The Jagged Frontier

• AI can excel at a task that seems hard and fail at one that seems easy.
• People can’t see where the edge is, so they carry trust from tasks inside it to tasks outside it.
• Inside the frontier, AI help improves speed and quality. Outside it, AI help can make people worse.
• Map the frontier for your own tasks with direct tests. Don’t infer it from general benchmarks.

Human skill tends to be smooth: someone who can write a strong strategy memo can usually do the simpler tasks around it. AI capability is jagged. The same model can draft a solid analysis and then botch a detail any junior analyst would catch. General benchmark scores flatten this into one number, which makes them a poor guide to which of your tasks are safe to hand off.

Field experiment with consultants. Dell’Acqua et al. (2023) ran an experiment with 758 Boston Consulting Group consultants on 18 realistic tasks. On tasks inside the AI’s frontier, consultants using GPT-4 completed 12.2% more tasks, worked 25.1% faster, and produced results rated more than 40% higher in quality. On a task outside the frontier, consultants using AI were 19 percentage points less likely to produce correct solutions than those without it. The study was later published in Organization Science (Dell’Acqua et al., 2026).

Task type matters. Vaccaro, Almaatouq, and Malone (2024) found human-AI combinations tended to lose ground on decision tasks and gain on creation tasks, another sign that value depends heavily on which task you’re looking at.

An older version of the idea. Hans Moravec (1988) observed that it’s comparatively easy to get computers to perform well on things like intelligence tests and board games, and hard to give them the perception and mobility skills of a one-year-old.

1. List the tasks in your workflow and test the AI on each one directly, including the ones that seem trivial.
2. Share the map with users: where to rely on the AI and where to double-check.
3. Re-map after model updates. The frontier moves.
4. Design interfaces that make uncertainty visible near the edges.

For vendor reviews, model cards, and launch reviews.

• Were tasks that are easy for people tested, not only hard ones?
• Where are the known failure pockets, and are users told about them?

The headline numbers come from one experiment with consultants on one set of tasks, and the frontier moves as models change. Treat it as a reason to test task by task, not as a fixed map.

The term comes from Fabrizio Dell’Acqua, Ethan Mollick, Karim Lakhani, and colleagues’ field experiment, first released as a Harvard Business School working paper in 2023.

1. [1]
   Dell’Acqua, F., McFowland III, E., Mollick, E., Lifshitz, H., Kellogg, K., Rajendran, S., Krayer, L., Candelon, F., & Lakhani, K. R. (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)
   Open ↗ (opens in a new tab)
2. [2]
   Dell’Acqua, F., McFowland III, E., Mollick, E., Lifshitz-Assaf, H., Kellogg, K., Rajendran, S., Krayer, L., Candelon, F., & Lakhani, K. R. (2023). Navigating the jagged technological frontier: Field experimental evidence of the effects of AI on knowledge worker productivity and quality.Harvard Business School Working Paper 24-013 Working paper
   Open ↗ (opens in a new tab)
3. [3]
   Moravec, H. (1988). Mind children: The future of robot and human intelligence.Harvard University Press Book
   No link
4. [4]
   Vaccaro, M., Almaatouq, A., & Malone, T. (2024). When combinations of humans and AI are useful: A systematic review and meta-analysis.Nature Human Behaviour, 8, 2293-2303
   Open ↗ (opens in a new tab)

• v1.05 Oct 2026Published.

Full changelog · Spot a mistake or a better source? Report it. Corrections are logged in the changelog.