This whole AI thing is simpler than you think

At least three-quarters of the speaking invitations I get these days are about AI. But lately, they’re for different reasons. Companies used to bring me in under the assumption that artificial intelligence was going to change everything. So they’d ask me to talk about the jobless future, prompt engineering, or automating marketing online. Today, they’re asking a different kind of question: What went wrong? Where are the promised productivity gains? In other words, why isn’t AI helping our company do stuff? 

And if I were to answer honestly, I’d tell them the simple truth: It’s because you and your people don’t know what you want to do with it! This is not a technology problem or even a people problem, but an organizational one. Most companies are trying to run 21st-century AI-enabled companies on 20th-century industrial age architectures. The goals and the values are simply not aligned. 

Weak returns

The failure of AI to increase productivity isn’t just cherry-picked anecdote or idle speculation. According to Microsoft’s own study, only 25% of AI initiatives achieved expected returns over the past three years. Computer scientist Gary Marcus has been writing about the hype and misplaced hope around large language models (LLMs) for just as long. But the problem here is not that AI can’t do great stuff; it’s that organizations are not yet thinking in a human-first way about technology. 

It’s what Chris Perry, my partner at Andus, calls “the Human OS.” Companies are trying to integrate AI into their existing workflows without considering the human systems that can not only utilize it, but also grow and improve with it. This means more than training workers on AI interfaces or running trials of AI-derived ads. It means remaking the very structure of your organization around these new opportunities. 

It’s a transition less technologically focused than the term OS may imply. You don’t reprogram your workforce around the needs of AI. Rather, you develop a different kind of workplace. Who thought up how a bank works, with a safe in the back and tellers behind special windows? It was a way of creating an interface between lenders and borrowers in a new economic system. Now it seems obvious. Or take the grocery store, with aisles, shopping carts, and cashiers. It is an operating system designed to allow humans to browse, gather, purchase, or even socialize (in the fruit section). These are architectures for human interaction under capitalism and then industrial production. And they worked, largely because the people designing them were taking the human consumers into mind.  

Lessons from the Industrial Age 

We don’t generally devise our operating systems for labor with the same considerations for the people trying to function within them. Industrial age values of efficiency and productivity led to the assembly line, the sweatshop, and the typing pool. These are environments in which humans are cogs in a machine, and the more repeatable and uniform their tasks, the better. 

If anything, the values of the industrial age were to remove human beings from the value equation altogether. Assembly lines allowed for unskilled, replaceable workers to replace higher-paid craftspeople. New machines were always valued for their ability to replace labor, or at least allow labor to be outsourced to less expensive labor in Asia. 

This approach to new technology won’t work in the age of AI. And even talking this way will make workers less likely to consider the possibility that you are trying to do anything other than replace them with this stuff. You are trying to augment instead of replace them, right? I sure hope so, because the alternative—a world where you are using AIs instead of people—means your only competitive advantage is the size of your contract with the AI company. 

Moving from “more’ to ‘better”

Assuming you are on Team Human here, then the object of the game is to help your people come to terms with how their own capabilities can be enhanced with this stuff. Sure, they can churn out more PowerPoints or spreadsheets in an hour, but that’s not the enhancement on offer here. The possibility is for them to do categorically better, more considered, and more developed work. Instead of using AI to create more volume of what already is (industrial age efficiency), what about using it to help imagine what does not yet exist? 

Again, this is the opposite of industrial age repeatability and workforce reduction. We have that one down. It’s time to move on from “more” to “better.” But to do that, your company needs to retrieve its core competencies, rediscover its values, and reclaim its culture. Oddly enough, this is the real AI opportunity: to double down on the human creativity driving your enterprise. It’s the very opposite of what we’ve been doing with technology until recently. 

But this means building a different kind of human readiness into the very architecture of your organization. You can buy a bunch of AI capabilities, but you need to match them with human processes that can benefit from them in real and sustainable ways over time. It will take more than one article to explain how that works, because it covers everything from talent development and workflows to rituals and incentives. 

For now, think of it this way: Machines process data fast and accurately. Humans don’t process so much as respirate and metabolize. In a successful organization, these two functions—processing and metabolizing—can support each other. The way to get there is not simply to buy more AI capability and work on training, but to engage with your people in a way in which they know what they want to do better. They must understand that their nervous systems and sensibilities are going to be valued and trusted as they explore uncharted territory, and that their explorations are being integrated into the company’s institutional memory—whatever their apparent contribution to growth or efficiency. 

View the full article