“We should do something with AI.” It is the sentence that launches a thousand stalled initiatives. It feels like ambition. In practice it is the surest way to spend money and months and end up exactly where you started, because it skips the one step that actually predicts whether AI works: starting from a real problem.
What the research found about order
My 2014 doctoral work was a qualitative descriptive multiple-case study built on in-depth interviews with 19 people across adopting and resisting Fortune 500 organizations. One of the clearest patterns in who actually moved came down to communication, and specifically to whether a leader recognized a real problem before reaching for a tool. The leaders who adopted tended to see that their communication was fragmented across too many disconnected systems, named that as a concrete failure they were living with, and adopted a unified platform as the fix for that specific pain. Recognizing a genuine problem and acting on it was one of the conditions most associated with adoption.
The leaders who could not see the fragmentation, who had no named problem driving them, had nothing for adoption to attach to. A tool with no problem behind it is just a cost. The order was not a small detail in the data. It was a dividing line. Pain first, tool second, was a marker of the adopting case. Tool first, problem never, was a marker of the resisting one.
The same trap, now with AI
AI makes this worse, because the pressure to do something with AI is louder than anything social platforms ever generated. So leaders buy the tools, announce the initiative, and then watch it quietly die, because it was never attached to a problem anyone actually felt. The numbers bear out how common this is. McKinsey found that 88 percent of organizations now use AI, but only 7 percent have scaled it, and the work most likely to produce real return, redesigning a workflow around a specific problem, has been done by only about one in five (McKinsey, 2025). Most organizations bought the tool and skipped the problem.
Boston Consulting Group’s research underscores why problem-first wins. The companies generating real value concentrate on a few high-priority problems and go deep, while the laggards scatter across many shallow use cases, and BCG found that leaders pursue roughly half as many initiatives as their less successful peers while getting far more out of them (BCG, 2024). Fewer problems, chosen well, beat more tools deployed widely.
The businesses getting real value do the opposite of scattering. They start from the thing that is visibly broken: the follow-up that gets forgotten, the quote rebuilt by hand on every job, the booking that slips through because nobody had time to confirm it. Then they build the system that removes that specific pain. The technology is the last decision, not the first. This is one of the five forces in the 5 drivers of AI adoption.
What this looks like in real work
That is exactly how the work goes when I build for a client. A youth riding school was losing hours every week stitching bookings, waivers, and payments together by hand, with no reliable way to even tell which family was which. The problem came first. We built one flow that handles intake, signed waivers, payment, and scheduling together, and the hand-stitching stopped. A roofing company had a different problem: after a software change, their estimating system had lost its connection to pricing, so the team was rebuilding every quote by hand. Again the problem came first. We traced it to the real cause and fixed the system underneath so quotes pulled correctly on their own.
Neither of those engagements started with “let us use AI.” Both started with a specific, expensive breakdown that someone could describe in a sentence. The technology was chosen to serve the fix, which is the only order that reliably works. You can read the roofing example in full in how a roofing company stopped rebuilding every quote by hand.
The move
Before you evaluate a single tool, find your most expensive recurring breakdown, the one that quietly eats hours every week or loses customers you never hear about, and write it down in one plain sentence. That sentence is your AI project. Everything else is shopping. If you cannot write the sentence yet, that is the first thing to fix, not the tool.
To find the breakdown worth solving first, the Omnine AI Readiness Assessment takes about three minutes and points you at the problem with the most leverage in your business.
References
Goodwin, M. R. (2014). A qualitative descriptive multiple-case study: Fortune 500 leaders’ social business platform adoption (Doctoral dissertation, University of Phoenix). ProQuest Dissertations Publishing (UMI No. 3648813).
McKinsey & Company. (2025). The state of AI in 2025: Agents, innovation, and transformation. McKinsey Global Survey on the state of AI.
Boston Consulting Group. (2024). Where’s the value in AI?