A personal agent can become a practice-wide habit when colleagues can understand, coordinate around and continue its work without depending on one enthusiastic user. More accounts would not establish that change. For Heidi Work, the organisational test is whether useful delegation becomes part of the practice's ordinary routines.
Heidi's September 2026 strategy statement, checked on 28 September 2026, describes removing clerical work while retaining human medical judgement. Its preview excludes the UK and Europe as checked on that date. This is a prospective implementation analysis, with fictional workflows rather than evidence of a current UK Heidi Work deployment.
The doctor has changed, but the surrounding practice has not
Imagine a fictional clinician who uses an authorised agent to request an already-issued hospital letter before a planned review. The clinician finds delegation convenient and can see that the request is progressing. Reception staff, however, still receive the instruction through the old task list and assume that they need to make the request themselves.
A nurse preparing the review cannot see whether the document has arrived. A colleague covering the clinician's absence sees an open task but does not know whether another action would duplicate work. The original user reports a successful experience while the surrounding team has acquired a coordination problem.
That scenario does not establish a failure of autonomous task execution. The agent could perform its authorised step correctly. The difficulty is that the practice has not agreed how the new activity becomes visible within existing routines.
The implementation question is therefore broader than whether the clinician likes the product. It is whether the people who previously received, acted on or checked the instruction can recognise what has changed and what they now need to do.
Trace the handovers rather than counting accounts
A useful workflow walkthrough would begin before delegation. Where did the instruction originate? Who would previously have received it? Which information allowed that person to proceed? Where would they have recorded progress, and how would the next colleague have known the work was ready?
The team could then walk through the same fictional task with the agent involved. The aim is not to produce a lengthy responsibility chart. It is to identify practical points where information must cross between people and systems.
In the hospital-letter example, a request acknowledgement and receipt of the actual letter are different events. A receptionist needs to know whether another request is unnecessary. The person preparing the review needs to know whether the relevant document is available. A clinician needs to know whether missing information still affects the planned encounter. One generic status may not answer all of those questions.
An implementation could succeed with modest shared information rather than a complex new dashboard. The relevant test is whether colleagues can answer their working questions reliably, using the systems they are expected to use, without asking the original champion each time.
Reluctance can reveal a workflow defect
A colleague who declines to adopt the new routine may be protecting an established process that still performs an important function. They may also be uncertain, unconvinced or simply unfamiliar with the product. Observation and conversation are needed to distinguish those explanations.
For example, an administrator might continue a manual list because it is the only place where unresolved work remains visible across shifts. A nurse might repeat a check because a completion indicator does not distinguish a requested document from a received one. A locum might avoid the process because nobody has explained which tasks it covers.
Those fictional explanations point to different remedies. Removing the list without replacing its function would create a gap. Improving status information could remove a duplicate check. A short introduction may be enough for a temporary colleague, provided the workflow is otherwise intelligible.
A useful interview would ask staff to demonstrate what they do and what information they lack. Asking only why they are not using the AI could turn a legitimate coordination problem into an apparent attitude problem.
NASSS offers a framework, not a prediction of success
The NASSS framework, published in November 2017, examines nonadoption, abandonment and the challenges of sustaining and spreading healthcare technologies. Developed through research and healthcare technology case studies, it considers the technology, value proposition, adopters, organisation, wider context and adaptation over time, alongside the condition or care context involved.
It is useful here because it broadens the inquiry beyond an individual's enthusiasm. It is not a scoring system that can predict Heidi Work's success from a product announcement, and it was not developed as an evaluation of this agent.
Applied as a set of questions, the framework encourages a practice to examine who benefits, who has to change, which dependencies remain and whether the workflow can adapt as circumstances change. The proposed implementation exercises in this article use that perspective without claiming that the framework proves any particular design will work.
The most valuable champion may understand the handovers
An enthusiastic doctor can help initiate a pilot, but enthusiasm is not the only useful implementation skill. An experienced administrator, practice manager or nurse may understand the local exceptions and cross-team dependencies that determine whether a process becomes routine.
That is an implementation hypothesis, not a rule that one professional group will always adopt first or lead best. The important capability is translating actual work into a process that other people can follow.
In the fictional document-request workflow, an administrator may know which acknowledgements are routinely mistaken for completed requests. A nurse may identify what information is needed before the review. A clinician may recognise when missing information requires a change in the plan. A workable process should incorporate those different perspectives rather than assume that configuring the doctor's interface completes implementation.
The champion should also help make the process independent of themselves. If every exception requires their personal intervention, a successful pilot may conceal a fragile service.
Test what happens when ordinary disruptions occur
A practical sustainability exercise would examine the workflow when the original clinician is absent, when a locum joins, when reception shifts change and when supplier support becomes less intensive. These are proposed tests of organisational resilience, not reports of problems encountered by Heidi customers.
The team would look for understandable records of current work, a usable route for exceptions and a way to continue through an agreed alternative when the agent is unavailable. The test is not whether all staff can reproduce the champion's technical knowledge. It is whether they can carry out the parts of the process their role requires.
Changes in staffing or local routines should also trigger review of the workflow assumptions. An instruction that worked in one arrangement may become ambiguous when responsibilities or communication channels change. Sustained adoption therefore needs a process for noticing and resolving mismatches, not just an initial training session.
A practice could document those observations through brief walkthroughs and staff feedback rather than relying entirely on usage analytics. The absence of support requests would not prove that everything is working if colleagues have quietly reverted to the old method.
A habit is shared reliance, not universal enthusiasm
Practice-wide adoption does not require every person to spend equal time in the agent. Some staff may initiate work, others may receive its outputs, and others may encounter only exceptions. What matters is whether the shared process is dependable and comprehensible.
For an individual clinician, a successful narrow workflow may already be valuable. For a practice seeking organisational change, the evidence should go further: colleagues can see relevant progress, duplicate handling is understood and reduced where appropriate, and the routine survives ordinary absences and changes.
Heidi Work could win an enthusiastic user through a convincing demonstration. Becoming an organisational habit would require the less visible work of fitting that delegation into how the rest of the practice actually operates.
Frequently asked questions
Does practice-wide adoption require every member of staff to use the agent directly?
No, different roles may initiate tasks, receive outputs or manage exceptions without using the same interface. The shared workflow needs to be understandable and dependable for everyone affected, not equally visible in each person's login history.
Should a practice appoint its most enthusiastic doctor as the champion?
Enthusiasm can help, but knowledge of local handovers, exceptions and team routines may be equally important. The appropriate champion or implementation group should be chosen for those capabilities rather than profession alone.
How can a practice tell whether adoption will last beyond a pilot?
It can observe whether the routine continues when the original champion is absent, new colleagues join and implementation support reduces. Those observations should include staff experience, workarounds and unresolved actions rather than only active-account counts.