Announcing the Artificiality Summit 2025!
Announcing the Artificiality Summit 2025! Don't miss the super early bird special at the end of the email...
Our World of Workflows research discovers that the benefits of AI support are not evenly distributed but rather significantly skewed toward businesses and entrepreneurs that are already succeeding.
Few things are more challenging than being an entrepreneur. They find themselves wearing multiple hats, from marketing to finance, customer service to recruiting, and technical solutions to strategic vision. AI promises a personal, real-time coach, an expert in every conceivable domain, accessible at any hour. Early studies show that AI can be extremely effective as a proficiency booster, especially for novices. But they also show that those same novices make bigger mistakes with AI because they can't readily judge the quality of AI answers. And most studies so far have been constructed to be within specific narrow boundaries. What we all want to know is how effective is AI advice when it's delivered "in the wild", dealing with real business problems in real time?
A recent study conducted by researchers from Berkeley Haas and Harvard Business School shows that expert AI assistance is uneven. The research anticipated a universal improvement in entrepreneurial outcomes, but the results revealed a more different picture: the benefits of AI support are not evenly distributed but rather significantly skewed toward businesses and entrepreneurs that are already succeeding.
Could it be that the direction of influence is reversed? High-performing entrepreneurs might be asking simpler, more direct questions to AI, focusing on refining and optimizing strategies that have already proven successful. These queries, which are more focused, might be where AI excels. More successful entrepreneurs may also be applying more sophisticated judgment to the AI’s output.
Conversely, struggling entrepreneurs might have more complex, foundational challenges. When they turn to AI for advice, insights, or solutions that could turn their fortunes around their queries might surpass the current capabilities of AI. Or anyone, for that matter.
The study merges the success of the entrepreneur with that of their business, operating under the assumption that the health of the business is a direct reflection of the entrepreneur's performance. This is inferred from the nature of the questions posed to the AI. Entrepreneurs of thriving businesses tend to ask straightforward questions, indicative of fewer, less complex challenges. In contrast, those leading less successful ventures pose more difficult questions, hinting at deeper, systemic issues. This methodology implicitly links the entrepreneur's prowess directly with their business's outcomes, without separately evaluating their personal success metrics outside the immediate context.
In the broader context of productivity studies, this research adds to the discussion about the definition of "low performing" and its implications. The term can encompass a wide range of scenarios, from novices who lack experience to seasoned professionals who face difficult obstacles. This diversity in interpretation makes it confusing for leaders who want a simple answer to the question: how will AI make my people more productive?
This conflation of terms—ranging from inexperience to outright ineffectiveness—complicates the interpretation of productivity research, as it fails to distinguish between the various stages of professional development and the diverse challenges each group faces.
There's a lesson here on a critical aspect of AI integration in business: the technology's effectiveness is contingent not just on its advanced capabilities but also on the specificity and clarity of the questions it's asked. This study suggests that the real kicker might well be a combination of level of skills retried to handle the complexity of the problem and skill and experience of the human.
The Artificiality Weekend Briefing: About AI, Not Written by AI