Editorial Article
Seven Hundred Hours a Year: the Prior-Authorisation Bill AI Is Being Bought to Cut
Seven hundred hours. That is how much administrative time CMS attributes to prior authorisation for an average provider in a single year — roughly USD 34,000 at the agency’s own estimate of USD 20–50 per hour, spread across 13 hours of every week.
Set that number beside the more fashionable one. Image-based verification of dispensed medication is the application that gets the demonstrations, the conference keynotes and the venture money. Prior authorisation gets a footnote. Yet on CMS’s arithmetic the administrative workflow represents one of the largest single pools of removable labour in pharmacy, and it is being dissolved by a dated regulatory obligation rather than by a clinical one.
A date, which this market almost never gets
From 1 January 2027, certain health plans regulated by CMS must implement and maintain a set of standardised APIs — Prior Authorization, Patient Access, Provider Directory, Provider Access and Payer-to-Payer. Health plans have to expose the coverage rules and documentation requirements as structured, machine-readable transactions. Providers get real-time access to what a payer actually wants.
Pharmacy does not often get a delivery date. Market forecasts are ranges, adoption curves are arguments, and procurement cycles slip quietly. A regulatory deadline is different: it converts a discretionary efficiency purchase — the kind that gets deferred when a health system’s capital committee meets — into a compliance obligation with a fixed horizon.
One honest qualification, because it changes how much of this is really pharmacy’s. The CMS programme governs prior authorisation for medical items and services on the medical benefit. A substantial share of pharmacy authorisation runs through pharmacy-benefit rails and NCPDP standards rather than through the CMS APIs. So 1 January 2027 is not a pharmacy mandate. It is a hard catalyst on the medical-benefit side and a very strong directional signal for everything adjacent: once structured authorisation is the norm on one benefit, the fax-and-portal version on the other looks like what it is.
Why this application, and not verification
The reason prior authorisation is the fastest-spreading use of AI in pharmacy is not that it is the most valuable. It is that it is the least contested.
Clinical verification sits inside device regulation where the software informs a decision on an individual patient. It requires validation against a site’s own error history, it appears on the FDA’s AI-Enabled Medical Device List if it clears, and — decisively — in most US states the final verification of a dispensed prescription remains a licensed pharmacist function. Automation is permitted only within approved exception-based frameworks. That caps how much labour a verification model can actually remove, no matter how accurate it becomes.
Administrative triage has none of those constraints. Reading a payer’s criteria and assembling the documentation is document understanding applied to a process that is still, in large parts of the system, a fax. There is no device classification, no pharmacist signature requirement, no clinical liability to underwrite. The barrier is payer heterogeneity, which is an integration problem rather than a regulatory one.
That is why the growth in this market is concentrated where it is. The technology that gets discussed is the technology that is hardest to deploy.
The labour it is competing with
The demand side is not speculative. The American Society of Health-System Pharmacists surveyed hospital pharmacy directors in autumn 2024 and found 56.2 per cent of hospitals reporting shortages in pharmacy leadership roles, at the same time as pharmacist deployment into clinical settings expanded sharply. Emergency department involvement rose from 10.9 per cent a decade earlier to 46.5 per cent; critical care reached 68.5 per cent, oncology 56.9 per cent, antimicrobial stewardship 48.1 per cent.
Hospitals want pharmacists doing clinical work and cannot hire enough people to lead that shift. Every hour spent assembling a prior-authorisation packet is an hour subtracted from the work the survey says they are trying to expand. The trade is not abstract; pharmacy directors described it to us in almost exactly those terms.
What the deflation means for this particular fight
There is a trap here, and it is the same trap that governs the whole market. AI in pharmacy grows 18.82 per cent a year — USD 1,840 million in 2025 to USD 7,310 million by 2033 — while revenue per AI-enabled site falls from USD 35,385 to USD 18,460, or 7.81 per cent annually. Administrative applications are the least defensible of all from a pricing standpoint, because a document-understanding workflow is precisely the kind of capability a general-purpose model can absorb.
Which means the vendors who win the 2027 window will not be the ones with the best extraction accuracy. They will be the ones already inside the pharmacy management system or the EHR, where the output can act on an order without anyone retyping it, and where re-validation costs more than the licence saving. In prior authorisation the moat is the integration, not the model.
The countdown that actually matters
The industry will spend 2026 arguing about which AI application in pharmacy is most valuable. The calendar has already answered a narrower and more useful question: which one has a deadline.
Read the full analysis in Global Artificial Intelligence in Pharmacy Market Report, Size, Share and Forecast 2026–2033. Sources: CMS electronic prior authorisation programme; ASHP national survey, American Journal of Health-System Pharmacy; FDA AI-Enabled Medical Device List; Pheonix Research modelling.
Frequently Asked Questions
How much does prior authorisation cost a provider?
CMS estimates prior authorisation costs providers USD 20-50 per hour and consumes about 13 hours a week - roughly USD 34,000 and 700 hours of administrative time per provider per year.
What changes on 1 January 2027?
Certain health plans regulated by CMS must implement and maintain standardised APIs, including a Prior Authorization API, from 1 January 2027. It applies to medical items and services on the medical benefit, so it is a directional catalyst for pharmacy rather than a pharmacy-specific mandate.
Why is administrative AI spreading faster than clinical verification?
Because it is the least contested. It carries no device classification, no licensed-pharmacist signature requirement, and no clinical liability to underwrite. The barrier is payer heterogeneity, which is an integration problem rather than a regulatory one.
What makes a prior-authorisation AI vendor defensible?
The integration. A document-understanding workflow is exactly the kind of capability a general-purpose model can absorb, so extraction accuracy is not defensible. Being inside the pharmacy management system, where output acts on an order without retyping, is.
