Most clinical software is designed for billing and compliance, then handed to clinicians as a data entry obligation. The predictable result is workarounds: paper notes typed up later, shared logins, and fields filled with whatever passes validation fastest.
Count the clicks, then remove them
In a ten-minute consultation, every additional interaction competes with the patient. Counting clicks for the most common task and halving the number does more for adoption than any feature. Defaults, smart pre-fill and remembering the last selection are unglamorous and decisive.
In a ten-minute consultation, every extra click is taken from the patient.
Design for interruption
Clinical work is interrupted constantly. A form that loses input when a user navigates away, or that cannot be saved half-complete, will be worked around within a week. Persist state continuously and make resuming trivial — this is the difference between a system used and a system tolerated.
Integration is most of the project
New clinical software almost never stands alone. FHIR where available, HL7 where not, and file-level integration where the vendor offers neither. Assessing this early matters because it frequently determines the architecture and it is the most common source of schedule overrun in healthcare builds.
Consent and access are data-model concerns
Treating consent as a checkbox produces systems that cannot answer who was permitted to see what, and when. Model consent as versioned, granular and auditable, and enforce access against it in the data layer, so the answer to an access question is a query rather than an investigation.
AI belongs beside the clinician
Drafting notes from a consultation, summarising a long history, surfacing relevant prior results — these save real time and keep the judgement with the clinician. Systems that present a conclusion without the reasoning or the source get distrusted quickly, and distrust is not recoverable in this setting.
Pilot with the sceptics
Piloting with enthusiastic early adopters tells you the best case. Piloting with the busiest and most sceptical clinician tells you whether it works. If it survives that, it will survive the department — and if it does not, you have learned it before rollout rather than after.
