The state of AI in healthcare (2025)

For years, healthcare was labelled a digital laggard. Not anymore. Menlo Ventures’ new report, ‘2025: The State of AI in Healthcare’, argues that the sector has “flipped the script,” now deploying AI at more than twice the rate (2.2x) of the broader economy. In just two years, adoption leapt from 3% to national leadership, evidence that AI is no longer a curiosity but a core operating strategy.

In this blog, Dr Richard Dune analyses the Menlo Ventures’ report and highlights what it means for the wider healthcare sector, particularly in the UK.

A step-change in adoption - Backed by real spend

The study (run with Morning Consult) surveyed 700+ U.S. healthcare executives across providers, payers and life sciences, and supplemented findings with 20+ stakeholder interviews. The topline is stark:

• 22% of healthcare organisations have implemented domain-specific AI tools - 7× up on 2024 and 10× on 2023

• Health systems lead on 27%, ahead of outpatient providers (18%) and payers (14%)

• Sector AI spend has nearly tripled to $1.4bn in 2025, with providers accounting for c.75% of that total.

Menlo’s authors spell out why momentum has accelerated: administrative overheads are crushing margins, clinicians are burning out, payer costs are rising, and life sciences productivity remains stubbornly flat. “AI offers the potential for improved efficiency, economics, and outcomes,” they conclude, and the dollars are following.

From pilots to production - Big systems, bigger bets

This is not a theory. The report highlights a series of landmark deployments that mark a clear pivot from experimentation to scale:

• Kaiser Permanente rolled out Abridge’s ambient documentation across 40 hospitals and more than 600 medical offices, “the largest generative AI rollout in healthcare history” and Kaiser’s fastest technology implementation in over 20 years

• Advocate Health reviewed 225+ AI tools and activated 40 use cases, from Microsoft Dragon Copilot for documentation to imaging and call-centre automation, projected to more than halve documentation time while automating prior authorisations, referrals and coding

• The Mayo Clinic is committing over $ 1 billion across more than 200 AI projects spanning administrative, diagnostic, and patient-care applications

• SimonMed is piloting over 50 AI systems across intake, scribing, and revenue cycle management.

The authors’ message is blunt: “Organisations that move quickly… are capturing advantages in cost structure, patient satisfaction, and clinical outcomes. Those that move slowly risk falling irreversibly behind.”

How leaders choose AI - Maturity, risk and near-term value

Unlike the EHR era, which was centralised, regulatory, and slow, the current wave is decentralised and pragmatic. Menlo finds that high-performing organisations use three filters:

1. Maturity of technology - Choose production-ready tools that perform reliably at scale
2. Level of risk to patient care - Start with low-risk administrative use; expose patient-facing tools only with deeper scrutiny
3. Short-term value delivery - Go after quick, credible wins to build organisational confidence.

Crucially, cost is secondary: buyers “will pay a premium for trusted AI solutions” where the risks of failure, operational disruption, patient harm, reputational damage, are existential.

Procurement is speeding up (for providers)

The report shows that procurement cycles are compressing by 18% for health systems (8.0 months → 6.6 months) and 22% for outpatient providers (6.0 months → 4.7 months). Payers, by contrast, have lengthened to 11.3 months, reflecting regulatory caution and concern about escalating claims and coding intensity. In other words, providers are now the fastest movers in healthcare tech, a historic reversal.

Where budgets go - Scribes, RCM and the ‘front door’ of care

Follow the money and you find the pain points:

• Ambient clinical documentation (ambient scribes): $600m

• Coding and billing automation: $450m

• Patient engagement and prior authorisation: both 10–20× YoY growth. 

Why? Because these categories deliver immediate, measurable ROI: less clinician “pajama time,” recovered revenue, faster access to care and reduced denial rates. Ambient scribes alone have created the first breakout market: Nuance/Microsoft (33%), Abridge (30%) and Ambience (13%) lead, but Menlo cautions the field is “contested” and switching costs are still low, which is why vendors are racing to expand into coding, revenue integrity and prior auth.

Startups vs incumbents - 85% of spend goes to challengers at least for now

A striking finding: 85% of generative AI spend currently flows to startups. AI-native challengers are winning on performance, speed and focus, while many incumbents are layering AI onto legacy platforms. But don’t count the giants out: Epic, Oracle Health and athenahealth are building embedded AI and ambient scribing, and many customers say they prefer to buy AI from their incumbent EHR for workflows beyond scribing (e.g., coding, prior auth, scheduling, navigation). Expect rapid platform competition over the next 12–24 months.

The bigger opportunity - Converting services spend into software

Menlo’s team estimates that U.S. healthcare administration totals $740 billion annually, yet IT accounts for just $63 billion, a vast ocean of manual work ripe for automation. Two routes are emerging:

• Augment existing IT in medical documentation ($19.6bn) and back-office RCM ($18.8bn) by inserting AI “intelligence layers” (e.g., Abridge, OpenEvidence, Commure, Smarter Technologies) between clinicians and EHRs or billing teams and claims processors

• Unlock services budgets in prior authorisation, patient engagement and front-office RCM, where software has historically captured only 3–5% of spend. Here, AI agents can compress processes from days to minutes, cut handoffs and improve access.

What this means for the wider UK health and social care sector

Although the Menlo dataset is U.S.-centred, the implications for the NHS, independent healthcare, and CQC-regulated social care are immediate:

1. Workforce relief is the killer use case. Ambient documentation and call-centre automation directly address burnout, free up clinical time and reduce agency reliance, priorities echoed in UK winter-pressure planning
2. Revenue and funding integrity translate to sustainability. For NHS providers, this means reducing unwarranted variation, improving coding quality, and accelerating discharge through streamlined administrative processes; for social care, it’s eliminating repetitive coordination tasks and enhancing bed flow across interfaces
3. Governance must keep pace. Rapid experimentation requires robust digital governance: model risk management, DPIAs, safety cases, bias monitoring, audit trails, and ongoing competency/CPD for staff who procure, operate and oversee AI
4. Platform thinking beats point solutions. As EHRs and specialist startups converge, providers should adopt modular architectures and open integration standards, avoiding lock-in while capturing quick wins
5. Inspection readiness will include AI assurance. Under the CQC Single Assessment Framework, evidence of safe, effective, well-led services increasingly includes how AI is governed: policies, role-based training, incident response, and measurable outcomes.

Practical actions for providers now

Target low-risk, high-ROI use cases first. Start with documentation, triage and admin scheduling; ring-fence clinical-decision tools for pilots with strong oversight

• Stand up an AI governance board. Include IG, clinical safety officers, DPOs, digital/IT, quality and risk, and workforce/training leads

• Instrument outcomes. Measure time saved, backlog reduction, coding accuracy, patient access and staff wellbeing, linking metrics to board-level KPIs

• Integrate with compliance. Keep your policies, SOPs, DPIAs and training records centralised and inspection-ready.

“Healthcare’s AI moment is here… Providers are seeing products that deliver ROI and witnessing peers adopt at scale,” the report notes. The message for UK leaders is clear: move with discipline, but move.

About The Mandatory Training Group

The Mandatory Training Group is a leading UK provider of accredited statutory and mandatory training, governance consultancy, and digital compliance solutions for the health and social care sector.

We help organisations strengthen workforce competence, modernise governance, and prepare for inspection under the CQC Single Assessment Framework.

Take the next step

• ComplyPlus™ Software - Our all-in-one regulatory compliance management system (LMS + TMS + policies + audits + dashboards)

• Health and Social Care e-Learning - CPD-accredited courses mapped to UK frameworks

• Governance and Compliance Consultancy - Digital transformation, policy management and inspection readiness.

Reference

Yap, G., Xiao, D., Hu, J., Sanday, J.P. & Beatty, C. (2025) - The State of AI in Healthcare.