The AI pilot should earn the next decision
Varun Sharma | Founder perspective
An impressive demonstration can start a conversation. A useful pilot should help someone decide whether the proposed change deserves to continue. For medicines businesses and health organisations, that means understanding the work, the comparison and the responsibilities around an AI-supported process before attention becomes a purchasing commitment.
Begin with the work someone recognises
A pharmacy owner may want to understand where a service loses momentum. A medicines team may need a clearer preparation process for a value review. A manufacturer may want evidence organised around a development decision. The user, beneficiary and budget holder are not always the same person. A pilot should identify each of them and the work that changes. A general claim that AI saves time is too distant from the practical question: which complete task, for whom, under what conditions and with what effect on the next person's work?
The baseline must survive a fair comparison
Compare the proposed process with a credible existing approach. Include preparation, review, correction and follow-up, not only the seconds in which a model produces an answer. Preserve a record of exceptions and unsuccessful cases. NICE's evidence standards framework asks developers to describe impacts against current care or system processes and to consider deployment and financial consequences (NICE, 2022). The relevant framework and evidence requirements depend on intended purpose. The commercial lesson is broader: an alternative has to improve a defined job, rather than win a demonstration designed around its easiest task.
Time, capacity and cash are different outcomes
A pharmacist who spends less time preparing a review may gain capacity for another task. That does not automatically reduce payroll or create a cash saving. Training, subscriptions, checking and process change also consume resources. Keep those effects visible. A customer needs to know whether the value is throughput, service quality, reduced preparation burden or an actual budget change. An investor needs to understand which effect could support willingness to pay and whether it could persist beyond the unusually attentive environment of a first pilot.
Governance is part of what gets implemented
Where a health IT system has relevant clinical-safety responsibilities, the manufacturer and deploying organisation have distinct work to do. NHS England's review of DCB0129 and DCB0160 highlights the need to modernise their application in the context of evolving technology (NHS England, 2026). Applicability must be determined for the intended system and use. A pilot should therefore identify responsibility for checking outputs, handling incidents and deciding when use must stop. Those arrangements are part of the operating proposition; a reassuring label is not evidence that they exist.
Agree what would change your mind
Before starting, ask the buyer what result would justify a further evaluation, a purchase discussion or a stop. Choose a limited task, relevant cases and an observation period that includes ordinary variation. Define what would count as an error, an unacceptable burden or a failure to deliver. Keep the decision rule separate from enthusiasm for the product. A negative result may show that the task was poorly chosen, the workflow needs redesign or the proposed value is insufficient. Each is more useful than declaring success because a demonstration ran without interruption.
The commercial product is the useful change
This is why PHARMORIS is being developed around the idea of a Medicines Action Layer. The ambition is to help connect evidence, accountable judgement and practical action around medicines work. The initial conversation should remain specific enough to evaluate: one decision, one responsible organisation and a comparison that matters. Any evidence study or evaluation would require its own agreed scope and approvals. The value of a pilot is the quality of the next decision it enables, including the possibility that the correct decision is to change direction rather than scale.
References
NICE (2022) Evidence standards framework for digital health technologies: How to meet the standards. Available at: https://www.nice.org.uk/corporate/ecd7/chapter/how-to-meet-the-standards (Accessed: 13 October 2026).
NICE (2022) Evidence standards framework for digital health technologies: How to use the framework. Available at: https://www.nice.org.uk/corporate/ecd7/resources/evidence-standards-framework-for-digital-health-technologies-user-guide-9024813901/chapter/how-to-use-the-framework (Accessed: 13 October 2026).
NHS England (2026) National review of clinical risk management standards DCB0129 and DCB0160: supporting information. Available at: https://www.england.nhs.uk/long-read/national-review-of-clinical-risk-management-standardsdcb0129-and-dcb0160-supporting-information/ (Accessed: 13 October 2026).
A decision worth discussing
NEUVIOR welcomes a specific question from an organisation considering its next programme, process or investment. The useful starting point is the decision, the people responsible and the evidence that could change it.
