case, AI does the volume. The human decides where it earns a place and where it doesn't.

This is also where hiring most often goes wrong. Our report calls it the confidence-versus-competence problem: the candidate who describes AI fluently but can't operate under real conditions. 31% of hiring managers say telling the two apart is their biggest challenge. 

You don't catch it by asking which tools someone uses. You catch it with a scenario-based task that mirrors the actual job, then watching whether they know when to put the tool down.

Pillar 2: Learning and digital agility

The human skill here is adaptability, and a particular kind of it: the resourcefulness to stay useful when the ground keeps shifting. Tool-specific knowledge now has a short half-life. Mastering one platform isn't fluency. Knowing how to learn the next one is.

A software engineer feels this fastest. The framework they relied on last year gets a breaking update, a new model deprecates the library they built around, and the EQ on display is whether they wait to be retrained or go figure it out, test the alternative against their own code, and write up what changed so the team isn't relearning it five times over. A marketer's analytics tool sunsets a feature mid-campaign and the same instinct decides whether they stall or adapt. A recruiter's sourcing tool changes its model and the agile ones have already pressure-tested a replacement.

AI is the moving target here, not the helper. The skill is staying oriented while it moves.

Test candidates on this by changing the rules mid-task: introduce a constraint, take away a tool, and watch. Weak candidates tweak the output. Strong ones step back, reframe, and rebuild the whole approach. That's the difference between someone who knows a tool and someone who can learn any of them.

Lou Adler, creator of Performance-based Hiring, puts the burden of proof on the candidate. "You gotta prove to me," he said during a recent TestGorilla webinar for our Hire for the AI Era audience. "Oh, I used AI to do this. Well, give me an example of when you did it. What'd you do? How'd you learn that?" Vague answers fall apart under that pressure. A real one holds its shape because the person actually lived it.

Pillar 3: Systems thinking and problem solving

The human skill here is written on the tin. Can they think at a systems level and solve problems effectively? Can they use AI beyond their own desk and see how its output lands across teams, data, and users?

A software engineer can ship an AI-generated feature in an afternoon. The systems thinker asks the questions the model can't: who consumes this downstream, what assumptions are baked into it, what's the blast radius if it's wrong at scale, and where does a human checkpoint need to sit before it ships. A product manager does the same with a recommendation engine. A data analyst does it before a model's output feeds a decision three teams away.

AI makes individual output cheap, which is exactly why this matters more, not less. When everyone can produce fast, the rare skill is producing something the wider system can actually absorb.

It's also the hardest pillar to detect. 23% of hiring managers flag difficulty assessing it, and it's the most consequential to miss. It guards against the individually productive hire who can't scale: the engineer who 10x'd their own output but left documentation so thin nobody else could follow it, and bypassed security and review on the way. Individual productivity that can't operate inside a system isn't a gain. It's a liability.

Pillar 4: Responsible and ethical AI use

The human skill here is conscience paired with foresight: the instinct to spot where AI use creates risk before it becomes an incident. It's spotting bias and governance gaps early, knowing when a human must stay in the loop, understanding the compliance rules for your industry, and documenting the decisions so they hold up later.

A software engineer wiring an LLM into a product decides whether to log how user data flows through it, or to find out the hard way during an audit. A recruiter deploying an AI screening tool asks whether it's been checked for demographic bias and what happens to the candidates it filters out. None of these are technical questions. They're judgment calls about consequences the model never sees.

Hung Lee, curator of Recruiting Brainfood, made the same point at our Hire for the AI Era event. AI, he argued, is not end-to-end but "middle-to-middle. In other words, on either end of that workflow is a human being." Someone has to set the plan, monitor the process, and sense-check the output before it ships. The judgment lives at the edges, where a person decides what the model is allowed to do and whether to trust what it produced.

AI is fast and confident and has no sense of what it's risking. The human supplies the part it lacks.

This is the most frequently cited number-one skill in the report, ranked the single most important pillar by 20% of respondents, ahead of even applied AI use. Practitioners are ahead of job descriptions here. The market hasn't caught up to what hiring managers already know they need. It guards against the fast mover who creates slow-burning liability. Teams that chase speed without guardrails don't dodge the consequences. They defer them.

Pillar 5: Human-AI collaboration and communication

This is the one Hassabis put above all others: the ability to connect with people. Here, it means treating AI as a teammate whose work you can explain, and making your own work with it legible to the humans around you. 

A software engineer who uses AI heavily but leaves no trace of how creates a black box that only they can maintain. The high-EQ version documents which prompts and inputs they used, which outputs they verified by hand, and where their own judgment overrode the model, so the next engineer can pick it up without starting over. A data analyst does the same with an AI-assisted report. A customer success lead shares the AI workflow that's working instead of hoarding it.

AI multiplies what one person can do. Communication is what lets that multiplier reach the rest of the team.

It guards against the AI fluency silo. A brilliant individual who can't communicate their AI use creates knowledge concentration, not knowledge distribution. The effect only compounds when fluency is transferable. When it isn't, it walks out the door when they do.

Stop hiring the best storyteller

AI fluency is the baseline now. Every serious candidate will use the tools, and most will talk a good game about them. The question worth asking isn't whether someone uses AI. It's whether they bring judgment, adaptability, conscience, and communication to how they use it. That's the EQ that separates a good AI-fluent hire from a great one, and it shows up across all five pillars.

The pillars turn that from a feeling into something you can measure. They shift the hiring question from "do you know AI?" to three harder, better ones: can you deliver with it, can you be trusted with it, and can you help others reach your level?

These aren't interview questions. They're evaluation dimensions, and they need a different kind of evidence than a thirty-minute conversation can produce. A polished talker clears that bar easily. A high-EQ operator only proves it in the work.

So before your next AI-era hire, change what you ask. Stop asking which tools a candidate uses. Start asking them to walk you through a workflow they redesigned: what broke, what they verified, and where they decided not to trust the model. It costs nothing to change today, and it surfaces judgment over vocabulary instantly.

Then back it with structure. See how TestGorilla measures the five pillars of AI fluency across technical and non-technical roles in the State of Hiring for AI Fluency report.

Hassabis is right that EQ may matter more than IQ. The trick is proving a candidate has it before you hire them, not after.

Want to see how you can leverage TestGorilla to start assessing your candidates for AI fluency? Try it free today.