Why AI Matters for Healthcare & Pharma

Prior authorisation: 14 days, $31 per transaction, and climbing

Prior authorisation — the process of getting a health plan's approval before delivering a service — has become one of healthcare's most expensive administrative burdens. The numbers are staggering:

The time cost: The average prior auth request takes 14 days to process. For complex cases, it can take 30+ days. During that time, patients wait for treatment, clinicians chase paperwork, and health systems absorb the operational cost of managing the pipeline.

The financial cost: The CAQH Index estimates that prior authorisation costs $31 per transaction when factoring in staff time, follow-up, and administrative overhead. A large health system processing 50,000 prior auth requests per year is spending $1.55 million annually on a single administrative function.

The volume: The American Medical Association reports that physicians complete an average of 45 prior authorisation requests per week. 88% of physicians report that the burden has increased in the past five years. One in three physicians reports that prior auth has led to a serious adverse event for a patient due to treatment delays.

The denial-and-appeal cycle: Approximately 10-15% of prior auth requests are initially denied. Of those denials, roughly 50-60% are overturned on appeal — which means the initial denial was wrong, but the health system spent additional time and resources proving it. Each appeal letter requires assembling clinical documentation, citing medical necessity criteria, and referencing payer-specific guidelines.

This is a workflow that is repetitive, document-heavy, rule-based, and high-volume. It is precisely the kind of work where AI delivers immediate, measurable value.

Drug development: 10-15 years and $2.6 billion per approved drug

If you work in pharma or life sciences, the numbers are equally sobering. The average drug takes 10-15 years from discovery to FDA approval, with the Tufts Center for the Study of Drug Development estimating the capitalised cost at $2.6 billion per approved drug.

Where does that time and money go? A significant portion is consumed by documentation, review, and regulatory processes:

Clinical trial protocols run 200+ pages and require review for internal consistency, eligibility criteria clarity, endpoint definitions, and regulatory compliance. Protocol amendments — which occur in over 60% of trials — require the same review cycle again.

Literature reviews for a single drug programme may involve screening thousands of publications, extracting relevant data, and synthesising evidence. A systematic literature review for an FDA submission can take a team of medical writers 3-6 months.

Regulatory submissions (NDA/BLA for new drugs, 510(k) for devices) are massive document packages. An NDA submission to the FDA can exceed 100,000 pages. Every section must be internally consistent, properly cross-referenced, and compliant with FDA formatting requirements.

Pharmacovigilance — monitoring for adverse events after a drug reaches market — requires processing Individual Case Safety Reports (ICSRs) continuously. A major pharmaceutical company processes hundreds of thousands of ICSRs per year, each requiring review, classification, and potential regulatory reporting.

None of these are clinical or scientific decisions. They are documentation, review, and processing workflows. And they are all candidates for AI assistance.

3.2 million healthcare workers short by 2026

The documentation burden and prior auth crisis exist against a backdrop of a severe workforce shortage. The World Health Organization projects a global shortfall of 10 million health workers by 2030. In the United States alone, the projected shortage is 3.2 million healthcare workers by 2026, including:

• Nursing shortage: An estimated 200,000 registered nurse vacancies annually through 2031, according to the Bureau of Labor Statistics
• Physician shortage: AAMC projects a shortfall of 37,800 to 124,000 physicians by 2034
• Coding and HIM professionals: The shift to value-based care and increasing regulatory complexity is outpacing the supply of certified coders and health information management specialists
• Clinical research professionals: CRO capacity constraints and competition for clinical research coordinators are extending trial timelines

When you combine a growing documentation burden with a shrinking workforce, something has to give. Health systems cannot hire their way out of this problem — there are not enough qualified candidates. They cannot reduce documentation requirements — regulatory and liability demands are increasing, not decreasing.

The only viable path is making existing staff dramatically more productive. That is where AI fits: not replacing healthcare workers, but eliminating the administrative burden that prevents them from working at the top of their licence.

Every hour a physician saves on documentation is an hour returned to patient care. Every hour a nurse saves on charting is an hour at the bedside. Every hour a coder saves on chart review is an hour that can be applied to the growing volume of claims.

What Mayo Clinic, Cleveland Clinic, and Pfizer are doing with AI

The organisations that are furthest ahead in healthcare AI are not waiting for perfect tools. They are deploying AI today against their most painful operational workflows:

Mayo Clinic has integrated ambient clinical documentation — AI that listens to the physician-patient conversation and generates a structured clinical note. Early results show a 33% reduction in after-hours documentation time. They are also using AI for clinical trial matching, scanning patient records against trial eligibility criteria to identify candidates that would otherwise be missed.

Cleveland Clinic partnered with a major AI provider to deploy AI across multiple operational workflows: clinical note summarisation, patient message response drafting, and literature search. Their focus has been on reducing EHR inbox burden — the flood of patient messages, lab results, and referral notifications that physicians must process daily.

Pfizer used AI to accelerate its COVID-19 vaccine development timeline and has since expanded AI applications across its drug development pipeline. Their focus areas include clinical trial protocol optimisation (reducing the complexity and cost of trials), literature review and evidence synthesis (compressing months of work into weeks), and regulatory document preparation.

Kaiser Permanente deployed AI for automated ICD-10 code suggestion, reducing coding turnaround time and improving first-pass accuracy rates. Their CDI programme uses AI to flag documentation gaps before the chart reaches the coder, reducing query volumes and accelerating the revenue cycle.

Mount Sinai Health System has implemented AI-powered prior authorisation workflows that auto-populate clinical justification from the EHR, match patient data to payer criteria, and draft submission-ready letters. Early results show a 40% reduction in prior auth processing time.

These are not research projects or proof-of-concepts. They are production workflows at major health systems and pharmaceutical companies that recognised the unsustainable trajectory of documentation burden, staffing shortages, and operational costs.