Work & judgement
GPT-4 helped plan a launch but hurt a brand decision
In a field experiment with consultants, GPT-4 helped with a product launch exercise. On a separate case, consultants with access to the same tool chose the right brand less often.
Consultants working on a footwear launch exercise completed more tasks when they had access to GPT-4. In a separate part of the same study, consultants given the tool were less likely to identify which of a company’s brands had the best prospects for growth. For a team making a similar choice, the study suggests testing whether the tool helps with each part of the job before deciding where to use it.
One tool, two kinds of work
Fabrizio Dell’Acqua and colleagues reported the experiment in Organization Science in 2026. The study included 758 Boston Consulting Group consultants. The researchers randomly assigned them to two separate sets of exercises. Within each set, participants first completed an assessment without AI and then were randomly assigned to work without AI, with GPT-4, or with GPT-4 and a briefing on how to prompt it. Both sets used the version of GPT-4 available in April 2023.
The footwear exercise included 18 tasks, from coming up with a product to defining the market and drafting launch materials. Participants with GPT-4 completed more of the tasks, and human graders rated their responses more highly. In the brand exercise, participants had to make a recommendation from interviews and financial data. About 85% of the group without AI got the brand right, compared with about 71% in the GPT-4 group and 60% in the group also given the prompting briefing. Even when the recommendation was wrong, graders gave the AI-assisted responses higher marks for coherence and persuasiveness.
The researchers had chosen the exercises after pretesting what the tool could do. The contrast shows that AI assistance can have opposite effects within the same profession. It gives no estimate of how often that will happen across all consulting work or with another model.
Review the choice before you polish the case
Suppose a community arts organisation wants to promote a series of events. Ask a colleague to decide which audience to target and record the booking figures and assumptions behind that choice. AI can help draft messages for the selected audience. If the same tool proposes switching the campaign to another group, send that recommendation back to the colleague who assessed the bookings. The wrong choice can spend the printing and advertising budget on people unlikely to book.
If you want to trial AI on the decision itself, score the recommended audience and the supporting figures separately from the writing. Use a past or synthetic campaign in which the relevant figures are known to the reviewer. Compare the recommendation with and without AI on equivalent cases. If the assisted choice does not hold up, keep the tool on draft preparation and leave the decision with the person accountable for the campaign.
Sources and further reading
- Dell’Acqua et al., Navigating the Jagged Technological Frontier, Organization Science (2026)
- Harvard Business School, published article PDF
Sources checked 5 November 2026. Prepared with AI assistance and editorial review.