When it comes to AI ROI, IT decision-makers not convinced Many business leaders remain unconvinced that AI is worth the expense despite continued hype from an industry that has bet billions on developing the tech and desperately needs to recoup that spending. Research commissioned by Lenovo which polled nearly 3,000 biz execs and senior IT decision makers, confirmed that quantifying …
I genuinely had no idea what you meant, until I saw this:
https://www.bbc.co.uk/news
Google has re-edited an advert for its leading artificial intelligence (AI) tool, Gemini, after it overestimated the global appetite for Gouda.
The commercial - which was supposed to showcase Gemini's abilities - was created to be broadcast during the Super Bowl.
It showed the tool helping a cheesemonger in Wisconsin write a product description by informing him Gouda accounts for "50 to 60 percent of global cheese consumption".
Edit; Correct URL this time, oops!
Anyone involved in making a decision about any AI related activities in an org is facing a few uncomfortable facts
Garbage in garbage out data quality problems even if you sandbox an LLM
Use case clarity
RoI statement even vaguely based on reality
Ooh shiny, I want and I want it now brigade
My approach has been to let users play with CoPilot on a sandboxed environment in Non-Prod, use it to recraft your 30 page report, use it in PowerBi to make prettier dashboards etc etc
It scratches their itch, they get to see how it's not going to do all the work for them, company gets to say they are using AI and we get to soldier on with some real meaningful data activities.
The issue is that some AI systems, which are used for very specific purposes with well-understood aims can be very useful. 'Unrolling' burnt manuscripts from Herculaneum is one*. The problem for businesses is that they are sold an AI that will work for them in a much more general way. When they discover that they need to specify the area for the AI in great detail, and understand it very well first, it alls starts getting much more difficult.
* https://www.theguardian.com/science/2025/feb/05/ai-helps-researchers-read-ancient-scroll-burned-to-a-crisp-in-vesuvius-eruption
"The papyrus, known as PHerc. 172, is one of three Herculaneum scrolls housed at the Bodleian libraries. The document was virtually unrolled on a computer, revealing multiple columns of text which scholars at Oxford have now begun to read. One word written in Ancient Greek, διατροπή, meaning disgust, appears twice within a few columns of text, they said."
It is an expensive additional cost for something with inherent flaws and so offers limited or minimal benefits in most cases.
Just give it up. It got further than the Metaverse and NFTs, but nobody is going to be blowing $500bn on this, because they won't get $500bn and $1 back. Not now. Not ever.
The next big thing may well be a switch back to simpler, less networked tech and paper. Because nobody is going to access, ransom or delete your paper records from Moscow whilst you sleep.
>"quantifying AI's return on investment continues to be one of the greatest barriers to its adoption"
I don't think that sentence is true. "Quantifying" would mean that the problem is figuring out exactly how big the ROI is. But I think the real problem is figuring out how to get a ROI in the first place. I suspect the real meaning of that sentence is "we can't figure out a way to massage the data so that a positive ROI comes out".
Are these customers the same people who are the employees of the businesses to whom you are selling AI services as a tool to remove/reduce/replace same said employees.
I'm guessing handing over money to MS, Google et al will not be at the top of the former employees list of priorities to be paid from the unemployment cheque.
The Register 
Biting the hand that feeds IT
Copyright. All rights reserved © 1998–2025
