When Estefania Angel started working as an executive assistant at a large tech company a few months ago, she noticed something counterintuitive: while her company’s job was to help other enterprises set up AI to streamline their in-house tasks, her company didn’t use those systems internally itself.

Using AI apps in Slack, Outlook, and Google to track various assignments and ping colleagues, Angel got the attention of her superiors. One even asked Angel to teach her how to use AI at work. 

“We started tracking a whole project that she was doing,” says Angel, who works as an executive assistant (EA) with EA service company Viva Talent, streamlining the project’s workflow.

That was just the first step. Ultimately, through Angel’s use of AI to make a variety of office tasks increasingly efficient, more and more of her colleagues began adopting those AI-driven processes until it became the company norm. 

It wasn’t company executives driving AI adoption—but rather lower-ranking, self-taught employees who helped AI use cases trickle upward.

This bottom-up AI adoption tracks with wider trends: Last year, McKinsey found that “the biggest barrier to scaling [AI] is not employees—who are ready—but leaders, who are not steering fast enough.” McKinsey researchers surveyed 3,613 employees and 238 C-level executives and learned that the latter seriously underestimate how much the former use AI. C-suite executives, for example, believed 4% of employees used gen AI for at least 30% of their work, when employees’ self-reported percentage was three times higher.

While EAs can be big drivers of AI filtering up to executives because of their proximity, sources speaking with Fast Company noted how recruiters, data workers, individual contributor (IC) coders, project coordinators, and even valets have sparked widespread AI adoption across their organizations.

Executives are largely not the ones on the frontlines of AI at work. Employees engaged in day-to-day tasks are steering their companies’ AI adoption—from the bottom up.

“A wild game changer”

At human resources software company Justworks in New York, one IC recently started using an AI agent that investigates the company’s code repositories for potential bugs. Once it finds a likely problem, it reports back to another AI agent, which does some QA testing and, if it detects a bug, opens a request for a fix.

That IC ended up “largely automating like 80% of the on-call process for his team,” says Justworks senior engineering manager Ryan Taylor. It’s been a “wild game changer,” Taylor adds—what began as “just an experiment” by the IC is now something his team is working on rolling out more widely across the company.

Cortney Hickey, executive operations director at automation software company Zapier, says she and her colleagues “have been influencing our execs in many ways” on AI, like in designing how decisions move across the organization: For example, Zapier’s recruiting team “has done a lot with AI that’s also trickled up,” including how the company generates pre-meeting briefs. 

Chris Morrison, who started as a valet at the upscale grocery chain Erewhon in Los Angeles in 2017, ultimately ended up developing AI systems that now aggregate the entire company’s data. Good at his valet job, he shortly got promoted to driving the company’s CEO, who “slowly started to realize that I was good at computers,” Morrison says. 

Having had on-the-ground experience at many of Erewhon’s stores, Morrison started driving less and working more with the CEO’s EA to set up pipelines and databases for Erewhon, using AI to automate and streamline tasks based on his knowledge of the company’s operations. 

These pipelines spread outward to colleagues and upward to superiors. Today, Morrison is a business analyst and AI lead at Erewhon.

Boots on the ground

It’s only natural for AI functions to move bottom-up at a company, because workers know their domains more intimately than the people who oversee them.

At Justworks, Taylor has seen “a lot of AI initiatives come from low-cost, quick-iteration experimentations,” he says. It’s simple to play around with AI to see how it can make your workday easier, and when something succeeds, other people at the office tend to notice and even start using it themselves.

“In these support or operations roles, you just spot the friction,” says Zapier’s Hickey. When she and her colleagues do, they’ll start piloting AI solutions. “Eventually, you test it with an exec, and they’re like, ‘I want more of this.’”

Jodie Mears, a UK-based EA at infrastructure software development company Bentley Systems, mentors EAs around the world. She’s been hearing, however, that a lot of her mentees’ executives don’t want them to use AI. “They feel like it’s cheating,” Mears says, rather than streamlining. Still, she says it’s best practice not to hide your AI use as an employee.

Employees “will battle between not wanting to admit that they used AI or an automation to make their role faster in fear of downplaying or downgrading their traditional duties,” Mears says. Though some fear that employees bringing on AI to take over some of their job functions represents an existential danger to their role, it might in fact make them more valuable. They become the “translation layer between the tools and how [they] work with leadership,” Hickey says. 

As Fineas Tatar, co-founder of Viva Talent, puts it, “My EA teaches me new things all the time. Especially when it comes to anything AI-related.” For instance, he says his EA has helped him reduce his meeting prep time from 30 minutes to just two through AI agent-created pre-meeting briefs.

“Let me have a piece of that”

Even though so many AI-assisted workflows originate below the C-suite, executives and managers can participate productively in employees’ AI iterations. Toward the start of this year, Justworks higher-ups noticed employees eagerly adopting AI, and decided to foster that process by giving them small budgets to spend on their trials with AI products. 

“Company leaders start with enablement, being like, ‘This is an industry thing. We need you to be leaning in on this. Here’s a budget,’” Taylor says. Leaders noticed that these developments were worth funding, “but the actual implementation and change has to very much come from the ICs and boots on the ground.”

By March, Justworks was hosting an internal hackathon in which employees were encouraged to “do whatever you want, but you’ve got to build it using AI,” says Taylor, resulting in some useful implementations, such as the ability to extract specific, uniform information from a number of differently formatted documents (like a bunch of CVs).

Similarly, Zapier hosted a company-wide hackathon where employees were encouraged to take a week to build with AI and then share their work. “Execs have an important role of empowering, but I don’t think they are necessarily the ones providing the foundational ways of working,” Hickey says.

For Mears, higher-level encouragement about her use of AI has come in the form of praise and approving nods. Workers also share their AI innovations in a dedicated chat. “The prompts that get shared are really quite invigorating,” Mears says. Team members inspire and adopt each other’s AI implementations. And with nearly every AI tool she’s used, she’s found her executive asking, “Show me what you did to free up your time. Create one of those for me.”

“You don’t have to be a particular IT whiz to use this,” she adds. “I think that’s the biggest revelation that trickles up to the C-suite: ‘Wow, my EA is doing that. Let me have a piece of that.’”