95 per cent of organisations are getting zero return from AI according to MIT
ft.comNah. My wife is in finance at a pretty well known startup. She has been buzzing about how some AI products she's worked with have made direct impacts on her bottom line. She's been using Enterprise ChatGPT and she loves it. I think there's plenty of opportunity in places like finance where projects are limited by headcount and bandwidth, and creating features for long-tail work.
While I doubt that statistic as well, the pedant in me is duty-bound to point out that your wife seeing a return on its use is not actually very strong evidence counter to the claim that 95% of organizations are not. She might just be in the 5%.
Careful, the underlying report is saying 95% of enterprise AI automation systems are not giving measurable results for AI projects, and:
- does not include individual use of ChatGPT, Claude, etc (called shadow AI economy in the paper)
- in-house AI tools fail 2x more often than external AI solutions
- finance and other backoffice departments are the ones seeing noticeable performance gains with AI use
- lack of AI adaptability, memory and evolution over time is making it hard for humans to adapt to AI when it should be the other way around
- users prefer the chat interface, not your AI-enabled webapp
- 90% of users prefer humans (duh) for mission-critical work
The report does not include the companys who are loosing money due to AI, or the costs that are bieng transfered onto the general public via higher powerbills and such.and then unproven adoption by government, and the comming bursting bubble. A full forensic financial analyis will amost certainly reveal significant losses overall, with a few nitche areas that are truely successfully benifiting. Focus on the latter, NOW, might be a good move for those in the game.
Completely agree. There was a similar report published last year, lots of fluff, seemingly tailor made to scare retail investors. FT's quality has really taken a hit in recent years. It's sad to see them publishing nonsense like this.
A more evenhanded take on the same report:
https://nanda.media.mit.edu/ai_report_2025.pdf
https://web.archive.org/web/20250818145714/https://nanda.med...
first link is broken
Included for completeness, breakage is known.
I think some of it is the return is hidden. Things that took people an hour now taking half that. Presentations that were bland are now improving. The fact I can do some quick analysis of some data then analyze it in a number of ways for insights makes me more efficient.
Instead of 1x you are getting 1.1x out of employees that leverage AI in some fashion. yeah it's not much but over time and enough staff that adds up and as people become more familiar with how to prompt and use ai it will become that much more of a productive and valuable tool.
Slop does not improve anything. Many people have a visceral negative reaction to it. Every blog post I see with some crappy AI image attached, the content is drastically devalued for me. Some workflows are actually slower now. People become deskilled. Some start to fall into delusion.
This technology is not an industry win.
There's a lot of organisations in which AI using being used well either.
IT is generally a force multiplier, not a revenue generator.
I've seen some pretty bad AI related proposals.