Diving overview:
- OpenAI urges companies to rethink how they measure the return on investment in artificial intelligence, saying traditional software metrics fail to capture the business value the technology creates.
- OpenAI CFO Sarah Friar said in a blog post Friday that organizations need to consider factors such as the number of tasks successfully performed by AI and the total cost of completing them.
- “For years, software has been measured by adoption of seats, active users, updates, etc.,” Frier said in a separate post on LinkedIn. “AI is different. AI needs to be measured by the work accomplished.”
Dive Insight:
The move comes as companies face increasing pressure to demonstrate that their AI investments are delivering tangible business value.
According to Gartner, global spending on AI is expected to reach a total of $2.59 trillion in 2026, an increase of 47% year over year.
A PwC survey released in January found that only 12% of CEOs said AI had both cost and revenue benefits. Overall, 33% of respondents reported an increase in either costs or revenues, while 56% said they have not seen significant financial benefits so far.
The issue has become more vivid as scrutiny of AI token consumption, or the amount paid by AI processing companies to use large language models, has increased.
Palantir CEO Alex Karp recently claimed that many business leaders are becoming increasingly dissatisfied with the economics of LLM and are questioning whether rising token costs are reflected in ROI.
“Companies are just fed up,” Karp said in an interview with CNBC.
Karp made the comments while touting Palantir’s proprietary technology and acknowledged that his company has a commercial interest in the discussion.
In a blog post, Friar introduced the concept of determining the value of AI based on a metric called “useful intelligence per dollar.” You need to answer four questions:
- Can AI complete important tasks?
- What is the cost of each successful task?
- Can people depend on the results?
- As AI usage increases, does each AI dollar create more value?
“Tokens create value when they are transformed into works that people can use,” Friar wrote. “As models become more capable, they can perform longer and more complex tasks, such as maintaining context, making multi-step inferences, working across multiple tools, and adapting on the fly.”
