Complex Insights in the AI Era

Good insights are harder to produce than most people think. We’re seeing a surge in tools promising faster, better insights, and many of them are genuinely powerful. Yet turning those capabilities into relevant insights is often less straightforward than it may seem. While AI can accelerate much of the process, it is not yet fully automated, and important nuances still require careful judgment. That’s where the human in the loop remains essential.

Based on my experience, three areas make the difference.

1. Creating a good prompt.
Even with good tools, this is still crucial. Many of the topics we work on are quite complex: they are layered, ambiguous, and highly industry-specific. Turning that into a clear line of inquiry is not easy, nor is it trivial. If the prompt is not well framed, outputs quickly become generic.

2. The sources.
Access to information is not the same as access to reliable information. As many have experienced across different tools, hallucinations, generic or misaligned inputs are common simply because the underlying sources are not on point.

And even when sources are strong, insights don’t come from aggregating them in a straight line. The best analysis comes from combining multiple angles: structured reports, blogs, articles, company analysis, and often less obvious signals that are harder to surface.

3. The interpretation.
This is where the human in the loop ties everything together: filtering what matters, connecting dots across sources, and most importantly, ensuring the output actually answers the underlying business question and is adapted to the team or individual asking it.

Don’t get me wrong, the tools available today are powerful and extremely helpful. But they don’t fully solve the hardest part yet: getting to truly relevant insights when the topic itself is complex.

And this is also the promising part. Tools are clearly moving in the right direction, but they are not quite there yet. If the challenges around framing, source quality, and interpretation can be addressed, the opportunity is not only to make research faster, but to make insight work significantly stronger in the future.