AI and automation in pharmacy practice are taking over dispensing, order screening and inventory, which could free pharmacists for clinical work such as prescribing and medication review. That shift happens only where health systems pay for the clinical work. Where they pay only for dispensed items, automation cuts pharmacist jobs instead of upgrading them.
Walk into a busy community pharmacy at five in the evening and you will see the profession’s problem in one frame. A pharmacist with a clinical degree, trained in pharmacology, interactions and dosing, is checking that the label on a box of amlodipine matches the prescription. Behind her, a queue. On the counter, a patient who wants to ask about a new rash and a technician fielding a phone call about a stock shortage.
For decades, pharmacy leaders have said automation would take the counting, labeling and checking off pharmacists’ hands so they could practice as clinicians. Robots that pick and pack have been in hospital pharmacies for years. Central fill and hub-and-spoke dispensing are expanding. AI now screens medication orders, predicts stock-outs and flags risky combinations.
Here is the uncomfortable thesis. Automation does not move pharmacists toward clinical roles by itself. It moves them wherever the money is. In systems that pay pharmacies mainly per item dispensed, automating dispensing reduces the need for pharmacists. In systems that pay for consultations, prescribing and medication reviews, it releases them to do those things. The technology is identical. The destination is a payment decision.
What does AI actually do in pharmacy today?
Three categories dominate.
The first is dispensing automation: robots that store, pick and label packs, and centralized facilities that assemble prescriptions at scale for local pharmacies to hand out. This is mostly conventional automation rather than AI, but it is where the biggest labor shift happens.
The second is clinical decision support: software that screens orders for interactions, duplications, dose errors and allergies. This has existed for decades, and its record is a warning. A widely cited systematic review by van der Sijs and colleagues in the Journal of the American Medical Informatics Association in 2006 found that clinicians overrode drug safety alerts in 49% to 96% of cases. Rules-based alerts fired so often, and so often irrelevantly, that people stopped reading them.
The third is newer machine learning: models that rank medication orders by risk so a pharmacist reviews the dangerous ones first, predict which patients will need intervention, or forecast demand. A 2023 systematic review in the European Journal of Hospital Pharmacy found 13 studies of machine learning models for detecting inappropriate hospital prescriptions. Only two had been evaluated in actual clinical practice, and nearly all were rated at high risk of bias. The authors concluded it was still too early to predict the impact of these tools in real pharmacy practice.
So the honest picture is this: dispensing automation is mature and labor-saving, clinical AI in pharmacy is promising but thinly evidenced, and the older generation of decision support taught the profession how easily software creates noise.
Will AI replace pharmacists?
It will replace a large share of what many pharmacists currently spend their day doing. That is different from replacing pharmacists.
The product-focused tasks, counting, labeling, final accuracy checks on routine items, are exactly the kind of well-defined, high-volume, checkable work that machines do well. In many countries, accuracy checking by trained technicians, supported by barcode scanning and automation, is already accepted practice.
The clinical tasks, deciding whether a medicine is appropriate for this patient, adjusting doses for kidney function, deprescribing in a frail older adult on twelve drugs, treating a urinary tract infection without a GP visit, need judgment, conversation and accountability. AI can support these tasks. It cannot hold the license or the liability for them.
The question is whether the second category grows fast enough to absorb pharmacists displaced from the first. That depends on regulation and payment, and England and the United States offer a revealing contrast.
How is England moving pharmacists toward clinical roles?
England has spent several years deliberately doing what pharmacy leaders elsewhere talk about.
In January 2024 it launched Pharmacy First, which allows community pharmacists to assess and treat seven common conditions, including sore throat and uncomplicated urinary tract infections in women, and supply prescription-only medicines where appropriate, paid through an NHS service fee rather than per item. Between launch and the end of February 2025, pharmacies delivered around 5.4 million Pharmacy First consultations, according to figures reported by The Pharmaceutical Journal.
In October 2025, legislation took effect allowing hub-and-spoke dispensing between different pharmacy businesses, so an independent pharmacy can have prescriptions assembled at a large automated hub run by another company. That is the automation half of the bargain: it takes assembly work out of the local pharmacy.
The prescribing half arrives this summer. The General Pharmaceutical Council has reformed initial education so that, from summer 2026, most newly qualified pharmacists will join the register as independent prescribers. NHS plans envisage bringing independent prescribing into the community pharmacy contract as an extension of Pharmacy First and contraception services.
The design logic is coherent. Automation takes assembly out of the pharmacy, regulation expands what pharmacists may do, and a service payment makes the clinical work financially viable. None of this is guaranteed to succeed. Community pharmacies in England have faced severe funding pressure and closures. But the pieces point in the same direction.
Why is the US picture different?
In the United States, pharmacists are not recognized as providers under Medicare Part B, so in most settings they cannot bill Medicare for clinical services the way physicians, nurse practitioners and physician assistants can. State scope-of-practice rules vary widely, and some states allow pharmacists to prescribe contraception or treat certain conditions. But the dominant revenue model for community pharmacy is still the dispensed prescription, squeezed by pharmacy benefit manager reimbursement.
Under that model, automation’s savings show up as fewer pharmacist hours per prescription. The pharmacist who is freed from checking labels has no billable clinical service to move into, so the business case says reduce staffing, not redeploy it. That is how you get the paradox of highly trained clinicians working in chaotic stores while health systems complain of shortages in medication management, chronic disease support and deprescribing.
This is the same lesson we drew about task-shifting generally: moving work to a different professional works when the system around it, including training, supervision and payment, is built for it. It fails when it is done as a cost cut.
A worked example: two pharmacies, same robot
Consider two hypothetical community pharmacies, each dispensing about 400 items a day with two pharmacists and three technicians. Each installs the same dispensing robot and joins a central-fill arrangement, cutting pharmacist time spent on assembly and routine checks by roughly half.
Pharmacy A operates under per-item payment only. The owner looks at the numbers and realizes the same volume can be handled with one pharmacist and two technicians. One pharmacist post goes. Patients see a faster counter, but no new clinical services, because none are paid for.
Pharmacy B operates in a system that pays a fee for minor illness consultations, structured medication reviews and prescribing. The freed pharmacist time goes into a consultation room. If the pharmacist handles, say, 20 consultations a day that would otherwise have gone to a GP or urgent care, the pharmacy earns new service income and the local health system gains appointment capacity.
The robot is the same in both. Pharmacy A’s automation shrinks the profession. Pharmacy B’s automation changes it. The difference is entirely in the payment rules, which are set by governments and payers, not vendors.
What is the strongest counterargument?
The strongest objection is that pharmacists moving into clinical roles duplicates work physicians and nurses already do, fragments care and adds cost. Critics also worry that prescribing in a retail setting, with commercial pressure and incomplete records, risks inappropriate supply, including antibiotics.
These are fair concerns. Pharmacist prescribing works best with access to the patient’s record, clear protocols and good referral routes. England’s own evaluation work on Pharmacy First, including analysis of how consultations are recorded in GP records, has shown how uneven information flow can be.
But the objection is an argument for doing it properly, not for leaving a clinically trained workforce at the label printer. Most health systems face growing demand for medication management from ageing, multimorbid populations. The workforce best trained for that work is pharmacists. The question is whether the system lets them do it.
What this means for health systems, pharmacists and technology buyers
For health ministries and payers, the lever that matters is payment for clinical services. Before funding automation or hub-and-spoke dispensing, decide what paid clinical work pharmacists will move into. Without that, you are funding job losses in the name of modernization. Our piece on value-based care from the payer’s chair explains why payers reward what they can measure, and pharmacist clinical services need measurable outcomes to earn their place.
For hospital pharmacy leaders, apply the lesson of alert fatigue before buying clinical AI. Ask how many alerts a tool will generate per pharmacist per shift, how it has been validated in practice rather than retrospectively, and whether it ranks risk rather than adding more flags.
For pharmacists, the opportunity is clinical credentials and documented outcomes. Prescribing qualifications, specialist training and data showing avoided admissions or GP visits are what convince payers.
For pharmacy owners, automation is worth buying when there is somewhere valuable to put the freed hours. Where there is not, it is worth being honest that the investment is a staffing reduction.
Frequently asked questions
Will AI replace pharmacists?
AI and automation will take over much routine dispensing work, including picking, labeling and routine accuracy checks. They are unlikely to replace the clinical judgment pharmacists provide in medication review, prescribing and managing complex regimens. Whether pharmacists move into those clinical roles depends mainly on whether regulation and payment systems support them.
How is AI used in hospital pharmacy?
Hospitals use automated dispensing cabinets and robots, decision support that screens orders for interactions and dose errors, and increasingly machine learning models that prioritize high-risk orders for pharmacist review. A 2023 systematic review found most machine learning prescription tools had not yet been evaluated in real clinical practice, so local validation remains essential.
What is Pharmacy First in England?
Pharmacy First is an NHS England service launched in January 2024 that lets community pharmacists assess and treat seven common conditions, including sore throat and uncomplicated urinary tract infections, and supply certain prescription-only medicines under protocol. Around 5.4 million consultations were delivered in its first 13 months, according to The Pharmaceutical Journal.
Can pharmacists prescribe medicines?
It depends on the country. In the UK, pharmacists can qualify as independent prescribers, and from summer 2026 most newly registered pharmacists will be prescribers at registration. In the US, prescribing authority varies by state and is often limited to specific medicines or protocols. Patients should ask their pharmacist what services are available locally.
What is alert fatigue in pharmacy?
Alert fatigue happens when decision support software generates so many warnings, many of them irrelevant, that clinicians begin overriding them routinely. A 2006 review found drug safety alerts were overridden in 49% to 96% of cases. It is the main reason newer AI tools should prioritize and rank risks rather than add more alerts.
The robot in the back room does not decide what kind of professional the pharmacist becomes. The fee schedule does.
Where you practice, does freeing a pharmacist from the dispensing bench create a clinician or a redundancy?
