I did not think I would hear this question as often as I do. Yet it seems to be the number one question people ask of me, right after “so, what do you do?” and I tell them, “I work in IT strategy advisory and consulting,” or some variation of that pitch. What do you do, and then, what do you think about AI?
Back in the late 90s, the fad question was, “what’s your take on Y2K?” Then a few years later, “what do you think about the dotcom boom?” followed shortly by the dotcom burst. A few more years passed, and touchscreen phones without a keyboard were suddenly all the rage for discussion. “They’ll never displace a BlackBerry!” As Microsoft CEO Steve Ballmer was famously quoted, “$500 fully subsidized with a plan… that is the most expensive phone in the world and it doesn’t appeal to business customers because it doesn’t have a keyboard!” Even a few years later, iPads and social media became the technology flavors that everyone was talking about, while relatively few people really understood what they would eventually become.
It seems that every five to ten years we get another significant technology disruptor that everybody talks about, only a select few really understand, and somehow the things dismissed as fads often end up becoming completely normal parts of everyday life.
I think AI is going to follow a similar path.
The frenzy around AI certainly feels like a fad. Every company suddenly needs an AI strategy, every piece of software apparently needs an AI button, every executive presentation needs to mention it somewhere, and a surprising number of businesses seem determined to prove how innovative they are by finding work that a machine can do instead of a person. Eventually that excitement will settle down, just as it did with the internet, smartphones, cloud computing, social media, and nearly every other technology that was once going to “change everything.”
The technology will still be here when the excitement settles.
My concern is what companies do while we are waiting for that normalization to happen.
Some organizations are making significant workforce decisions based on an assumption that AI can simply replace people. Customer service teams are reduced, junior roles disappear, creative work gets automated, and productivity forecasts are built around the belief that fewer employees plus more AI must equal a more efficient company. The spreadsheet can make that argument look remarkably convincing.
Reality has been a little less cooperative.
Klarna became one of the better-known examples after aggressively promoting AI-powered customer service and reducing its workforce. By 2025, CEO Sebastian Siemiatkowski acknowledged that the cost-cutting approach had gone too far and began recruiting people again so customers could always reach a human being when necessary. This does not prove that AI failed at customer service; but it does demonstrate how quickly a technology strategy can become an operating-model problem when the human part of the equation is underestimated.
The economics are also considerably more complicated than “AI is cheaper than employees.” Gartner reported that at least half of generative AI projects had been abandoned after proof of concept by the end of 2025, citing poor data quality, inadequate risk controls, escalating costs, and unclear business value. Gartner has also warned that organizations that misunderstand how GenAI costs scale can miscalculate costs by 500% to 1,000%. (https://www.gartner.com/en/articles/genai-project-failure)
That is a very expensive error.
There is another problem that concerns me even more: when companies remove expertise, they may also remove the people best equipped to make AI useful. AI can generate an answer very quickly. Someone still needs to recognize whether that answer is good, or even truthful. That distinction becomes increasingly important as the work becomes more complex. Gartner has found that experienced employees in roles such as software engineering and law can receive some of the greatest productivity benefits from generative AI because they already know what good work looks like and can validate the output effectively. The expertise that looks expensive on a payroll spreadsheet may be exactly what allows the organization to use AI safely, intelligently, and productively.
I see the same principle in my own work. AI can help me research, explore an idea, challenge assumptions, summarize information, prototype software, or work through several possible approaches much faster than I could have ten years ago. Its value comes from combining that capability with decades of experience that tell me when the answer is wrong, when the recommendation does not make sense, when important context is missing, or when something technically possible would be a terrible business decision. Removing the experience and keeping the AI reverses that relationship.
There are already signs that people are growing tired of what happens when organizations automate the human contribution too aggressively. LinkedIn now gives users a “Seems like AI slop” option for posts and comments that feel generic, repetitive, or lacking in authentic perspective, and the company says more than one million members used it shortly after launch. LinkedIn has also adjusted its algorithms to reduce the distribution of content it classifies that way.
There is something wonderfully ironic about using AI to identify content that has too much AI in it.
The lesson for businesses should be larger than social media content, though. Customers still value judgment, experience, personality, empathy, context, and trust. Employees bring those things into processes every day, often without anyone formally documenting that they are doing it. Replacing the employee may remove far more organizational knowledge than the job description suggests.
That is why I believe the better opportunity for AI is performance improvement.
Start with the people you already have. Look at the work they do, the processes that frustrate them, the repetitive activities that consume their time, the information they struggle to find, and the decisions that take longer than they should. Then ask where AI can make those people faster, better informed, more consistent, or more capable.
That approach also fits a principle I have written about before in Strategy Before Technology: understand what the business is trying to accomplish before deciding which technology should be applied. AI deserves exactly the same discipline. A company should be able to explain the business outcome it expects before reorganizing itself around a technology simply because everyone else is talking about it.
AI will almost certainly eliminate some tasks, change some jobs, and eventually make certain roles unnecessary. But technology has been doing that for centuries. It will also create new work, increase what individuals are capable of accomplishing, and gradually disappear into everyday tools until we stop pointing at everything and proudly announcing that it contains AI.
That is probably when AI becomes truly valuable. The fad will eventually calm down. The tools will improve, the economics will become clearer, and companies will get much better at understanding where AI creates genuine value. In the meantime, organizations should be extremely careful about trading away years of human expertise for a technology whose greatest potential may actually depend on having that expertise sitting beside it.
Use AI to make your people better at what they do. Improve the process, remove the repetitive work, accelerate the research, strengthen the analysis, and give good employees better tools.
You may discover that the people you were planning to replace were the competitive advantage you needed AI to amplify.