A Looming AI-Enabled Performance Divide

Why Business Leaders Must Prioritize Tool Curation and Workforce Training to Avoid Falling Behind in 2025.

In the latter part of 2024, a practical consensus on the present – if not future – state of generative AI capabilities emerged. It held that the tools had advanced so quickly while the adoption of those tools in the workforce lagged so far behind that even if capabilities never pushed beyond the state of the art circa October of 2024, the impact on the economy would be profound as firms eventually caught up to a future of work that had been decided some time ago.

That late–2024 consensus – and the data backing it up – has been useful in that it’s at once grounding and motivating. It can pull the focus of AI-curious entrepreneurs and executives away from arcane and ever-shifting debates about AGI timelines (surely important – but mostly for AI researchers and the governments that should be crafting some kind of meaningful AI regulations) and back to the urgent and practical questions of tool curation, upskilling & training, and rethinking productivity and possibility within their organizations.

That’s the real opportunity and the clear imperative for leaders in 2025 – and likely beyond. But I worry that the clarity of that imperative may be dulled by some of the narratives about how we’ll collectively catch up to the AI-enabled future.

Describing what he’s called a great normalization, analyst Azeem Azhar envisions that “widespread adoption and practical integration into every aspect of work and personal life will accelerate, reshaping productivity patterns and business strategies across sectors.”

Even further smoothing the processes of adoption and integration at scale, the opening essay in the recent Tech Trends Report from Deloitte suggests that

“We won’t ‘use’ AI. We’ll just experience a world where things work smarter, faster, and more intuitively — like magic, but grounded in algorithms. We expect that it will provide a foundation for business and personal growth while also adapting and sustaining itself over time.”

Maybe. Eventually. There’s definitely validity to the idea that exponential technologies become not only more powerful and cheaper but also – and crucially – easier to use over time. That was obviously true with the history of computation, and it’s already been true at least once for Gen AI (with ChatGPT being the critical “interface moment”).

But what if AI tools are becoming more powerful much more quickly than they are becoming easy to use? And what if the market is so flush with wave after wave of AI tools that many users are overwhelmed by choice and unwilling to invest in developing real proficiency with a tool that might be obsolete in a few months?

The near-term picture of adoption and impact is likely to be more uneven and generally more fitful than many reports suggest. And accordingly, the productivity/performance gains are likely to be unevenly distributed even among workers and firms who are attempting to leverage the new tools. A new pre-print paper from researchers at Berkeley and Harvard found that the performance impact of Gen AI tools on a group of entrepreneurs ranged from +15% to –8%, painting a picture that doesn’t look like a rising tide lifting all boats & such but rather, a storm surge where some would surf and others struggle to stay afloat. So to speak.

What that suggests to me for 2025 is the prospect of a deep AI-enabled performance divide developing while the market continues to churn through tools and releases along the way to some future state of consolidation. And the threat of being caught on the wrong side of that divide should absolutely underscore the urgency for business leaders to get serious about tool curation and training within their organizations.

@Jeffrey