The challenge
A UK care home group with several homes plans its activities around what residents enjoy. Some sessions fill up. Others are offered week after week and quietly turned down, and the pattern stays hidden until a manager sits down with a pile of attendance sheets.
The group wanted care home activity tracking software that showed which activities residents refuse and what staff say about them. Reading every attendance list and enrichment chart by hand was slow, so it rarely happened across a whole month or a whole home.
There was a firm condition. Activity records describe older people, often frail, so any AI in the system had to do its job without ever learning who those people are.
What we built: care home activity tracking software with AI
We built the interests and activities module for the group's care platform on React, Node.js, Express.js and MongoDB, for desktop and mobile. Then we added an AI breakdown on top of the records staff already keep.
Interest-based invitations. Staff tick each resident's interests. When an activity is scheduled, the system suggests residents who might enjoy it, and staff add or remove names.
Mandatory absence reasons. After each session, staff mark every resident as attended or absent. An absence can't be saved without a reason, such as "not interested".
Enrichment chart entries. Residents who attended get an entry covering engagement, mood and staff comments.
The AI breakdown. For each activity, the AI shows how many times it was offered against how many times it was refused, and brings negative staff comments to the top.
The breakdown is a report and nothing more. It never cancels a session, edits a schedule or touches a resident's record.
How it works day to day
Picture the wellbeing coordinator planning next month's program. Instead of paging through attendance sheets, they open the breakdown and see each activity's offered and refused counts side by side, with the critical comments listed first.
A session that residents keep declining stands out right away, and the comments beside it hint at why. The coordinator and the home manager talk it through and decide what to change. The software points; people choose.
Because absence reasons are mandatory, the report's evidence is gathered as staff close each session. Nobody fills in an extra form for the AI.
How resident data stays private
Only redacted, de-identified text is passed to the AI model. That rule shaped the whole design.
Identifiers come out first. Before anything leaves the group's platform, the system removes or replaces personal identifiers: names, dates of birth, NHS numbers, phone numbers and addresses. Each resident becomes a placeholder, including when a name appears inside a staff comment.
The model sees de-identified text only. It counts offers and refusals and ranks comments without knowing whose they are.
Results are matched back in-house. The breakdown returns keyed to placeholders, and the group's own platform joins it to the right residents for staff to read.
People check the output. The home manager and wellbeing coordinator review the report before anything changes in a resident's week.
We designed this flow with UK GDPR in mind, sending the model the least information the task needs.
The result
Activities, attendance, absence reasons and enrichment entries now live in one place, on desktop and mobile. For any activity, planners can see how often it was offered, how often it was refused and the comments that explain the gap.
They get that view without exposing a single resident's identity to the AI model. We have no measured outcome figures to share, so we don't quote any. What changed is the capability: refusal patterns that used to hide in paperwork are now a report the team can act on.
What the research says
Digital care records are now the norm. The Department of Health and Social Care estimates that 81.1% of CQC-registered adult social care provider locations in England had a digital social care record as of October 2025 (DHSC provider statistics). Those records are the raw material for reporting like this.
Data security weighs on providers. In the government's 2025 adult social care provider technology survey, 41% of respondents named the cost of cyber and data security as a barrier to adopting care technology over the next five years (DHSC technology survey).
Placeholders still need protecting. The UK Information Commissioner's Office notes that pseudonymized data is still personal data for whoever holds the extra information, while pseudonymization can reduce the risks to the people concerned (ICO guidance). That is why the key to each placeholder never leaves the group's platform.
Where AI fits next
Neither of these is built yet. The AI could suggest alternative activities drawn from the interests of the residents who refuse most often. Managers could also compare breakdowns across homes to see which sessions work in one home and struggle in another.
Planning something similar?
Before you add AI reporting to care home activity tracking, settle four questions:
Which records does the AI genuinely need, and which identifiers can come out before it sees anything?
Do staff record the reason behind a refusal, or only the absence?
Who reads the AI output, and what are they allowed to change because of it?
Where does the link between placeholders and real residents live, and who can open it?
This report looks across sessions; for one resident over time, see our AI activity summaries for care homes, and for residents who rarely take part, the AI resident engagement audit. Read our guide to keeping business data private with AI, see how we approach AI automation, or explore our healthcare software work.
