Care Connection
Workforce & rostering · live in the partner trial

Find the roster gaps. Get to the cause.

Rostering intelligence reads your rostering system’s own export and shows where shifts go unfilled, why, and what it is costing. Deterministic analysis does the counting. AI explains it in plain language, with every number checked against the evidence before you see it.

See it on your export The cash-leakage picture

Review-only by design. Works from an export. De-identified before anyone sees a screen.

From a live deployment

Eight homes. Ninety days. One leak hiding in plain sight.

The figures below come from a live deployment across eight residential homes, de-identified. Yours will differ. The pattern probably will not.

The upstream leak
99%

of vacant duties were never offered to anyone.

4,617 of 4,643 unfilled shifts in ninety days. The bank and casual pool never saw them.

The process that works
87%

of shifts that were broadcast got filled.

The fill process is not the problem. The leak sits upstream of it.

Designed-in gaps
56%

of vacant duties sat on template lines with no funded person.

Placeholder lines no roster action can fill, inflating every gap report.

Template versus contracts
41%

of cancelled duties were ones the work patterns could never deliver.

5,141 duties cancelled as unachievable before anyone was sick.

Premium cover
34%

of worked shifts came from the bank.

At one home it was half. Recurring bank hours are contracted lines in disguise.

Firefighting by text
111,633

SMS offers sent to fill 2,853 shifts.

Thirty-nine messages per acceptance, and half of all assignments made by hand.

De-identified figures from one provider’s rostering export, last ninety days at the time of analysis. They describe that deployment, not the sector, and are not a recoverable savings rate.

The leak in hours and dollars

An unfilled shift is never free.

It is paid for somewhere else: in premium cover, in care minutes you have committed to deliver, and in leaders’ time spent plugging the roster shift by shift.

Premium cover

Bank, agency, casual loading and overtime

A gap found on the day is filled at the highest price available. A gap found two weeks out is usually filled at ordinary rates. Offers sent one to two weeks ahead were accepted at six times the rate of offers sent inside 24 hours.

Care minutes at risk

Funded minutes you did not deliver

Every unfilled care duty is time you are funded to provide and have committed to report. The same screen that shows the gap shows which grade and which shift it sits in, so the care-minute exposure is visible before quarter end.

Management time and fatigue

The hidden roster

Half of all assignments made by hand, 4,024 rule warnings in ninety days, and tens of thousands of texts. That is leaders’ time, and it is fatigue exposure for the same people who keep saying yes.

A financial control, not a staffing cutThe target is cost that added no care value: shifts never broadcast, template lines nobody can fill, cover booked late. Care requirements stay central to every option, and a person decides every action.
How it works

Facts first. Then the AI explains.

Your rostering export becomes a de-identified fact store: every unfilled and cancelled duty, every offer and reply, every absence day, rule warning and bank booking. Deterministic tools do the counting and rank ten root-cause hypotheses. The AI reads that evidence and writes the explanation.

AI narrates, never computesEvery figure in an answer is checked against the evidence before it reaches the screen. Anything the model says that the data does not contain is flagged, in the open.
De-identified at loadStaff appear as numbers only. No names, contact details or health information reach the database, the model or the screen.

  1. Load the export

    Your rostering system’s standard export becomes the fact store in about ninety seconds. Nothing is installed in your systems.

  2. Rank the causes

    Ten hypotheses are scored for every home: never offered, unfunded template lines, unachievable patterns, absence pressure, grade scarcity, bank dependence and more.

  3. Ask in plain language

    “Why are we short on evenings?” routes to the right analysis, runs it, and returns a grounded answer with the drill-downs behind it.

  4. Act and track

    Save the finding, agree the lever and the owner, and watch the same screens the following month.

What you see

Six screens, one question: where is the gap, and why?

Illustrative screens from the deployment above, de-identified. Facility names are fictional.

Gap radar: every home ranked by gap rate with never-offered share, absence, warnings and bank share
Gap radar. Every home on one screen: gap rate, what is open ahead, never-offered share, absence, rule warnings and bank share. Click a row to open its root-cause board.
Root cause board: ranked hypotheses with scores, evidence and drill-down buttons
Root cause. Ten hypotheses scored and ranked for one home, each with its evidence, the lever that addresses it and a drill-down.
Fill funnel: vacant to offered to accepted to filled, with acceptance by lead time
Fill funnel. Vacant, offered, accepted, filled. Acceptance by lead time shows when to broadcast.
Ask intelligence: a plain-language question and a grounded answer with headline, evidence and ranked causes
Ask intelligence. A plain-language question, a grounded answer, and a badge confirming every number was verified against the evidence.
AI enterprise brief generated across all homes
AI enterprise brief. The organisation-wide answer: systemic causes, local causes, outliers and the levers to pull first.
Twelve-month trend of vacant duties, gap rate, sick days, warnings and bank bookings
Trend. Twelve months of vacant duties, gap rate, sick days, warnings and bank bookings, with forward roster periods marked.
Across the organisation

Systemic or local? Now you can tell.

The same root-cause ranking runs for every home, then the numbers are pooled. A cause that scores high at most homes is systemic and belongs to head office: the template, the broadcast rule, the leave policy. A cause that scores high at one or two homes is local and belongs to a conversation with a manager.

In the deployment above, four causes were systemic and three were local. Two homes were outliers on gap rate and bank share. That is a different board paper from “agency spend is up”.

Score 60+30 to 59Under 30
Which causes are systemic and which are local?Which template lines are vacant every week?Is sickness clustering around weekends?How reliant are we on the bank, and on how few people?What if three care workers are on leave for 14 days?
Heat map of root-cause scores by facility, labelled systemic or local
Cause by facility heat map. Systemic rows belong to head office; local rows belong to one manager.
Guardrails

Built to be trusted in a board pack.

Review-only

Nothing is written back to your rostering system. People decide; the platform prepares the evidence.

Every number cited

Each screen names its source tables and definitions. Each AI answer is checked against its evidence.

De-identified

Staff numbers only. No names, contacts or health detail in the database, the model or on screen.

Hypotheses, not verdicts

Cause scores rank what to investigate first. They are evidence for a conversation, not causal proof.

Talk to us

See it on your own export.

A two-week pilot. You provide a de-identified rostering export; we return the gap radar, a root-cause board for every home and an enterprise brief. Scope and data handling are agreed first.

  • Works from your rostering system’s standard export.
  • De-identified before load. Staff numbers only.
  • Review-only. Your people decide every action.

Start the conversation

We reply within two business days.

We use your details only to reply to this enquiry. No newsletters. See our privacy policy.

Thanks — we'll be in touch

We reply within two business days.