Beyond the pilot: How CIOs are operationalizing AI across health systems in 2026 (Qventus)
TL;DR
The second annual CIO report from Qventus, a vendor of AI care-operations automation. It rests on a survey and one-to-one interviews with more than 60 CIOs, chief AI officers, CMIOs and other senior IT leaders at medium and large US health systems (PDF p. 3). Five findings:
- Few have scaled. 42% are actively deploying AI across several use cases, but only 4% have “achieved scaled AI implementation with measurable outcomes.” 45% name difficulty scaling pilots as a top obstacle (p. 5).
- ROI is hard to measure. Four in five struggle to determine or measure the ROI of AI. 74% need ROI within a year to justify an investment, but 38% do not expect it for 13 months or more (p. 5–6).
- EHR dependency is the top barrier. 74% say reliance on their EHR vendor’s AI roadmap is a top obstacle. Asked to choose between waiting 18 months for an EHR feature and a proven third-party tool in three months, 22% would wait, down from 52% in the 2025 survey (p. 6, p. 10).
- Vendor sprawl, and a wish for one partner. 51% spend 11–25% of IT bandwidth on vendor management and integration; 72% would prefer one comprehensive AI partner, but 11% work that way today (p. 9).
- Agents: mostly wait and see. 57% describe their stance on agentic AI as “healthy skepticism,” 34% are investing, and fewer than 6% have live deployments with measurable ROI. Yet 55% say some AI initiatives already run “with limited oversight from staff” (p. 8).
Key data
| Finding | Figure | PDF page |
|---|---|---|
| Pressure to operationalize AI rated 7+ out of 10 | 65% | p. 2 |
| Actively deploying AI across multiple use cases | 42% | p. 5 |
| Scaled AI implementation with measurable outcomes | 4% | p. 5 |
| Piloting only in limited areas | ”nearly a quarter” | p. 5 |
| Struggle to determine or measure ROI | 80% | p. 5, p. 12 |
| No clear process for benchmarking AI performance or ROI | 39% (p. 5); “a quarter” (p. 2) | p. 2, p. 5 |
| Top obstacle: dependence on EHR vendor’s AI roadmap | 74% | p. 6, p. 10 |
| Top obstacle: IT resources to manage multiple AI vendors | 66% | p. 6, p. 9 |
| Top obstacle: no vendor willing to guarantee ROI | 53% | p. 6 |
| Would wait 18 months for the EHR feature (2025 → 2026) | 52% → 22% | p. 10 |
| Agentic AI: healthy skepticism / investing / live and proven | 57% / 34% / <6% | p. 8 |
| Some AI initiatives run with limited staff oversight | 55% | p. 8 |
| Prefer one AI partner / work with one today | 72% / 11% | p. 9 |
| ”Tech fatigue” is a barrier to adoption | 87% | p. 11 |
| Top 2026 priority: quality outcomes, compliance and patient safety | 60% | p. 7 |
PDF page numbers; the printed page number is one lower (the cover is unnumbered).
Method and definitions
- Sample. “More than 60” senior technology leaders: CIOs, chief AI officers, CMIOs and others, at medium and large health systems; the press release adds “U.S.” (p. 3; release). No breakdown by role, size or region.
- Method. A survey plus one-to-one interviews (p. 3). Fieldwork dates, question wording, response options and the number answering each question are not given.
- Definitions. “Scaled AI implementation with measurable outcomes” is not defined further; neither is “actively deploying.” “Limited oversight from staff” is not defined either.
- Sponsor. Qventus, which sells an “AI teammate” platform for care operations (p. 13). The report recommends risk-sharing contracts, consolidation onto one platform partner and “automated care operations” (p. 12), which is the category Qventus sells.
- Previous edition. A first report appeared in June 2025; it is not ingested. The report compares one question with it (EHR waiting, p. 10).
Scope and reliability
- Vendor research, small N. With about 60 respondents, the 4% is two or three health systems. No margins of error.
- Recommendations follow the seller’s product. The survey questions (one partner versus several, waiting for the EHR versus a third party) frame the choice Qventus competes in.
- An internal inconsistency. The share without a clear benchmarking process is “a quarter” on p. 2 and 39% on p. 5.
- Two different 4%s. Besides the 4% that have scaled, another 4% say they have “the appropriate resources and processes” for vendor management (p. 9). Quote the scaling figure from p. 5.
- Borrowed context. Several framing figures come from other sources: McKinsey’s 88% of organisations using AI (p. 3) and a 2025 MIT report’s “95% of generative AI pilots failed,” cited via MLQ.ai (p. 4).
Dynamic-capabilities reading
digital-seizing/balancing-digital-portfolios: the report is about scaling up from pilots (42% deploying, 4% scaled) and about balancing internal and external options: wait for the EHR vendor’s feature or buy a third-party tool now (p. 10).digital-transforming/navigating-innovation-ecosystems: health systems manage four to seven AI vendors at once (25%) and say they want one comprehensive partner (72%) (p. 9).contextual/internal-barriers: IT bandwidth spent on vendors, “tech fatigue” among staff (87%) and the inability to measure ROI (80%) hold back scaling (p. 5, p. 9, p. 11).
How it connects
- enterprise-ai-adoption — the pilot-to-scale gap, measured among US health-system technology leaders.
- ai-maturity-measurement-comparison — a strict bar (“scaled with measurable outcomes,” 4%) next to a loose one (“actively deploying across multiple use cases,” 42%) in the same survey.
- ASTP Data Brief 80 covers US hospitals in 2024: 71% use predictive AI in the EHR and most name several bodies accountable for evaluating it. Qventus asks a smaller group of larger systems in early 2026 about scale and payoff.
- Dinakaran (Luminai) — a founder selling operations automation to large hospitals; Qventus’s respondents are the buyers.
- agent-oversight-and-delegation — 55% say some AI already runs with limited staff oversight, while 57% are sceptical of agentic AI (p. 8).
- McKinsey & Company — source of the 88% economy-wide adoption figure the report uses as context (p. 3).
- ai-in-health-care: the concept page that sets the health-care sources in this wiki side by side.
What was actually ingested
The full 13-page PDF, downloaded from Qventus’s own site, and the press release from Qventus’s newsroom (raw/articles/2026-04-09-qventus-cio-report-press-release.md). The Business Wire copy cited in the Dutch report returned HTTP 403; the newsroom copy carries the same release.
Linked entities and concepts
- enterprise-ai-adoption, agent-oversight-and-delegation — concepts.
- McKinsey & Company — cited for context.
Dangling (single-source mention, deferred): Qventus (publisher); Mudit Garg (Qventus CEO, press release); Joseph Sanford (UAMS); Deepti Pandita (UCI Health); James Whitfill and Matthew Anderson (HonorHealth); Morgan Jeffries (Geisinger); Dwight Raum (Rochester Regional Health); Menlo Ventures and MLQ.ai (cited).