At a glance
- Source
- Dell’Acqua, McFowland, Mollick, Lifshitz-Assaf, Kellogg, Rajendran, Krayer, Candelon and Lakhani (2026). Navigating the Jagged Technological Frontier: Field Experimental Evidence of the Effects of Artificial Intelligence on Knowledge Worker Productivity and Quality. Organization Science, 37(2), 403–423. doi:10.1287/orsc.2025.21838
- How it was done
- An experiment registered in advance. After a baseline task, consultants were assigned by chance to work without AI, with GPT-4, or with GPT-4 plus a short introduction to prompting. Randomised experiment. People were assigned by chance to work with or without the tool, and the two groups were compared.
- Who took part
- 758 consultants at Boston Consulting Group.
- Where
- A global management consulting firm; realistic consulting tasks designed with the firm.
“AI assistance improves performance for some tasks but worsens it for others, even within the same knowledge workflow and with a seemingly similar level of difficulty.”
What it found
The authors call the edge of what AI can do a “jagged frontier”. Tasks that look equally hard can fall on different sides of it.
- Inside the frontier. On 18 realistic consulting tasks, from creative to analytical, consultants using AI completed 12.2% more tasks. They finished them 25.1% faster. The quality of their answers was significantly higher.
- Outside the frontier. On one complex management task, consultants using AI were 19 percentage points less likely to produce the correct solution than those working without it.
When the result holds
The gains held on tasks the AI model of the time could do well. The loss appeared on a task chosen to be beyond it. Tasks on the two sides can look equally hard, so which side a task falls on has to be tested.
Limits
- The tasks were realistic but were not live client work.
- Only one task was tested outside the frontier.
Practices that use this study
This study supports:
- Keep a person checking anything that goes to a customer or carries legal weight Before AI output leaves the company or is relied on in a decision, a named person checks it against real sources and signs it off. Make the check a step in the process, with time allowed for it. Do first · Effort: medium (weeks, a few people)
- For each kind of task, test whether a person with AI beats either one alone Do not assume that adding a person to AI, or AI to a person, gives the best result. Test it for each type of task before you settle the process. Do next · Effort: medium (weeks, a few people)