The challenge
A UK care home group assesses every resident with structured forms covering needs, risks and preferences in detail. Turning those answers into a readable care plan summary was still a manual job for nurses. The group wanted an AI care plan generator for care homes that would start the summary for them, while leaving the final say with the nurse.
Most of what a plan needs was already in the assessment. The effort went into rewording it, checking nothing had slipped through and doing it all again at the next review.
And the next review always comes. Each form has its own schedule, so the same rewrite returned every time a resident was reassessed.
What we built: an AI care plan generator for care homes
Our team delivered the form builder and the form completion flow in 2024, using React, Node.js, Express.js and MongoDB. We then added an AI step connected to OpenAI's ChatGPT API.
What it reads. The answers in one resident's completed assessment, from a form built in the group's own form builder.
What it writes. Clearer suggested wording for those responses, and a draft care plan summary based on them.
Where the draft goes. Into the resident's care plan summary, marked as a draft. It is not part of the care plan yet.
Who accepts it. A nurse reviews the draft, amends anything wrong or missing and acknowledges it. Only then does it stand.
The AI step sits between two human checks. Before it, the assessment is signed: whoever completes it enters their password as a final signature.
Where a form needs a witness, a pop-up logs the witness's details, role and signature with date and time. After the AI step, a nurse is always the one who accepts the plan.
How it works day to day
Take a nurse after a resident's scheduled reassessment. The new answers are in, and a fresh draft summary is waiting in the resident's care plan. It is written from what the assessment actually says, not from a blank page.
The nurse opens the draft beside the form it came from. They check each point against the answers, correct anything the AI got wrong, add what only they know about the resident and acknowledge it. Until that moment, the draft carries no weight in the record.
When the form's review date comes around again, the cycle repeats from the new answers, and the nurse checks a new draft.
The result
Care plan summaries now start from a draft built on the resident's own assessment answers. The nurse's time goes into checking and correcting, not retyping.
Nothing in a draft counts until a nurse has reviewed and acknowledged it. The assessment underneath carries its author's signature and, where required, a witness. We have no measured time savings for this feature, so we don't quote any.
What the research says
Care teams are already trying generative AI. A University of Oxford collaboration brought together more than 70 people and organizations from the care community. It lists help with generating care plans among the current uses of generative AI in adult social care (Institute for Ethics in AI, 2025). A reviewed step inside the care record gives that use a clear owner.
Staff find their own AI when none is offered. Microsoft and LinkedIn's 2024 Work Trend Index found that 75% of knowledge workers use AI at work. It also found that 78% of those users bring their own AI tools (Microsoft Work Trend Index 2024). Building the draft step into the group's own platform gives nurses an approved route instead of a personal chatbot.
Nearly right is not the same as right. In Stack Overflow's 2025 Developer Survey, 66% of respondents named AI answers that are almost right, but not quite, as their top frustration with AI tools (Stack Overflow, 2025). That is why a draft only counts once a nurse has reviewed, amended and acknowledged it.
Where AI fits next
Neither idea below is built yet. The nurse could see what changed since the last acknowledged plan, so a review focuses on the differences. The AI could also flag assessment answers that conflict with recent chart entries, with every flag going to the nurse to decide rather than into the plan.
Planning something similar?
Before you add AI care plan drafting to your care records, settle these questions:
Which assessment answers should feed the draft, and which should stay out of it?
Who may acknowledge a draft, and how is that recorded?
Can staff tell a draft from an acknowledged plan at a glance?
What happens to an acknowledged plan when the resident is reassessed?
The drafts live inside the care planning platform we delivered for this group, next to an AI question builder for form authors and AI writing help for staff completing forms. Our AI automation page explains how we add AI with a human sign-off, and our healthcare software page shows related builds.
