Michael and I have been doing a deep dive into AI as we build out an AI-enabled ecosystem to help us optimize and scale our search practice, reduce administrative friction, and spend more time in the market facing off with clients and candidates. This WSJ article is worth a read. It highlights a practical reality of AI that’s going to matter a lot more to HR than people think.
https://www.wsj.com/tech/ai/what-is-inference-explaining-the-massive-new-shift-in-ai-computing-ed65a2fe
We’ve spent the last couple of years talking about AI in terms of what it can do. The more important story is where the money is going and how the cost of AI compares to traditional human labor. AI used to be about training models, which required a big upfront investment. Once that was done, the cost largely stabilized. Now the focus is on inference, which is what happens every time the model is actually used.
Every prompt, every workflow, every automated task creates an ongoing cost that scales with usage. The more companies rely on AI to do real work, the more they’re effectively paying for it in a way that starts to resemble labor. At a certain point, this stops being a technology conversation and starts becoming a workforce conversation. If a system is producing output continuously, tied directly to business activity, and showing up as a recurring cost, it begins to look a lot like capacity for which you would have otherwise hired.
We’re not fully at a place where AI replaces payroll. But we are getting closer to a world where companies are managing two parallel cost structures: human labor and machine generated output. That has real implications for how we think about hiring, org design, and productivity. As companies start deploying AI at scale, they’re going to need a much clearer understanding of the economics behind it. Not just what they’re spending on the technology, but what human labor it’s actually replacing or augmenting. It’s one thing to invest in AI. It’s another to prove that it’s delivering the same or better output, more efficiently, and at a lower cost.
At some point, this becomes a question of accountability. If AI is being positioned as a substitute for capacity, companies need to be able to demonstrate that tradeoff clearly. Otherwise, you’re not driving efficiency, you’re just layering cost on top of cost.