Most work doesn’t need the Ferrari of AI, Wharton and Accenture find, but companies may be paying for it anyway
Reported by Crypto Briefing and Fortune2 sources ↓
A Wharton-Accenture report finds that more than 90% of tasks AI agents could take on are routine enough for less expensive models. A Wharton co-author says companies may be using their most sophisticated model across entire workflows.
Key points
- The study assessed more than 18,000 tasks and work activities across 923 occupations
- Only 6% to 11% of agent-capable work in each of 19 industries was typically complex enough for frontier models
- Estimated five-year costs for 100,000 employees using 35 agents ranged from about $155 million in health care to $520 million in insurance
Sources
Details
The researchers filtered tasks by the human interaction and physical activity involved, leaving 11,339 for assessment. They rated complexity using factors including interdependent steps, legal or audit requirements, systems involved, sensitive data and data volume.
The report says AI computing is a major cost in its industry estimates. Its authors recommend matching models to tasks during agent design; they say dynamic routing between models is difficult to track in live operations and few enterprises can use it at scale.
Written from coverage as of
Images
- Fortune
Stocks in this story
Change since open: 20 Nov to 21 Nov 2026
Last refresh