An AI Disruption Score with One, Probably Irrelevant, Omission
April 15, 2026
Another dinobaby post. No AI unless it is an image. This dinobaby is not Grandma Moses, just Grandpa Arnold.
I love consulting and services firms. An AI “disruption score” from AlixPartners caught the attention of Business Insider. “A New Scorecard Shows Which Software Companies Will Win or Lose in AI” explains how a savvy investor or AI hungry professional can choose which horse to ride in the smart software derby. The article points out that the SaaSpocalypse may be a bit of a problem and notes “one that could reshape private equity portfolios in painful ways.” Yep, SaaSpocalypse. I think this means bad news, but when consulting firms explain something there will be an upside and a downside. The girl scout bake sale will raise funds for a camping trip. The cookies could contain poison.
Yes, sir, we have an AI horse, an AI pony, and a sheep dressed up like AI. Which is the winner. The buyer needs help figuring out which of these animals can pull his delivery wagon. Thanks, Venice.ai. The weird colors are an innovation, right? Well, good enough.
So what is this “disruption score”?
The write up says that the “score” depends on “two main factors: data and vertical specialization.” How does one quantify “data” and “vertical specialization”? That’s the magic of many consulting firm explanations. General Eisenhower like the four square grid labeled with measurable data like men, supplies, and ammunition. The standard consulting grid uses icons like cows and stars. You get the idea. Abstractions sell explanations.
Alix scores companies on a spectrum. Risks are calculated. Those with high scores have moats, presumably capable of stopping invaders armed with bows and arrows. It is possible that an invader from China might have a laser weapon or information about a CEO’s off-site behavior, but that moat will definitely prevent a problem. Well, there is risk in human endeavor and technology.
The write up quotes someone from Alix as saying in a slide deck:
“Protection from AI disruption is far higher when companies own proprietary data, systems of context, ecosystem leverage, embedded workflows, and operate in regulated or critical domains,” the firm wrote in an exclusive presentation prepared for Business Insider.
Business Insider comments:
The results are stark. Only about 14% of companies analyzed had strong moats across both dimensions, while roughly a quarter had weak defenses on both fronts, leaving them highly vulnerable as AI-native competitors scale rapidly.
The Alix method makes something I thought was obvious “clear”; to wit:
The firm’s framework highlights that not all “systems of record” are equally safe. As the findings show, even enterprise resource planning (ERP) systems, long considered defensible, are only rated as having medium-strength moats. AI agents may reduce the number of human users, and therefore software licenses, while stripping away higher-margin add-ons and interfaces, according to AlixPartners’ Milicevic.
“Safe” is an interesting word. Does a user of smart software know if the provider is learning from what’s in the customer’s moat? My answer is, “Whom do you trust with your data? I trust few people and I don’t trust vendors. Period.”
The Business Insider write up makes a bold statement:
The firm [Alix] ultimately groups companies into four categories: “fortress” businesses with strong moats; “survivors” that must quickly build or acquire AI capabilities; firms likely to be sold to AI-native buyers; and those facing potential wind-down. The findings drive home a tough message: The era of growth-at-any-price SaaS is ending, and AI is accelerating a divide between a small group of defensible platforms and a much larger pool of exposed assets.
I want to end by addressing the “one, probably irrelevant, omission”. Smart software using the large language model methods output hallucinations, made up data, and misstatements. In short, the LLM is about “good enough” information. The problem is that some consulting firm clients may not spot inaccuracies. Not every CEO and MBA is an A student in every technical and business issue. Add to this the reality that orchestrated smart software may just accept outputs from one source without validating the accuracy of the information.
One can talk about moats, proprietary data, and yada yada. One can have matrices and scores. But the one, probably irrelevant, omission is that today’s AI systems are just good enough. Everyone with skin in the game wants the one-trick ponies to be stallions. Upon inspection some of those ponies and stallions are mules dressed up for cosplay.
One can do an analysis. Just make certain that the beasties are not in disguise.
Stephen E Arnold, April 15, 2026
Comments
Got something to say?
-
- Subscribe to Beyond Search
Feature archive
News archive
-
