HBR Insights - LLM 'Trendslop' in Strategic Analysis

LLM tools are increasingly used by executives and their strategic advisors. This research demonstrates significant inherent bias in the major models.

HBR's March 2026 issue had a really interesting piece about the reality of using LLMs in strategic analysis. We all know intuitively that the models are biased for any number of reasons - training corpus, algorithms, even the CPUs used for training and inference. But intuition has limited use in guiding decision-making.

That's why we were so taken with this paper by Angelo Romasanta, Llewellyn D.W. Thomas, and Natalia Levina. They ran large-scale simulations against all the major models to test bias in evaluating binary choices that executives are faced with every day. It is one of the first benchmarks we found that steps outside of the technical realm and into the practical realm for 'the rest of us'.

The full article can be found here