Shift Handover Summariser
Turns a shift's scattered progress and incident notes into a one-page ISBAR-structured handover — vitals, meds, intake, mobility, falls, mood — with the source note quoted on every line and clinical risks flagged for the RN.
It reads the shift's progress notes, structures them into ISBAR categories with the source line quoted on each item, and surfaces the handful that need an RN's eyes before sign-off.
Where this works well
The slow problem this makes visible is handover fidelity. A morning shift in residential aged care generates dozens of progress-note lines per resident — obs rounds, med rounds, intake, mobility, mood — and the verbal handover to the incoming team runs on memory and notes that are often written up hours after the event. The things most likely to be lost are exactly the things that matter: a low BGL treated with juice, a refused analgesia dose, an unwitnessed fall with no obvious injury, a change from cognitive baseline.
This earns its keep where you run structured shift handovers across a residential facility with more than a handful of residents, your carers already write progress notes in the clinical record (PointClickCare, Leecare, AlayaCare or similar), and a registered nurse signs off the handover. The clearest beneficiary is that RN: instead of transcribing six hours of notes into an ISBAR shape under time pressure, they get a draft already grouped into Identify / Situation / Background / Assessment / Recommendation categories, with the riskier items pulled to the top, and spend their attention on the clinical judgement instead of the clerical assembly.
Where it works badly
It is only as good as what the note records. If your carers capture the substance verbally and write thin notes — "obs done, settled" — the summary will be thin and falsely reassuring, because there is nothing in the source for it to surface. The honest test: pull last Tuesday's morning-shift notes for one wing and ask whether a nurse who wasn't there could reconstruct what actually happened from the notes alone. If they can't, the summariser can't either, and fixing the note-writing is the real first job.
It also reads the time it is given. A fall that happened at 0635 but was written up at 1300 is summarised as a 1300 entry — the tool surfaces the timestamp in the note, and cannot know the event was earlier. And it is built for the structured progress note, not the free-text incident narrative; a long, emotionally charged incident note may get a flag ("looks like it may be reportable — RN to review") rather than a confident structured line, which is the correct behaviour, not a fault.
What it doesn't do — and shouldn't
It drafts; the registered nurse decides. It surfaces which residents had a flagged event, quotes the note behind each line, scores its own confidence, and raises an amber escalation block — but it does not approve a handover, clear a resident as stable, or sign anything to the next shift. The Approve / Edit / Reject step is a person, deliberately.
The line matters most around two decisions. First, escalation: the tool can flag that an unwitnessed fall or a refused dose needs the RN's eyes, but whether to call the GP, start neuro obs, or review the medication chart is clinical judgement that stays with the nurse. Second, reporting: it can flag that a noted event looks like it may meet a Serious Incident Response Scheme reportable category, but the decision to report — and the lodgement to the Aged Care Quality and Safety Commission within the 24-hour (Priority 1) or 30-day (Priority 2) window — is an accountable-person decision, not an automated one. Both of those are consequence-bearing and regulated, which is exactly why the human stays on them.
What your data has to look like for this to work
Concretely: progress notes need to live in a queryable clinical record, attributable to a resident and a shift, with timestamps the tool can read. Vitals need to be recorded as values (BP 122/78, BGL 3.9 mmol/L), not just "obs done". Medication events need to distinguish given from refused. Falls, skin integrity and intake need to be recorded as facts in the note, not left to the verbal handover. And the note needs to separate an observation from a recommendation — most facilities have some of this consistently and some of it not at all.
That gap is usually the real work, and it is usually about how information is captured at the bedside, not about buying a new tool. Most providers find that tightening what a progress note must contain — and getting carers writing the substance, not the shorthand summary of it — is a bigger and more valuable piece than the AI layer that sits on top. That is the work we help with first; the summariser is what becomes possible once the notes are worth summarising.
Could it summarise a note in a way that misses something clinically important and a tired nurse just signs it?
That is the failure we design against. Every line quotes the source note it came from, so the handover is reviewable rather than a black box, and clinical-risk items — an unwitnessed fall, a refused analgesia dose, a low BGL, a change from baseline — are pulled into an amber 'escalate to RN before sign-off' flag rather than buried in a tidy paragraph. The registered nurse still reads and approves it. It moves the risky items to the top; it does not decide they are safe.
Our progress notes are a mess of abbreviations, shorthand and half-sentences. Will this even work?
Partly, and the gap is the honest first job. It handles the common shorthand carers actually write — obs WNL, 4WF, BGL, cont. managed, PEG flush — but it can only structure what the note records. If falls, intake or a refused dose live in a carer's head or a verbal aside and never hit the note, no summariser recovers them. We usually start by looking at a real week of your notes to see what is and isn't being captured before any AI touches it.
Does this replace the registered nurse running handover?
No. It drafts the written handover so the RN spends the conversation on judgement — which resident to watch, what to escalate, what the notes don't say — instead of transcribing six hours of notes from memory. The clinical handover conversation, and the sign-off, stay with the nurse. The recaptured time goes back into care, not off the roster.
How current does the data have to be — what if notes are written hours after the event?
It summarises the shift you point it at, using the timestamps in the notes, so late-written notes are summarised late. That is a real limitation worth naming: a fall noted at 0640 but written up at 1300 will sit in the handover with its written time, not its event time. The tool surfaces the timestamp it found; it cannot know an event happened earlier than it was recorded.
Where does our resident data go — is it sent off to a public AI service?
No. It runs inside your Microsoft 365 tenancy on Copilot with a private retrieval layer, so progress notes stay within your environment and are not used to train a public model. Resident clinical information is sensitive health information under the Privacy Act, and SIRS-reportable detail in particular has to stay inside your incident management system — the design keeps it there.
If a note describes something reportable, does it lodge the SIRS report?
No, and it should not. It can flag that a noted event — an unwitnessed fall, an unexplained injury, an allegation — looks like it may meet a SIRS reportable category and route it to the RN, but the decision to report, and the lodgement to the Aged Care Quality and Safety Commission within the 24-hour or 30-day window, stays with an accountable person. It shortens the time to noticing; it does not make the call.
Estimated build: 3–4 weeks. Most of it is template work we've already done.
Fixed scope, fixed price, fixed dates.
The cost band reflects the engagement shape, not a per-feature line item. We work on fixed scope, fixed price, fixed dates — see the services catalogue for what falls inside each band.
Considering this for your org?
The honest place to start is a bite-sized first piece — one contained change, low risk. Tell us where it hurts; we’ll play it back, scope it, and show you what’s possible.