Recap: From Potential to Practice: How to Get Employees to Use AI
With warnings of an AI bubble looming and studies suggesting most generative AI pilots never make it into production, many executives are asking the same thing: We’re spending so much on AI—why aren’t we seeing real returns?
A major reason, experts say, is that employees still aren’t adopting AI at scale. To explore how companies can change that, The Information’s Kevin McLaughlin spoke with three leaders who focus on AI adoption:
- Alicia Lenart, VP, HR, Atlassian
- Brian Benedict, co-founder, Eliza
- Colleen Blake, COO, ServiceRocket
The Barriers to Adoption
One recent MIT study estimated that 95% of GenAI pilots fail to scale—a finding Alicia Lenart said Atlassian’s 2025 AI Collaboration Index: CXO Insights report echoed. It found that only 4% of executives report transformational improvements from AI. A core issue, Lenart argues, is companies focusing more on their tech stack than on the culture needed to support adoption.
“Having people do a training session for AI and hoping they’ll adopt AI and walk away is a fail,” she said. “You have to let people try and play.”
Brian Benedict, who helps large enterprises roll out applied AI, said many clients still treat AI as an add-on and overlook the restructuring meaningful implementation requires.
“Right now, people want to just bolt AI onto an old process,” he said. “What you really need to do is design your entire workflow.”
In his view, that means explicitly incorporating AI steps, checkpoints and automated handoffs into processes, rather than sprinkling prompts on top of legacy systems and hoping for magic.
The Elephant in the Room
No number of enterprise AI tools can overcome the biggest barrier to adoption: fear. Employees are less likely to experiment with a technology if they think it might replace them. Lenart said companies need to address that concern directly, and one of the most effective ways is for managers to demonstrate their own use of AI. In fact, Atlassian’s findings reveal that employees who see their managers model AI use are four times more likely to work with AI throughout the day.
“If managers are actually using AI and showing it to their employees, that goes a long way to dispelling the myth that AI is going to take jobs,” she said.
Colleen Blake, who sees similar hesitation among ServiceRocket’s staff, said companies should anchor AI discussions in concrete processes rather than abstract anxieties. Her team tackles this by identifying the weakest link in a workflow, then using AI to analyze bottlenecks and surface options, while humans decide what to do with them.
“AI can help you analyze the data and see the possibilities, but you need people to change the strategy,” she said.
From Experiment to Workflows
To move from isolated wins to companywide adoption, the panelists said leaders have to normalize experimentation, including the messy parts. Lenart said it’s just as important to surface failures as successes.
To help make this possible, Atlassian runs several dedicated Slack channels where employees post AI hacks, examples of where the tools fell short and requests for help. The company also has “AI champions” focused on leading and accelerating this cultural shift.
Blake said ServiceRocket takes a similar approach through “AI recipes,” where staff document their experiments—good and bad.
“It’s a great way for people to learn, and [it] encourages interaction and dialogue,” she said. “People can say, ‘Now what? How can we take this to the next level?’”
The Road to Agentic AI
For companies to see a real payoff from their AI investments, Benedict says employees need to graduate beyond chat windows.
“When you think of the transference from chat to API, that’s where a lot of companies kind of bridge the ‘Are you a power user or not?’ question,” he said.
To help people get there, Lenart said Atlassian leans on hackathons and structured experimentation, giving cross-functional teams dedicated time to prototype AI-powered workflows and agents. Success, she says, is more likely when people view AI as part of the team, not a threat to it.
“We feel strongly that it’s not going to be human or AI. It’s going to be human and AI,” said Lenart. “Maybe the number of AI agents will increase, but there’s still got to be humans in the loop.”