Only 33% of 502 senior AI decision-makers at AI-active enterprises said their biggest program met or beat its AI business case. An HFS Research survey with Rackspace, fielded in August and September 2026, puts that hit rate at 37% among firms using AI in one function, 28% where it spans several functions, and 56% where AI is embedded in standard operations. Those figures are success rates inside separate groups, not how many enterprises sit at each stage, and not a within-program drop as AI spreads.
Updated November 15, 2026
Key Takeaways
- Only 33% met or beat their biggest program’s AI business case.
- Hit rates: 37% one function, 28% several, 56% in standard operations.
- Productivity (58%) and cost cuts (42%) outpace CX (28%) and revenue (21%).
- 96% report significant governance gaps; 78% rank governance failures above technology.
How do business-case results differ by group?
Among enterprises using AI in a single function, 37% met or exceeded the case, according to MarketScale’s account of the HFS survey, from results published November 12. The share who met or exceeded the case is 28% for companies deploying across several functions and 56% where AI is part of standard operations. HFS cautions that these are different groups, a pattern rather than proof that maturity causes better returns.
In that several-function group, MarketScale reports HFS’s reading that ownership of the whole system tends to go missing. Respondents worked at firms with at least $100 million in annual revenue, across eight industries and four regions. All were running pilots or more, and 83% had AI in production or beyond. The survey was blind, and EconoTimes notes the results describe AI-active enterprises, not the wider market.
Why do productivity gains outrun revenue?
Of respondents, 58% cite workforce productivity gains and 42% cite measurable cost reductions. Only 28% report better customer experience, and 21% report measurable revenue growth. On these shares, efficiency is the common payoff. Revenue as the minority result is what respondents reported, not proof that growth cases fail.
The first four rows after the headline are business-case hit rates within separate company groups, not the share of firms at each stage. Value rows are overlapping shares, not a ranking that sums to 100%.
| Survey result | Share |
|---|---|
| Biggest program met or beat its case | 33% |
| Met or exceeded case: one function | 37% |
| Met or exceeded case: several functions | 28% |
| Met or exceeded case: standard operations | 56% |
| Workforce productivity gains | 58% |
| Measurable cost reductions | 42% |
| Better customer experience | 28% |
| Measurable revenue growth | 21% |
What does running AI as one system require?
Ninety-six percent reported significant governance gaps. For 78%, governance failures pose a bigger threat than technology failures to scaling AI. The most cited shortcoming was oversight of autonomous AI after deployment, at 55%. Another 44% lack a single, consistent way to run AI in production.
HFS groups the barriers as fragmented accountability, unpredictable costs, and governance gaps. EconoTimes reports that HFS traces weak results to unclear ownership, poor repeatability, and limited visibility. Phil Fersht, HFS founder and CEO, said, “Enterprise AI has an accountability problem that nobody in the industry wants to admit.”
What should CxOs settle before the next review?
A separate HFS 2027 research agenda says 87% of C-suite leaders are investing in AI faster than they can prove its value. It quotes Fersht: “In 2027, enterprises, providers, and technology firms all have to stop selling activity and start proving outcomes.” That 87% is not a result from the 502-person survey.
HFS proposes a Services-as-Software Operator, an internal team, providers under one remit, or a managed-services partner, to own live performance, cost, and governance. The enterprise keeps business context, policy, decision rights, and legal accountability. EconoTimes reports stated preferences, not purchases: 61% preferred one accountable partner across models, platforms, and infrastructure, and 73% said they would pay more than raw infrastructure costs for a fully managed service with governance accountability.
EconoTimes notes that Rackspace’s partnership, and its description of itself as an AI-stack operator, point to a commercial interest in the model HFS recommends.
Three checks follow from the published pattern:
- Name who answers for performance, cost, and governance after go-live, and against which measure.
- Decide if production is run one way, or each function keeps its own method.
- Split productivity and cost evidence from CX and revenue results fewer firms report.
Settle those checks before funding the next function. An AI trust scorecard before go-live matches the constraint 78% called a bigger threat to scaling than technology failures, beside the overconfidence gap in AI deals.
Frequently asked questions
What share of AI-active enterprises met their AI business case?
Only 33% of 502 senior AI decision-makers said their biggest program met or beat its business case. The survey covers AI-active enterprises, not the wider market.
How do business-case hit rates differ by deployment group?
The share who met or exceeded the case was 37% at firms using AI in one function, 28% where it spans several functions, and 56% where AI is part of standard operations. HFS says these are separate groups, not proof that maturity causes better returns.
Where do respondents say AI value shows up?
Productivity gains were cited by 58% and cost reductions by 42%, against 28% for better customer experience and 21% for measurable revenue growth.
Who does HFS say should run live AI?
HFS recommends an internal team, providers under one remit, or a managed-services partner to track performance, cost, and governance after go-live. The enterprise keeps business context, policy, decision rights, and legal accountability.