Nothing here is clinical advice, and nothing here should touch identifiable patient information.
This review gives you a contact order that your team can act on, plus a separate queue for decisions you are not authorised to make. It is for outpatient booking coordinators and access managers working from de-identified waitlist, availability and rule data. Nothing here is clinical advice, and none of the prompts should receive identifiable patient information.
Prepare the three inputs
Before you paste anything, remove names, contact details, dates of birth, addresses, record numbers and free-text history that could identify someone. Replace each person with a stable case ID. Keep that same ID in every table so you can match outputs back to your approved system.
You need three documents for a useful review:
- A waitlist export with the recorded urgency category, relevant dates and current contact status.
- An appointment availability export with slot restrictions, not just dates and times.
- The current booking rules for the service, including who can approve exceptions.
Do not rely on a note that says only “urgent” or “book soon”. The review can report that wording, but it must not translate it into a booking decision unless your written rules explain what it means.
Key point
Keep clinical decisions separate
Use recorded categories and approved rules to order administrative work. Send unclear urgency, suitability and priority questions to the clinician or manager named in the process.
Start with Check the waitlist and appointment files. It prevents a common failure: producing a neat contact list from a slot table that does not show appointment type, eligibility or held-slot status. Correct the source records first. Do not patch missing values in the prompt.
Choose the prompt for the task in front of you
Use the prompts in sequence for a full session, or use one when you already have checked inputs. The output of each prompt is an operational work product, not a replacement for the booking system.
| If you need to | Use this prompt | Keep open while reviewing |
|---|---|---|
| Find missing fields and conflicting instructions | Check the waitlist and appointment files | The source exports and current service rules |
| Give staff an ordered call list | Create a prioritised contact list | The approved booking rule for each slot type |
| Route cases nobody can safely book | Separate cases needing a decision | The escalation owner and question log |
| Deal with non-contact and unused capacity | Review unused slots and contact failures | Contact-attempt and release rules |
| Hand work to the next shift or team | Prepare the booking handover record | The completed action log |
The prompts use a deliberate order of priority. They do not put a vague note above a documented rule. They also do not let the model choose between equally ranked cases when there is one slot. That is a capacity decision for the service manager unless your rules expressly settle the tie.
Note
Use the rule in force now
Check that the booking rules and slot export are current before each session. Product behaviour and available features can be version-dependent, so check the xAI documentation overview if you need current platform details.
Check the contact list before acting
Treat the proposed list as a review sheet. Compare each top-ranked row with the source exports before a coordinator contacts anyone. The fastest useful check is to sample the first five entries, every entry with an escalation, and every slot marked as soon to be released.
For each sampled row, confirm:
- The case ID appears in the waitlist export and is still active.
- The recorded urgency category is copied exactly, rather than paraphrased.
- The appointment type, service, location or remote status, duration and eligibility match the slot restrictions.
- The stated match reason points to a real rule or field.
- The contact action follows the documented attempt and release process.
Check
A usable list has evidence
Every proposed booking should show both a slot ID and a clear match reason. If either is absent, do not treat the row as ready to action.
There are several signs that the output is wrong. A list is suspect if it ranks a case where the urgency field is blank, assigns a case to a slot with an unverified restriction, or turns a free-text note into a clinical conclusion. It is also wrong if it silently excludes cases because a field is missing. Those cases belong in the exception queue, with the missing field and owner stated plainly.
Watch for duplicated case IDs, stale availability and misleading date fields. A referral date is not always a waitlist entry date. A held slot is not always available. A contact attempt is not the same as a successful conversation. Check the field headings, not just the values.
Watch out
Do not paste patient-identifiable material
The prompts need case IDs and operational fields, not names or clinical narratives. Keep the identifiable record in your approved local system.
Record decisions in the right place
Copy the confirmed action back into the approved waitlist or booking system. Record the rule used, slot ID, contact outcome and escalation owner according to local process. The handover prompt is useful for continuity, but it does not replace the system of record.
Keep the decision queue separate from the contact list. A coordinator should not have to infer whether “review with clinician” means stop, call, or book. The prompt gives one interim status for each exception so the next action is visible.
When the review does not work
Stop the booking run if the source data cannot show whether a case is eligible for a slot, or if the rules conflict. Run the data-check prompt again with the missing columns or the correct rule document. If the problem is unclear urgency or suitability, send the exact question from the decision queue to a clinician. If it is a tie, capacity exception or unclear policy, send it to the service manager. Do not solve either by changing a recorded category or making an assumption.