Gartner issued a stern warning this week to organizations across all sectors that the cost of introducing artificial intelligence (AI) to the workplace could easily balloon by 500 -1,000 percent.
Speaking at Gartner’s flagship Symposium event in Australia, VP analyst Mary Mesaglio said: “Factors contributing to these inflated costs include vendor price increases and neglecting the expense of utilizing cloud-based resources.”
Mesaglio highlighted that the inappropriate use of AI, such as deploying it for simple queries instead of search functions, could also contribute to cost overruns. She added that the cost of AI workloads or queries varies according to how organizations structure their AI’s access to data, as allowing AI to process unstructured data may yield more information and better results but at a higher cost.
Gartner also warns that staff time saved by the deployment of generative AI may not automatically be put to productive use. To describe this phenomenon, Mesaglio coined the term “productivity leakage” to describe one potential downside to successful AI implementation. Citing Gartner’s research indicating that employees using AI could save up to 43 minutes per day, she questioned whether this time would be used productively or contribute to a 10-30 percent drain on AI’s benefits.
“I would use the time to get a [café] latte,” quipped Mesaglio.
Productivity gain not equal across all staff
Gartner also noted that the productivity gains from AI do not apply equally across all employees, with less experienced workers potentially benefiting more than their seasoned counterparts. This is another factor that could make it extremely difficult for firms to identify staff time saved and apportion that time to productive tasks.
Gartner also stresses that, before embarking on the introduction of AI, companies should first establish precisely what it is that they expect AI to do for their business.
“Organizations should adopt a pace of AI implementation that aligns with their industry’s rate of change. For industries not significantly impacted by AI, a slower approach may be sufficient,” says Gartner.
Last year, Gartner VP analyst, Don Schneibenrief, attempted an informal experiment where he tried to use AI as member of his analyst team and found that it was not yet to be trusted with consultancy-oriented tasks.
“AI lies with perfect grammar”
“It’s like that annoying team member we all know, the know-it-all. Except that this one seems to lie with perfect grammar.”
Unfortunately, as Cyber Intelligence reported earlier this year, AI has already gained an unenviable reputation for making errors and even entirely fabricating information and quoting bogus non-existent but credible-sounding sources – a failing referred to as “hallucinating.”
A recent Copilot for Microsoft 365 Transparency Note also warns: “We encourage users to review all content generated by Copilot for Microsoft 365 before putting it to use.”
And Microsoft should know, as they first popularized the widespread use of generative AI with the launch of ChatGPT in November 2022.
