Blindspot

What we know so far about how Agentic AI will impact the labor market, and where we're exposed

The labor market is already changing due to Generative AI and Agentic AI is widely expected to accelerate that change, but how much—and how soon? If you’ve been following The Agentic Era series, you know that this is the question I’ve been wrestling with since Anthropic ushered in the Agentic Era with the release of Claude Computer Use in late October. And judging by the questions I’ve received from readers, it’s the number one thing on your minds, too.

Agentic AI—the kind that doesn’t just generate content but can autonomously execute business processes—has a different-in-kind impact on the labor market from the Generative AI that preceded it. Agents (especially General Purpose Agents) dramatically reduce the barrier to automation, shifting the requirement from an engineering team to just a human-language description of a process to be automated. This means that any knowledge worker who understands a workflow can automate it, which in turn could accelerate business process automation at an unprecedented pace.

And yet, after months of scouring the latest labor market projections, it has become clear to me that we have a blindspot.

The Diverging Predictions

There are three major "labor market expectation" camps forming, and they talk about their expectations in very different ways.?

Sector 1: Economists and Researchers

Here’s a quick recap of the most prominent labor market projections available today:

• McKinsey Global Institute and OpenAI’s came out early with?the first large-scale, economy-wide bottom-up analyses back in 2023 and 2024, but neither has been updated recently and neither explicitly explains whether their analyses consider automation beyond what Generative AI alone made possible. An update here would be valuable, as these were rigorously composed bottom-up analyses.
• Daren Acemoglu (2024 Nobel Economics Laureate) has estimated that only 5% of worker tasks will be able to be profitably automated. But this projection haircuts from the first versions of the OpenAI and McKinsey research released in 2023. Stanford's Erik Brynjolfsson has a somewhat more optimistic expectation for the pace of GenAI uptake, with some high-quality projections released in 2024, but assumptions around for process automation due to Agentic AI aren't explicitly called out from GenAI.
• The IMF projected in January of 2024 that 40% of global jobs could be affected by AI but that estimate simply tallied the amount of knowledge work jobs (and didn't predict actual uptake of process automation).?
• The U.S. Department of Labor’s projections have changed for certain occupations since ChatGPT launched — but for the most part those have been limited to directly-AI-related occupations, which seems an underestimation.
• Two notable recent projections: The World Economic Forum’s Future of Jobs Report and LinkedIn’s Work Change Report are the most recent labor market-wide analyses (both January 2025), but again neither mentions Agentic AI explicitly but both are still definitely full of insights and worth a read and re-share. The WEF report, for example, is based on employer surveys conducted six months ago—before Agentic AI was widely recognized as imminent -- but still projected?job creation and destruction due to structural labour-market transformation will amount to 22% of today’s total jobs (on a base of 1.2B formal jobs), and?two-fifths (39%) of existing worker skill sets will be transformed or become outdated by 2030.?Excellent analyses, but they may already deserve an update.
• And then a fairly large amount of commentary from economists and researchers like the Harvard Colloquy podcast interview of David Deming.?

These reports are valuable, but they almost never explicitly model the impact of business process automation enabled by Agentic AI, which leaves us to wonder if the Agentic Era is already baked in, or if the economists are literally struggling to update their forecasts fast enough to reflect the pace of technological change.

Sector 2: Tech and Large Employers

At CES last month NVIDIA’s Jensen Huang gave a talk explaining that IT will ‘become the HR of AI agents’, a sentiment I heard echoed by many large enterprise CIOs at the Cisco AI Summit in Palo Alto last month. Although Jensen's take is a dramatic one, leaders at Microsoft, Google, OpenAI, and virtually every other major technology company are on record proposing that Agentic AI (specifically) are poised to?unlock massive productivity gains, create entirely new job categories, and accelerate human potential.

McKinsey published a special report on Agents back in July of 2024, Deloitte issued a report this past November predicting that 1/4 of large enterprises would field AI Agents in 2025, and MGI double-dipped with a report on Superagency just a couple of weeks ago (more on Superagency later in this article -- but just know that it's a great book I strongly recommend reading) but each of them stopped short of predicting the effect on jobs or labor skills.

And in my experience talking to dozens of CIOs in recent weeks, I've found that those more ambitious projections from Tech leaders and consulting firms are matching what I'm hearing from Business Leaders. Executives appear to be fully committed to expanded business process automation but expect it to take some number of years to fully integrate into an enterprise. They cite concerns around safety, reliability, access controls, and the need for comprehensive staff training before widespread adoption. Some of them lament that small startups will be faster at navigating this change, giving them a temporary edge. I suspect if WEF ran their employer survey today, this would emerge as the main message.

Sector 3: The Education and Workforce Development Sector

I've found that for the most part the education and workforce development sector is focused on building basic LLM Literacy and digesting the potential of Generative AI. As a sector, we've largely internalized that upskilling is needed --?JFF's recent publications are a great example of this.?But with few notable exceptions, the discussion among the majority of education and workforce development institutions hasn't advanced to really focus in on Agentic AI.?

Alignment between these three sectors is rather important -- it drives our ability to advocate for policies and practices that will build the workforce we need to create great new jobs.

The Gap in AI Labor Market Projections

The vast majority of labor market projections that account for AI were published in late 2023 or early 2024. That means they’re based on research conducted even earlier—before the rise of truly Agentic AI. These reports often name generative AI as a factor, but I have yet to find a single labor-market-wide projection that explicitly reflects Agentic AI automation.

This leaves us with a fundamental question:

• Did these reports already assume that business process automation would accelerate beyond content creation? Or,
• Have they not yet reflected a massive shift that could redefine the timeline and scale of AI-driven labor market disruption?

Either way, the lack of alignment around expectations is a gap that makes it harder to align the three sectors around a shared view of what's to come and how to make the outcome great.?

This is our blindspot.

The Consequences of Waiting Too Long

This gap in research isn’t just an academic issue—it has real-world consequences. AI-driven labor market shifts are already underway. Studies have shown that GenAI led to a 21% decrease in job posts for automation-prone jobs related to writing and coding in online freelancing platforms as far back as mid-2023.?I'm already hearing reports of developers and startups rapidly integrating Agentic AI into their workflows, and planning to grow on agentic workforces.

But the outcome could be truly wonderful if we lean in and leverage this technology to improve jobs and build new jobs. One of the best fresh takes on this is the book Superagency, by Reid Hoffman, who makes the case for how this could all go right -- it's a good read, and one that highlights the opportunity if we upskill. In fact, at this point, it looks like virtually everyone agrees that upskilling is needed. But how much? And how fast? And without robust projections, policymakers, businesses, and educators won't be able to come to terms with how rapidly we need to start this upskilling.

What Needs to Happen Next

1. New Projections: We need updated labor market analyses that explicitly account for Agentic AI’s role in business process automation.
2. Transparency in Assumptions: Future reports should clearly state the extent and type of automation they assume, particularly in knowledge work.
3. Urgency in Action:?Upskilling is already a broad consensus solution, but the pace with which institutions are moving to build AI skilling programs is highly inconsistent, and largely behind what industry wants to see. We should either come to alignment on the pace of change to expect, or play it safe by leaning into upskilling more decisively.

The Bottom Line

This cannot wait. We must close the blindspot.?

?? If you know of any projections that do explicitly consider Agentic AI, drop them in the comments.

?? If you know a researcher or economist, send them this article. If you are a researcher or economist, let’s work together to get this right, and I can help get your forecasts in front of the tech/employer and education/workforce ecosystems as quickly as possible to drive discussion and consideration.

?? If you’re an educator, employer, or workforce leader, start planning for an agentic future now, and tell your peers and colleagues to subscribe to The Agentic Era for help navigating the change to come.