Economists Mariana Mazzucato and Rosie Collington argue that consultants can, at best, give dubious guidance, and at worst, exacerbate government and private sector dysfunction. In their book The Big Con: How the Consulting Industry Weakens Our Businesses, Infantilizes Our Governments, and Warps Our Economies, the economists argue consultants emerged in a post-Ronald Reagan era of reduced regulations, necessitating third parties come in to save institutions who had lost faith in themselves.
Instead of righting the ship, Mazzacato and Collington argued, these consultants created just an “impression of value,” an illusion of helpfulness and little else, all while the government and private companies burned money to hire them.
In an era of AI, promising to save companies cash by automating white-collar jobs, the use of chatbots for guidance may be an appealing alternative to firms no longer willing or able to shell out for consultants. But emerging research shows that while you can ask AI what you would a consultant for a fraction of the price, its advice may not be worth taking, either. In fact, AI assistance might just present an old problem in a new medium.
A recent study led by the Esade Business School at the Universitat Ramon Llull in Barcelona, Spain, found that when various large language models (LLMs) were asked to provide guidance on a workplace issue, they gravitated toward a response that was most aligned with buzzwords, rather than providing guidance that best aligned with the scenario. Researchers dubbed the proclivity of AI to gravitate toward the same jargon to inform their judgements “trendslop.”
“An LLM is not the colleague who critically evaluates current ideas, looks into the contextual specifics, stress-tests assumptions, and pushes back when everyone gets comfortable,” the study authors wrote in a Harvard Business Review post summarizing their research. “On strategy, LLMs might be more akin to a freshly minted MBA or junior consultant, parroting what’s popular rather than what’s right for a particular situation.”
Recent layoffs among the “Big Four” consultancies, amid a wider industry slowdown, have suggested companies may already be losing value to potential clients. PwC slashed 150 business support staff in November 2025, around the same time McKinsey shed hundreds of jobs.
“As our firm marks its 100th year, we’re operating in a moment shaped by rapid advances in AI that are transforming business and society,” a McKinsey spokesperson told Bloomberg last year.
But the emergence of “trendslop” suggests AI is far from able to provide direction to companies seeking counsel from the technology, and this research exposes the bias LLMs struggle with.
How ‘trendslop’ manifests
In order to measure AI’s tendency to give responses aligning with trends rather than logic, researchers tested seven models, including GPT-5, Claude, Gemini, Grok, across 15,000 simulations and scenarios. Models were asked to choose between two solutions when presented with workplace tensions, such as if a company should prioritize long term versus short term growth, or if a firm should use technology to automate versus augment workers’ jobs.
Researchers predicted that if LLMs were providing advice based on the situation-specific details, there would be diversity in which solution the models choose. Instead, the seven models usually clustered their answers around the same strategy, indicating a preference for “modern managerial buzzwords and cultural tropes.”
Even when researchers reworded prompts or asked for pros-and-cons analysis, the AI models, in many cases, demonstrated a strong preference toward a similar business strategy. The study authors warn relying on AI as a consultant will not result in bespoke business solutions, but rather a cookie-cutter solution it could propose to any business when prompted, regardless of the specificities of a presented challenge.
“This reveals a real risk for leaders,” the researchers said. “An LLM can sound highly tailored to your situation while quietly steering you toward the same small cluster of modern managerial trends.”
Exposing LLM bias
The “trendslop” tendencies of LLMs are a result of biases they take on when the models are being trained, researchers noted. Because LLMs are trained on heaps of information from internet texts to social media to news, they tend to cling onto the positive or negative connotations attached to certain phrases or concepts, deeming “commoditization” as outdated and negative, and “augmentation” as progressive and positive.
In other words, when prompted to provide guidance on a tricky workplace scenario, AI isn’t analyzing the situation in question, it’s regurgitating key phrases based on how often it encountered while it was trained on data. In the case of ChatGPT, the study noted, the bot sometimes rejected providing a binary choice, instead recommending both solutions. Research published in Nature last year found AI sycophancy isn’t just unproductive, it can be harmful to science, confirming the biases of those prompting it instead of presenting users with data supported from scientific literature or other reliable, more impartial sources.
The “trendslop” researchers did not completely eschew the use of LLMs in navigating tricky workplace situations. They suggested models could still be helpful in generating alternative solutions or identifying blind plots in certain scenarios. If you’re aware of AI’s biases toward concepts like augmentation or long-term strategizing, you can challenge those biases to reveal more insightful guidance, according to the study.
“Leadership is ultimately about making hard choices in conditions of uncertainty and taking responsibility for them,” the researchers said. “AI cannot and should not be a substitute.”
