Build with AI, not about AI
The strongest AI projects don't lead with the technology. They lead with the problem and let AI be the mechanism, not the headline.
There’s a subtle but important distinction in how organisations approach AI projects. Some build about AI. Others build with AI. The difference determines whether the project delivers lasting value or becomes a showcase demo that nobody uses.
Building about AI
Building about AI means the technology is the point. The goal is to “do something with AI” or “have an AI initiative.” The project starts with the capability and works backwards to find a use case.
This tends to produce demos. Proof-of-concepts that impress in a meeting room but can struggle to survive contact with real operations. They can be impressive technically but hard to embed practically.
You can spot these projects by their framing. They lead with “we built an AI that…” rather than “we solved X by…”
Building with AI
Building with AI starts from the other direction. There’s a real problem. A bottleneck, a quality gap, a process that doesn’t scale. AI happens to be the right mechanism for solving it.
The technology is invisible in the description of what the project does. “Proposals now take 20 minutes instead of 4 hours.” “Competitor analysis runs weekly instead of quarterly.” “New team members produce client-ready work in their first week.”
Most people don’t care that AI is involved. They care that the problem is solved.
Why the framing matters
When a project is built about AI, it often lacks a natural owner. It’s less likely that someone’s workflow depends on it, or that anyone notices when it stops working. It tends to live on the side of the business rather than inside it.
When a project is built with AI, it’s load-bearing. Someone relies on it every day. It’s woven into how work actually gets done. It survives leadership changes, budget reviews, and strategy pivots because it’s delivering value that people can feel.
How to tell which one you’re building
Three questions to ask before starting any AI project:
Who will use this tomorrow? If you can’t name a specific person and a specific task, you’re building a demo.
What happens if it breaks? If the answer is “nothing, we’ll just go back to how we did it before,” the project isn’t solving a real problem.
Can you describe the outcome without mentioning AI? If the value proposition only makes sense when you explain the technology behind it, you’re leading with the mechanism instead of the result. This connects directly to why so many AI pilots fail: the framing was wrong from the start.
The practical takeaway
Start with the workflow. Find the friction. Design the solution. If AI is the right tool for that solution, use it. If it’s not, don’t force it.
The strongest AI projects we’ve seen don’t feel like AI projects. They feel like operations that work better than they used to.