Did the owner ask for it, or has it been done?
You arrive to cover the afternoon visit and find two notes: “check water and litter tray at each visit” and “water changed this morning”. Was the litter tray checked too? The instruction says what was requested; the second note records just one task for a particular visit. Do not fill the gap with “all done”, but do not assume that a task should be repeated without asking either.
This guide proposes a text AI trial to prepare handovers between home-visit pet sitters. The intended output is an internal record separating instructions, recorded actions and questions for each visit. The sitter checks and decides; the assistant does not observe the animal or certify its care.
Keep a visit log, rather than a convincing story
The Professional Pet Care Association describes a care sheet where sitters record their arrival and mark tasks as they complete them. It explains that teammates rely on the log to check what has been done. This is a practical reference for keeping tasks tied to visits, rather than replacing the log with a summary that loses detail.
Time To Pet documents post-visit reports with configurable questions. Walkies publishes tutorials on recording actions, completing checklists and sending a home-visit report. Those pages describe their own systems; we have not tested them here or established any connection to AI or Adaris.
Before adding AI, compare a shared sheet and a checklist with your actual problem. If ticking off tasks is enough, start there. The trial in this guide is relevant when you need to review narrative notes and prepare handover questions while retaining their sources.
Fictional case: two visits and a gap you must preserve
The animals, visits, identifiers and texts below are invented. The final column is our manual reference for correcting a future draft, not an observed AI response. It contains no addresses, home access details, treatments or clinical instructions.
| Record | Fictional input | Handover review reference |
|---|---|---|
| I1 · current instruction | For the cat Nube: check water and litter tray at each agreed visit. | Retain as an instruction. Do not mark any task as completed. |
| V1 · morning visit | The sitter records: water changed. No note about the litter tray. | Water changed according to V1. Litter tray not recorded: clarify with the sitter, without claiming completion or omission. |
| M1 · owner message | For this afternoon’s visit only: no need to water the plant. | Exception for today and that visit; it does not cancel other tasks or visits. |
| V2 · afternoon visit | Visit planned; no completion report yet. | Keep as planned. Do not tell the owner a visit has taken place. |
| N1 · handover note | The morning sitter left a question about where to put the litter scoop. | Open question. Do not invent a location or present it as an owner instruction. |
Prepare the handover in four steps
First, separate current instructions from each visit’s notes. In the trial, I1 is the instruction and V1 is a morning record. Keep the original text beside its identifier. A visit note does not by itself change the agreement with the owner.
Second, open a row for each visit and task. In V1, water has a recorded action and the litter tray has no record. “Not recorded” means documentary information is missing; it proves neither completion nor failure to act. The next action is to clarify with the responsible person, not ask AI to complete a plausible story.
Third, preserve the scope of changes. M1 concerns watering the plant this afternoon only. Do not extend it to the cat’s water, the litter tray or later days. V2 remains planned until there is a record for that visit; I1 is not evidence of completion.
Fourth, give the next sitter the specific questions. A manual reference draft would read: “Morning V1: water changed according to the note; litter tray not recorded, clarify with the morning sitter. Afternoon V2: planned, no report yet. M1 waives watering the plant this afternoon only. N1 asks where to put the scoop; answer missing.” Review it before use; it is not a sent message.
A trial instruction that keeps gaps visible
Try only the invented records: “Prepare an internal pet-sitter handover using I1, V1, M1, V2 and N1. Separate instruction, planned visit, recorded action, exception and question. Cite each datum’s identifier and keep each task tied to its visit. Do not turn instructions into completed actions. Where a record is missing, write not recorded and request clarification from the responsible person; do not infer failure or recommend repeating care. Preserve M1’s time scope. Do not invent times, locations, care actions or statuses. Return a draft for human review. Do not send messages or change the schedule.”
This is a trial proposal, not a validated capability. Check every sentence against the inputs. If it says “all Nube’s needs attended to” or “the afternoon visit has finished”, reject that sentence and log the error. Friendly wording does not supply the missing check.
What the person receiving the record must check
Check that each action retains its visit, task and source. Do not move “water changed” from morning to afternoon or use a note about one task to complete the others. Preserve N1’s question even if it disrupts a tidy summary.
Do not use the draft to decide feeding, medication, diagnosis or urgency. This example concerns records and coordination. Clarify care questions with the appropriate responsible person; for a real incident, follow the agreed professional procedure and contact the relevant person without waiting for AI to write a summary.
Do not add keys, alarm codes, addresses or real conversations to the trial. The example data is fictional. Real use needs permission, appropriate terms and information limited to what is necessary. The record alone does not prove that someone entered the home or performed a task.
A small pilot with a reference prepared first
Prepare several invented handovers: a recorded task, a task without a note, an exception for just one visit and a change awaiting confirmation. Write the expected manual record first. Then compare the assistant’s draft with that reference, without using it to organise real care.
Count invented actions, confused visits, overextended exceptions and lost questions. Record preparation and correction time against the manual sheet. These are proposed measurements: we have not run the pilot or observed savings, accuracy or care improvements.
The linked gardening guide develops another document-handover trial, with tasks by zone and conflicting notes. To test the pet-sitter record, use this guide’s fictional case. If reviewing an AI draft takes more work than maintaining a clear checklist, keep the checklist.
From a manual trial to a bounded integration
In a manual trial, you copy invented texts and check the output. Automation would require agreement on which records it may read, how each visit and task is identified, who confirms a correction and what approval is needed before anything is sent. Connecting your calendar or application requires assessing access and feasibility; we assume no compatibility with the cited software.
Adaris develops custom software, CRM and AI automation case by case. We can explore an internal handover record that retains original logs, displays questions and requires approval before communication. This is a bounded development proposal, not an existing proprietary pet-care application. Our working language is Spanish; we work remotely.
Related services
Sources
Does the next sitter have to piece together every visit?
Tell us how you record instructions, tasks and questions. We can assess an internal record with sources and human approval. Our working language is Spanish.