Most work doesn’t need the Ferrari of AI, Wharton and Accenture find, but companies may be paying for it anyway

What happened
A new report from the Wharton School and Accenture, released Monday, analyzed over 18,000 work tasks across 923 occupations in 19 industries to assess which tasks are suitable for automation by AI. The study found that more than 90% of tasks that could be automated by AI are routine and do not require the use of the most advanced AI models. Despite this, many companies are reportedly using the most sophisticated AI models for all tasks, regardless of complexity. The report recommends that companies match the complexity of the AI model to the complexity of the task to avoid unnecessary costs.
Key details
- The report examined 18,000+ tasks and activities across 923 job types.
- Only 6% to 11% of automatable tasks in each industry require advanced 'frontier' AI models.
- Estimated five-year AI costs for a company with 100,000 employees using 35 agents range from $155 million (health care) to $520 million (insurance).
- Routine tasks, such as enrolling people in programs or drafting copy, generally do not need advanced models.
- Dynamic routing of tasks to different AI models is technically possible but rarely done at scale in enterprises.
Why it matters
Companies deploying AI may be incurring higher costs by using advanced models for tasks that do not require them. This affects business leaders and IT departments, who face increased AI-related expenses and may not be achieving the best return on investment. Adjusting AI model selection to fit task complexity could lower costs and improve efficiency for organizations using AI at scale.
