Skip to content

SignalCross-industryData story

A Third Met the Business Case

HFS Research says a third of the AI-active enterprises it surveyed report that their most significant AI program met or exceeded its business case, but never says what a business case is. It's the first framework through the Multiplicity Test™. HFS ran the study with Rackspace.

Multiplicity Test™ scorecard for HFS: 33% of AI-active firms say their top AI program met or exceeded its business case. Nouns defined: No. Judgment: named, not filled. Self-reported.

Key figures

  • 33% Of AI-active enterprises say their most significant AI program met or exceeded its business case. Survey of 502 senior AI decision-makers, August to September 2026, enterprises with at least $100 million in annual revenue. (HFS Research release, 12 Nov 2026)

The study was conducted with Rackspace, which sells managed services, one way to fill the operator role the study calls for. Figures are self-reported survey answers.

What is an AI-active enterprise? Best we define that before tackling the story out of HFS Research, which published the results of a recent survey it took of enterprise AI leaders.

HFS does not say what AI-active means. It says who it asked. There were 502 of them. They ran AI at companies that made $100 million a year or more. Every company had AI running in some form. Most had it in production. Some were still trying it out. These were not the people waiting on AI. They were the ones already doing it. HFS knows this and says so. The findings, it says, "describe AI-active enterprises, not market-wide AI adoption." HFS expects the rest of the market to be further behind.

HFS Research says only 33 percent of the AI-active enterprises it surveyed report that their most significant AI program has met or exceeded its business case. That sounds like a verdict. Look closer.

What's a "business case"? What counts as "value"? The report never says. Each company judged itself against its own, so 33 percent is simply how many said they hit it.

That's why I'm starting the Multiplicity Test™, my own yardstick for judging how AI frameworks are built. It borrows from the later Wittgenstein, who gave up on carving the world into tidy, labeled pieces and decided meaning lives in use. Language, he wrote, has countless uses, and "this multiplicity is not something fixed." Many AI frameworks are still built the early way: name every part, then write the rules. The test asks two things of any framework. Are the nouns defined, so anyone could check a claim against them? And does it say where the judgment lives and who holds it, rather than pretending to write it as a rule?

HFS fails the first. On the second, it names an operator to hold the judgment, then says no provider type owns that role yet. Rackspace, which ran the study with HFS, sells that kind of operator service.

Still, they got one thing right: nobody is accountable for the joins between the parts. That's what I'd call the weave. They found the gap, but on a foundation they never defined.

Passing looks like this: a framework that defines its terms so anyone could check a claim, then names who decides when to act, and how. That's the bar I'll hold every framework to here. HFS is the first. It found the right gap and missed the foundation. The Index will keep score.

Get new AI Value Index records in your inbox.

Subscribe