I can believe it. Agentic AI only really succeeds in the long run when it’s built atop pristine data sources that are themselves automatically maintained and carefully curated.
And have you seen how large enterprises approach data? They threw that into a buzzword half a decade ago and haven’t improved upon it since.
My gut read:
* 40-50% will be outright cancelled by EoCY 27
* 30-40% will soldier onward with a specific model in use, on preserved systems
* 10-30% will be “successful”, actually creating business value by working on live data and in specific use cases (namely read-only chatbots pulling from CRM or other “live” systems)
* <10% will be “runaway successes”, in the form of permanently eliminating a job function from an enterprise or organization and producing significantly more value than they consume.
In business, perception often is taken for reality (just look at how long people clung to DOS-Windows when there were far better Amigas and Macs for sale).
I can believe it. Agentic AI only really succeeds in the long run when it’s built atop pristine data sources that are themselves automatically maintained and carefully curated.
And have you seen how large enterprises approach data? They threw that into a buzzword half a decade ago and haven’t improved upon it since.
My gut read:
* 40-50% will be outright cancelled by EoCY 27
* 30-40% will soldier onward with a specific model in use, on preserved systems
* 10-30% will be “successful”, actually creating business value by working on live data and in specific use cases (namely read-only chatbots pulling from CRM or other “live” systems)
* <10% will be “runaway successes”, in the form of permanently eliminating a job function from an enterprise or organization and producing significantly more value than they consume.
So a 60% success rate for these projects? Sounds like good odds!
If you define success simply as “not cancelled”, then yes, fantastic odds
A good fraction of the ones not cancelled will be the real pain.
What percentage of all software projects started in the past 12 month will be cancelled by the end of 2027?
I can easily believe that number.
However it's unclear how they arrived to it. Can anyone with a Gartner subscription take a look to see if this passes a smell test?
Probably going to be closer to a pareto distribution.
In business, perception often is taken for reality (just look at how long people clung to DOS-Windows when there were far better Amigas and Macs for sale).
Gartner.hasHistoryOfSettingPerceptionInFortune500Companies == YES
The "Gartner effect" is very real.
TIL: "agent washing"
This sounds about right. As fads go, it's a very expensive fad.
[dead]