This AI company built an AI-proof recruitment process—and just got acquired for $1.1 billion

If you’re looking for a job or hiring, the question is no longer whether AI is involved—but how aggressively you’re using it. 

Generative AI has wormed into every stage of recruitment, from drafting applications and filtering candidates to AI-led interviews. It’s the wild west out there. (And it’s getting wilder.) Both employers and prospective employees are exasperated.

Examples abound. Last year, Anthropic urged prospective applicants to not use AI systems when applying to jobs at the AI company, even asking them to sign a contract to confirm they read and understood the ask. Goldman Sachs has implemented blocks and employs AI detection software, while McKinsey actually requires candidates to use its internal chatbot Lilli for practical consulting tasks.

As all parties scrabble to jump through the new LLM-molded hoops, the mechanics of hiring are going under the microscope. CVs, take-home tasks, interviews—all of it is ripe for re-engineering. In a market obsessed with automation for automation’s sake, one particular Swedish AI darling is standing steadfast in its AI-resistant approach. 

“Jam around on ChatGPT—but it’s not going to save you”

Sana Labs, founded in 2016 by high school prodigy Joel Hellermark, builds software that helps organizations train employees and deploy AI assistants using their own institutional knowledge. The client roster features the likes of Spotify, Hootsuite, Strava, and Robinhood, and in November, it was acquired by Workday for $1.1 billion, a deal that’s reportedly one of Europe’s largest AI acquisitions. The business runs on generative AI, and yet, when it comes to recruitment, Sana has designed a system where leaning on LLMs is a liability.

As the AI-on-AI arms race between candidates and recruiters intensifies, Sana’s contrarian bet appears to be paying off. It has scaled to a 300-person team, adding 150 new hires in the past year alone, with an attrition rate that’s well below the industry benchmarks (in the European tech sector, the average sits at 17.4%).

Everything usually starts with a poach. Exceptional staff are rarely looking for a job, explains head of operations Olivia Elf, who’s been a ‘Sanian’ since 2021. The team spends weeks, months or sometimes years chasing down talent—once their interest is piqued, the window is narrow. “We need to make it as easy as possible for someone to apply,” Elf tells Fast Company. “We have two clicks, not 10.” 

Candidates upload a CV and add contact details. That’s it. Everyone hears back within 24 hours—often in half that time—whether they’re rejected or cleared to move on.

Next comes a rapid-fire intro call with an available Sana staffer. There’s no central recruiting figurehead: every single employee at the company is trained to interview, which Elf calls “the best life skill anyone can have.” Interviewers are encouraged to push boundaries—the more mind-bending the questions, the better. But first, Sana’s litmus test: the MEH framework. Are you a missionary or a mercenary (M)? Do you strive for excellence, always (E)? Do you demonstrate humility (H)? Pass that (in other words, don’t be a “meh” candidate), and you’re sent a role- or domain-specific take-home task. 

“We expect people to use every tool available—we’re not looking to cap anyone,” Hellermark, who founded Sana when he was just 19, tells Fast Company. “Setting artificial constraints in hiring would contradict the environment candidates are stepping into.”

Hopefuls are invited to a Sana hub in Stockholm, New York, or London (most work is on-site anyway, with relocation common) to present their take-home task. This is where reliance on LLMs is rendered useless. Feedback is instant, and candidates should expect to be challenged on their thinking when they present. “We get quite aggressive, but it’s done with care and humor, so the experience should simulate the experience of working here,” says Elf.

If a slide deck smacks of ChatGPT, interviewers press harder. “We want to know which parts they used AI for and we’re not going to tiptoe around that,” says Elf. In “You can jam around on ChatGPT before, sure—but that’s not going to save you.”

Of course, not every company can afford to fly candidates from around the world to prove their skills without AI in person. “It’s a luxury that can get expensive if you’re hiring at high volume,” says Andreas Bundi, who’s been a recruiter for 15 years and founded Berlin-based HR consultancy Bundls. “It’s more realistic for companies to use tools and good judgment in how they conduct remote interviews to minimize the risks of candidates cheating.”

To boot, Sana’s line of questioning is deliberately personal and reflective. They want to see deep into the mind palace, so ChatGPT is a useless shortcut. Favorite questions include: “Looking back at the most pivotal decision you made, what principles shape those?” “When you’re sitting on your couch scrolling, where do you end up?” “When’s the last time you changed your mind about something important?” All of which are questions candidates would struggle to answer with any gumption using an LLM.

Looking for: inquiring minds

On the surface, all this sounds quite quaint. But after years of remote work, highly automated hiring funnels, and now hacks to game every stage of the process, the basics are easy to overlook. Paranoia is high, while trust is low. Reports of job and employment-related scams have increased by 1000%, and actual losses run into the hundreds of millions of dollars. Everyone assumes everyone else is cheating.

Bundi has seen more companies balk at in-person task presentations and interviews, which he says is a natural response to the overreliance on LLMs: “I’ve seen a huge increase in fraudulent applications, and certainly more AI firms in my network are using third-party tools to detect when it’s being used in applications and interviews,” he explains. He adds that the use of AI auditing tools will rise, especially in fast-scaling firms. “But I don’t think the in-person thing is necessary—I could definitely imagine there is some candidate pushback to the carbon emissions, even if companies no longer seem to care about projecting an ethical image themselves.”

By now, Sana’s hiring philosophy is well-worn: the process has barely shifted since 2022, before the AI boom ruptured recruitment. There is, however, one new non-negotiable: candidates must be genuinely curious about AI itself. This is, after all, an AI company where employees are constantly stress-testing models and building automations for kicks. Being unilaterally curious outranks networks, prestige-level employers, and even formal credentials.

“Curious people are curious beyond ego,” says Elf. “They’re constantly reinventing themselves, comfortable naming their weaknesses, and actively working toward their own version 2.0.” As well as being a trait that only humans have, she adds, curious people are almost always the most fun people to work with, which for Sana is important to keep in mind during the hiring process.

As large language models metabolize the internet and spew out its average, the impact on hardcore users is coming into focus. Researchers have found that while LLMs can help with goal-directed tasks, like answering specific questions or copyediting documents, they hinder “divergent thinking,” or the ability to generate novel ideas. Curiosity has moved onto the endangered list, and more employers might want to consider doubling down.

Elf agrees. “AI is doing a lot of wonders, but it is absolutely a threat to rigorous thinking, one that’s reducing the need to look into your own head,” she says. 

While much of the hiring market optimizes for speed at great cost, Sana seeks depth and an infinite skillset. Curiosity, it turns out, is both a soft skill and a survival skill in an AI-native world. 

As we hurtle towards conformity, it’s also the hardest one to fake.

View the full article