The price of work, visible for all

LLMs are becoming increasingly proficient at providing salary information, offering a new, accessible tool for compensation benchmarking and negotiation

LAST SUMMER, WE wrote about a report from Payscale suggesting LLMs (ChatGPT, Claude, Gemini, etc.) were contributing to what the report’s authors called a “widening disconnect” around salary expectations.

Click here to view this article in the London Inc. Worklife newsletter

LLMs, they suggested, were chronically overestimating market salaries, and as we put it back then, “quietly pushing the 20 per cent or so of jobseekers who use tools like Gemini or ChatGPT to research salaries to unknowingly drive up their salary expectations.”

A new report from OpenAI (makers of ChatGPT) released last month gave us another look at this trend, and suggested not only that workers are increasing their reliance on LLMs to research salary and wage information (sending as many as three million messages per day about it), but that the tools themselves are actually getting much more accurate at the task of determining pay benchmarks.

Story Continues Below

 

“AI is a new type of labour-market intermediary,” reads the OpenAI report, titled ‘ChatGPT and the Price of Work’. “Rather than requiring a worker to search across multiple websites, interpret scattered salary pages or ask a socially costly question, a model can synthesize wage information and return a benchmark in seconds. ChatGPT is already being used in this way.”

And, according to OpenAI at least, workers are getting much better answers now than they were a year ago. In a benchmarking test, it claims ChatGPT was able to accurately provide market wage information to within 10 per cent of the actual average, 99.8 per cent of the time. “These results imply that ChatGPT can serve as a reliable national wage look-up layer for workers,” OpenAI concluded.

HR professionals say this is likely the new normal. “Gone are the days where it was a song and dance at the end of the interview process to see what you might get,” says Louisa Benedicto, vice-president of HR recruiting at Hays, speaking to HR Reporter. “It will become a more demanding task for managers, who will be required to have a rationale, have an explanation, have a basis for determining salary and compensation.”

Story Continues Below

 

It’s still very early in this transition, but the evidence suggests that for all the complaints people have about AI’s role in the jobseeking process, it is at least helping to smooth over salary negotiations. A study this month from Eastern Washington University found that 78 per cent of professionals felt more confident about salary negotiations, and 63 per cent felt their use of an LLM to prepare actually got them a better deal.

Some HR professionals think it could lead to a more transparent culture around pay, which would benefit everyone. “Employers who are less transparent would find themselves more vulnerably, in that the employee could be armed with all this data,” said Anil Verma of the University of Toronto. “I think employers should be fully aware of what information is available through AI, and to be prepared.” Kieran Delamont