Writing updates for families
Turning care notes into a warm, clear update for a family member. It involves a named person, so it needs an approved tool — and a check on what should not be shared.
Our rating
Try with care
Can help, but needs the right tool, a data protection check and a person reviewing every output.
What it looks like
Before
Care notes for the week: "Tues — low mood, didn't want lunch. Wed — better, sat in garden 40 mins. Thurs — refused personal care in morning, accepted at teatime. Fri — daughter rang, chatty after."
With AI
A short update noting a mixed week, the time in the garden, that mood lifted after her daughter's call, and that some days she preferred help later in the day — with the refusal described plainly rather than omitted.
An invented example.
Families want to know how someone is. Staff want to tell them. What gets in the way is that writing a good update takes twenty minutes that nobody has, so updates happen when there is time, which is usually when something has gone wrong.
AI can turn a week of care notes into a readable update. It is useful, and it needs more care than its apparent simplicity suggests.
Why this is “try with care” and not “good to try”
Three reasons, and they compound.
It handles real health information about a named person. That means an approved tool with a data processing agreement, and a DPIA. No free accounts, no personal accounts.
It has no idea what is private. Care notes contain things the person would not want their family to know. A tool asked to summarise the week will summarise the week. It cannot distinguish “had a fall” from “cried about her late husband”, and it will include both, warmly.
It softens bad news by default. These tools write pleasantly. Asked for a warm family update, the refusal of personal care becomes “preferred to wait”, and the difficult day disappears into an upbeat paragraph. The family reads an account of a good week that was not one.
Where it helps
Consistency. Every family gets an update, not just the ones who ring. The families who never ring are often the ones who most need contact.
Warmth on a bad day. A member of staff writing at the end of a hard shift produces a terse update. That terseness reads as coldness to a family two hundred miles away.
Structure. Highlights, how they have been, anything to mention, what is coming up.
Turning shorthand into English. Care notes are written for colleagues. A family should not have to decode them.
What to check, every time
Work through in this order. It takes a minute.
- Is anything in here private? Would the person want their son to know this? Where they can tell you, ask.
- Has anything difficult been softened out? Compare against the notes. If the week included a refusal, a fall or a low patch, it should be visible in the update.
- Are other people named? Neighbours, other residents, visiting professionals. Take them out.
- Are the facts right? The tool has repeated whatever the notes said, more confidently. If the notes were wrong, the update is wrong and now sounds authoritative.
- Is this the right family, this week? No tool knows that the daughter is about to go into hospital herself.
Consent, which is the bit most often skipped
Sharing information about someone with their family needs a lawful basis. Being a relative does not create one.
Where the person has capacity, the question is simple: what do you want us to tell them, and how often? Where they do not, your Mental Capacity Act duties apply as they would to any other decision about their care, and this page does not substitute for that.
And people are entitled to change their mind. An update that has become automatic is easy to leave running after the person would rather it stopped.
Tell people it is being used
The person, and usually the family, should know AI helps draft these. The ICO requires you to tell people before a new use of their data begins, and a family who discovers it later reacts far worse than one told at the start.
Most people, told plainly that a tool drafts the update and a care worker checks and corrects it before it is sent, find that entirely reasonable. See telling the people you support and our letter template.
How to trial it
Start with two or three families who already get regular updates and are happy to say what they think.
Write the update both ways for a month — once by hand, once with the tool — and compare before sending. You are not looking for which is faster. You are looking for whether the AI version ever left out something the family should have known.
If it did that once in a month, the checking step is not optional, and your process needs to say so.