Why healthcare services struggle to count the cost and success of AI

ARTIFICIAL intelligence-deployment within the healthcare industry is now at a faster 2.2 times rate compared with the broader economy in the US, said venture capital firm Menlo Ventures. Health systems are leading the adoption.

Despite having a slower start compared with other sectors, healthcare and biotech companies are beginning to embrace AI, with software such as OpenEvidence for evidence-based medicine, Abridge for ambient clinical documentation and Tempus AI that offers precision medicine solutions.

Hospital chief financial officers (CFOs) are being confronted, however, with the broader question beyond massive capital expenditure figures.

AI is built within the healthcare system architecture

In an industry pressured by broad macro transformation imperatives, the hook of “plug-and-play savings” by AI evangelists square up against the actual reality of how predicted efficiency savings rarely equate to actual cash liquidity.

Why? Management theorist Peter Drucker famously opined that healthcare is the most complex human organisation ever devised. The answer is in the architecture of healthcare, and how returns on investments, revenue generation and workflow processes are being designed.

On the surface, better AI tools, to relieve healthcare workers of mundane, repetitive, non-revenue generating yet essential tasks, should bring about an efficiency boost.

But such tasks are built within an architecture that requires integration. How will these AI solutions be linked within broader healthcare service practices?

Payers and insurance companies will know that having physicians review more patients an hour will mean an increase in consultations, referrals and prescriptions.

Incidental findings noted via imaging or additional tests will further lead to a cascade of care, and may drive costs higher from a system perspective.

On the manpower end, healthcare managers are trying to avoid a sudden loss of hard-to-train healthcare workers, yet how may such manpower be redeployed and reskilled for the extant care model?

And finally, risk. While clinicians globally are encouraged to adopt AI to supplement work; concerns about duty of care, negligence and not fully understanding the technology (“black box”) continue to overlay how professional services are being priced and adopted.

Moreover, risks of “unknown unknowns”, arising from differing AI models, could lead to real patient harm if validation and transparency are not addressed. Pricing in AI-driven cybersecurity threats further expose healthcare companies to significant upward cost pressures.

Billed beds versus speculative savings

Tenet Healthcare, a Dallas-based healthcare services company that operates hundreds of healthcare facilities, experienced significant growth and analysts’ target price revisions in the second quarter of 2026, driven by higher ambulatory revenue (from services not requiring an overnight hospital stay).

In healthcare, it is easy for analysts to measure two factors: volume and pricing revenue, all within established payer frameworks. But what healthcare AI solutions promise is prediction, seamless monitoring and time savings.

Investors continue to grade healthcare service providers via traditional financial metrics, yet AI companies are talking about speculative efficiency savings.

Instead of measuring things that do happen – procedures billed, beds filled, lab tests run – CFOs are asked to price probability against hypothetical costs. Moreover, hospitals rarely have an accurate cost ledger resulting from a minor 10-minute delay in care.

This leads to a value-based care lag. Financial rewards from using preventive health AI mean insurance companies pocket the upside tomorrow, yet healthcare service providers bear the upfront AI software costs today.

What, then, is the cost of driving AI transformation within healthcare services?

The price for organisations is the harder work of implementing change management, clinical validation and integration across stakeholders ranging from payers to regulators.

Unlike a traditional magnetic resonance imaging machine where depreciation is predictable and well-modelled, AI use in clinical settings suffers from model drift (when the performance of an initially accurate model degrades over time). You cannot set it and forget it.

Rather, revalidating the goals and impact of AI is the operational drag and friction for healthcare providers worldwide.

AI as a force multiplier

What, then, is the solution?

Biotech investor RA Capital has long argued that while 80 per cent of healthcare spending in the US is derived from healthcare services, only 8 per cent is on novel medicines. Yet, it is the latter minority spending that drives innovation.

Just as medical breakthroughs are a force multiplier for better patient outcomes, AI must similarly be viewed as the investment that generates momentum and improves standards of care.

Trying to mathematically isolate the financial return on investment of a prevented medical error, a shortened intensive-care-unit stay or a single nurse’s saved hour is a structural dead end.

In healthcare services, the numbers are rarely objective truths, but rather a flexible canvas used to justify whatever strategic narrative the board wants to sell.

Financial modelling in healthcare is often a rear-view mirror used to justify a forward-looking strategy, where the math simply mirrors corporate will.

The true test of AI impact is architectural transformation. When a venture capitalist invests in a breakthrough technology, they do not ask for a line-item ledger of how many software engineers the tool will save for the company. They look at structural scale.

For healthcare organisations, the question is therefore: Does this technology fundamentally redefine healthcare’s perennial manpower, financial and resource capacity constraints?

The writer, a medical doctor by training, advises on macro strategy for a global long-only investment firm. He used AI for ideation and editing before submitting the draft.