The default assumption about AI agents in the workplace is subtraction: fewer humans doing the same work. Doug Llewellyn, CEO of Data Society, is betting on a bigger company entirely, made up of a growing number of human employees alongside a growing number of agent “employees.” The workforce and survey data on AI adoption so far leans in his direction more than the shrinkage narrative does.
Most conversations about AI agents and jobs start from the same premise: agents replace tasks, tasks add up to roles, roles add up to headcount, so headcount should go down. Llewellyn starts somewhere else:
“I’m not sure the people won’t be employed. I think that the actual number of employees will be bigger. There’ll just be a bigger ratio of agentic employees to human employees. So I don’t necessarily think if you’re a company of 5,000 people right now, in 10 years you’ll be a company of 3,000 people. I think you could be a company of 10,000, and it’s 5,000 humans and 5,000 agents. That’s kind of how I look at it.”
Doug Llewellyn, Chief Executive Officer, Data Society
The distinction matters. A shrinking-headcount story and a changing-ratio story can look identical in a single department’s numbers, fewer people doing a given task, while adding up to completely different outcomes at the company level. One is a story about doing the same work with less. The other is a story about doing more work, total, with a workforce that now includes both kinds of employees.
The doom version of this conversation isn’t baseless. Change is real and uneven. The World Economic Forum’s Future of Jobs Report projects roughly 92 million jobs will be displaced by 2030 due to technology, demographic, and economic shifts. But the same report projects about 170 million new roles will be created over that period, a net gain of roughly 78 million jobs globally. The report also finds that nearly 40% of core workplace skills are expected to shift in that window, with 63% of employers already citing the resulting skills gap as a major barrier to transformation.
Net job growth alongside real displacement is a messier, less quotable story than “AI is coming for your job,” but it’s the one the data actually supports. It’s also closer to Llewellyn’s framing than either the optimist’s “nothing changes” or the pessimist’s “everyone gets replaced.”
Executive intentions point the same direction. JLL’s 2026 Future of Work Survey of more than 2,200 C-suite and corporate real estate leaders across 21 countries found 60% anticipate net headcount growth over the next three to five years, versus 40% expecting reduction. Sixty percent said they believe AI will reinvent human roles rather than replace them outright, and organizations further along in AI adoption were, if anything, more likely to keep hiring full-time employees and investing in entry-level talent, not less.
The “agentic employees” half of Llewellyn’s ratio is showing up fast, too. Microsoft has reported 15x year-over-year growth in active AI agents inside Microsoft 365, and 18x growth among large enterprises specifically. That’s the kind of number that makes the shrinkage story plausible on its face, more agents doing more work. But it’s happening at the same time LinkedIn’s 2026 Labor Market Report found employers created at least 1.3 million new AI-related job opportunities over two years, roles like AI engineers and data annotators that didn’t meaningfully exist before. Agents are scaling up. So, so far, is the human side of the ratio.
None of this guarantees Llewellyn’s specific numbers. Nobody knows yet whether a 5,000-person company becomes 10,000 total “employees,” 8,000, or something else, and the outcome will vary enormously by industry and role. What the data supports is his underlying claim: the more useful question isn’t whether AI agents will shrink your company. It’s what ratio of human to agentic work you’re actually planning for, and whether your organization is structured to manage both.
That second part is where most companies are furthest behind. Standing up 5,000 agents is a technology problem. Knowing which of those agents are still earning their keep, who’s accountable for what they do, and how a human workforce and an agent workforce actually coordinate is a governance problem, and it’s the one that determines whether “bigger ratio of agentic employees” turns into more output or just more sprawl.
Before betting on a bigger ratio of agentic employees, know what you’ll do to manage them. Data Society’s AI Advisory work helps organizations build the governance a growing agent workforce requires, and AI upskilling programs prepare human teams to work alongside it.
