AI is incredibly powerful and it is relatively easy to add a rudimentary integration to new and existing software. It’s easy to get caught up in the hype and see every problem as solvable with an AI hammer.

But just because you can, doesn’t always mean you should.

Don’t get me wrong, I use AI-augmented tools every day and am amazed at what they do for my productivity. I also create AI-augmented features in the software that I build.

However, if I reach for the LLM AI “hammer” first, I bypass the opportunity to achieve better results and user experience. By focusing on the root problem at hand and structuring my data a bit better, I could nrgate the need for AI and achieve better outcomes.

For example, if your software needs to match job seekers with job specs, you could reach for the AI hammer to do the work, but you don’t need to. Why? Because AI yields “poorer” quality results than getting human assistance breaking down the constituent parts of a user profile and the parts of the job spec into structured data, matching these structured pieces, and human oversight to make the final judgement on a ‘good’ match.

For example, if your software needs to generate a list of similar job titles to one listed in a job post, you could reach for the AI hammer to do the work, but you don’t need to. Why? Because it might be cheaper, quicker and yield adequate results using existing databases like the US O*Net database of careers and salaries.

The point is, AI can do many things, but it’s not a panacea. You might find you get better results by exploring the root user problem and structuring new or existing data to solve the problem more accurately.