It’s Time To Commoditize AI

I am tired of hearing business leaders tell me that they are just going to give their people access to AI tools and allow them to make decisions on how to leverage it for effectiveness. That is not an AI strategy, and it is not the way to build a culture that is technology forward and aligned. 

The problem is not the decentralization or the individual freedom. In fact those are the most positive features of this otherwise severely flawed non-strategy. The problem is simply this: Highly skilled executives, deal makers and relationship builders don’t do operations and tool interface. We have seen this phenomenon for decades in the sales world where the top performing sellers are chased from here till Tuesday to enter their data and deals in Salesforce, and they never do. However, at the end of the day, organizations put up with it because why? Well, because they are the top performing sellers. 

This is not an outlier or an idiosyncratic case. People work in different ways, and AI is no different than other past technologies. I can hear the arguments now, “AI is different”, “With AI, you don’t need to interface in an intensive way and it does the thinking for you”. Wrong. The same could have been said when SaaS tools made a splash on the scene 3 decades ago. Their whole appeal was how much easier they made interfacing with technology and data, and still, the operators used them, and the executives didn’t. Your executives and relationship builders don’t want to think about workflows and interact with tools, no matter how easy it may seem to the technologists and operators. And even if you find a magical unicorn executive who wants to, they won’t be very good at it. And even if you find the one-in-a-million super mega unicorn who is a true rainmaker, and who likes interacting with tools, and who is good at it. Why would you ever want to pay them to spend their time doing that? The answer is, you don’t, you want to pay them to do what they do best: to make rain.

If the above confused you in any way, let me clarify. I am a techno-optimist. What AI can do is mind boggling. The capabilities are astounding, but the outputs are only incrementally better than the last generation of tools we have all had access to for 25 years. Take this conversation I recently had with a start-up CEO in the media space for illustration purposes (paraphrased for brevity, clarity):

What’s the point? The point is that AI is incredible at what it does well, and it is not good at what it is not good at. We, as consumers of AI business tools, need to stop trying to squeeze all the value out of AI that we possibly can, and we need to define a set of functions that AI can reliably do well and then, and here is the important part, we need to define what our people do well and set up workflows and systems that are symbiotic with one another. I am not guessing at this. I am actively experiencing how well this can work, and I have seen it with AI tools and with other non AI tools. No difference. Same concept. 

If we try to work with AI at the margins of where it is valuable, we put ourselves in a position where the lack of reliability and variance of outputs causes our people to have to implement systems that overshadow efficiency gains and work counterproductively to any benefits we might have gained. Simply put, when AI works with no human in the loop and produces an output, a human has to make sure that output is ready for the real world, and if they were not in the loop, the amount of work it takes to understand how the output was produced and get in a position to validate or correct costs time and money, oftentimes more than was gained by employing the tool in the first place. This, done at scale, across an organization causes, untold efficiency losses, not gains.

AI is great at research, analysis, and data and insight surfacing. Executives and relationship builders are great at growing relationships, executive function and decision making, especially involving tough trade-offs, assuming responsibility, assuming risk. And then, the key. We need an entire new set of job descriptions and role definitions; not new people, but slightly changed role definitions for our operators. We need people to translate information from the rainmakers and executives so that the tools have the right inputs, and we need people to interact with the tools to get the right outputs and surface them to the rainmakers and executives so that they can spend more of their time doing what we pay them to do. These are the same engineers, administrators, project managers, and operators we have always had, but refocused on internal alignment between tool workflows and people workflows. This is a three layer organization with AI agents on the organizational chart enabling relationship builders and executives to perform and the operator layer actually making both sets of work possible. 

But, this can only become a reality if business leaders commoditize and specialize AI on their own. Don’t buy the cutting edge vision of what is possible. Don’t buy the marginal capabilities that we supposedly revolutionize the way work gets done. Simply, understand the functions that need to be accomplished in your organization. Be very clear about what AI is good at already, and buy specialized and proven tools that can do that work reliably and well. Then build people workflows that enhance those tool workflows and allow the tool workflows to enhance your people.