Nufar Gaspar

So here are two kind of expert craft, specific techniques from working with hundreds of executives that really get you from good to being exceptional in being able to get results with AI research.

So the first craft that I want to mention is wisdom of the craft.

I don't want you to ask one AI tool and then trust the results.

What I like to do is to send the exact same research across multiple models, sometimes even the same model or the same tool in several different sessions.

Then after you had multiple threads researching the same question with the various tools or the same tool in various instances, you aggregate and you aggregate on where they agree and you investigate where they diverge.

Then what I would strongly recommend that you do is that you use a separate model or a separate thread to fact check the aggregated result.

Because AI is much better at verifying than at generating correct research from the

So this is a weird phenomenon, but it works here as you add the data results.

And if there is a consensus across tools, it's likely factual.

From my experience, if you see 100% consensus, that's probably a true fact.

And if only one tool reports something, that's something that requires going deeper and researching again.

So that's also a cheat sheet on how not to have to validate each and every data point from AI by actually expanding the scope of the research and then narrow it down in an intelligent way.

The second thing that works very well is that before you act on any research or analysis output, whether it's external market research or internal data analysis, run it through these three questions.

The first one is grounded in real sources or is AI pattern matching?

The second one will be what's missing that I didn't think to ask.

And the last one, and that's the most important one, are you feeling comfortable putting your name to it if someone was asking for that?

And if you, even without good justification, intuitively don't feel comfortable to put your name to this, this means that you need to do some more work.

It takes about 30 seconds and it catches the majority of the AI research failures before they become bad decisions or a slop that you share with others.

And if you really want to take it even further, like a pro tip will be that the output shouldn't default to a wall of text or a boring bar chart.

You need to think with the modern tools, what's the easiest way for me or the consumers of this research to actually interact with the data, whether it's an interactive dashboard, an infographic, a reactive page that you can filter.