Care Inspectorate Self-Evaluation: Draft the Evidence Summary With AI

A self-evaluation for a small care service answers three questions under the framework's headings: how are we doing, how do we know, and what are we going to do now. AI can draft the evidence summary from records you supply, but the provider writes and owns every judgement and grade, and checks each statement against a record.

The Care Inspectorate grades quality indicators on a six-point scale, from 1 (unsatisfactory) through weak, adequate, good and very good to 6 (excellent). Your self-evaluation is more convincing when the grade follows from the evidence than when it is asserted. A service that says "good" and shows an audit, a changed practice and a resident's own words is more credible than one that says "very good" with nothing attached.

Follow me on Instagram@sagnikteaches

What the framework asks, in plain terms

For care homes for adults and older people, the quality framework (published 1 April 2022; its web pages were last updated on 5 October 2026) is organised around six key questions: how well do we support people's wellbeing; how good is our leadership; how good is our staff team; how good is our setting; how well is our care and support planned; and what is our overall capacity for improvement. Under each of the first five there are up to five quality indicators (key question 3 has three), and each indicator lists key areas, short bullet points describing the practice it covers. Inspectors look at a selection of indicators, chosen by the type of inspection, the quality of the service, the intelligence held about it and any risks they identify.

Connect on LinkedInSagnik Bhattacharya

The framework sits within the Health and Social Care Standards, which have applied since 1 April 2018. Other adult services, including care at home, other support services, housing support and nurse agencies, have their own frameworks, so use the one published for your registered service type, read its key questions and indicators, and check that you have the current version on the Care Inspectorate's website. The rest of this tutorial uses a small care home as the worked service, and the method carries over to other types with the headings swapped.

Subscribe on YouTube@sagnikteaches

The care homes framework also lists core assurances: areas inspectors always look at, because they are early signs of harm when care goes wrong. They cover the registration certificate, insurance and written aims and objectives; adult protection; infection prevention and control; medication; management of people's finances; accident and incident records; the development or improvement plan; complaints; staff recruitment; the physical environment; equipment and safety maintenance; planned care; management oversight; and meaningful connections. A self-evaluation that skips these, however polished the rest, leaves the obvious questions unanswered, so fill in the core assurances sheet first and start on the indicators after.

Fitting the draft into the Care Inspectorate's own tool

You do not need to invent a format. The Care Inspectorate's self-evaluation tools page for care homes (last updated 27 October 2026) offers a blank self-evaluation tool in Word, a worked example, a core assurances template, an improvement plan template, a focus group plan and sets of example questions for people experiencing care, relatives, staff and visiting professionals. The blank tool asks for the service, the manager, the date, the key question, the quality indicator and the key area, then repeats three boxes for each key area: how are we doing (with the six grades from unsatisfactory to excellent to choose from), how do we know, and what are we going to do now.

That layout tells you how to size each AI task: one key area at a time, never a whole key question in one prompt. Quality indicator 3.2, staff have the right knowledge, competence and development to care for and support people, has three key areas: staff competence and practice, staff development, and supervision and appraisal. Three key areas means three short drafts, each pasted into its own box, and three grade boxes that you tick yourself. The example question sets are worth using too. Paste the staff questions and ask the AI for a ten-question paper version for a team meeting, keeping the wording close to the original so each answer maps back to the indicator it came from.

Set up the evidence folder the AI will read

AI is only as good as the material you give it. Before any prompt, build a folder (or a shared drive area) with one sub-folder per key question and put in dated, real records. Remove resident names and anything identifying, or replace them with initials the service already uses internally.

Name each file with a date and a topic ("2026-09 medication audit summary"). Three to five good items per area is enough. If a folder has nothing in it, that is your first finding, not a problem for the AI to paper over.

A 14-bed home works through quality indicator 3.2

Illustrative figures for a small, made-up care home. The manager is working on quality indicator 3.2 under the staff team key question, which covers competence, development, and supervision and appraisal. Her evidence folder holds a training matrix export (1 October), a supervision log for the last six months, three team meeting notes and a staffing review.

The training matrix shows 21 staff, with mandatory training 90% in date overall but dementia awareness at 71%. The supervision log shows 19 of 21 staff had a supervision in the last 8 weeks, and the two missing are on long-term leave. The team meetings show that after a medication incident in July, the senior team ran a short refresher in August. These are the raw facts. The job of the summary is to arrange them under "How are we doing", "How do we know" and "What are we going to do now".

The prompt that drafts the evidence summary

This prompt keeps the AI in a narrow role: arrange and phrase what you have supplied, flag gaps, and leave the judgement and grade to you.

You are helping the manager of a small care home in Scotland write
a self-evaluation evidence summary for quality indicator 3.2
("Staff have the right knowledge, competence and development to
care for and support people"), under key question 3.

Use ONLY the EVIDENCE below. Structure the summary under:
1. How are we doing? (what the evidence shows, strengths and weaker areas)
2. How do we know? (list each source with its date)
3. What are we going to do now? (actions, owner, date: only if the
   evidence points to a need; otherwise write [MANAGER TO DECIDE])

Rules:
- Do not suggest or give a grade. The provider decides.
- Do not add any figure, date, quote or practice not in EVIDENCE.
- Mark anything unsupported or unclear as [NEEDS EVIDENCE].
- Plain English, 300 words maximum, past and present tense as the
  evidence supports.

EVIDENCE
- Training matrix 1 Oct: 21 staff; mandatory training 90% in date;
  dementia awareness 71% in date; moving and handling 95%.
- Supervision log 1 Apr to 30 Sept: 19 of 21 staff supervised in
  last 8 weeks; the 2 outstanding are on long-term leave.
- Team meeting notes 12 Aug: refresher on medication recording
  after a July incident; 12 of 16 on duty attended; the rest briefed.
- Staffing review 20 Sept: no unfilled shifts; 3 agency shifts in
  the last month.

An illustrative output: "How are we doing? Most mandatory training is in date, with 90% overall, though dementia awareness is lower at 71%. Supervision is regular: 19 of 21 staff were supervised in the last 8 weeks, and the two others are on long-term leave. After a medication recording incident in July, a refresher was delivered on 12 August, attended by 12 of 16 staff on duty, with the remainder briefed. Staffing has been stable, with no unfilled shifts and 3 agency shifts last month. How do we know? Training matrix (1 Oct); supervision log (1 Apr to 30 Sept); team meeting notes (12 Aug); staffing review (20 Sept). What are we going to do now? [MANAGER TO DECIDE: plan to raise dementia awareness from 71%]."

What you would fix: the AI was disciplined and did not grade. You add the action yourself: book dementia awareness for the 6 staff out of date (29% of 21) by 30 November, owned by the deputy, with the tracker as the check. You also move the staffing line out: no unfilled shifts and agency use belong to quality indicator 3.3, staffing arrangements, so that sentence goes on the 3.3 sheet. You also notice the draft does not say whether the August refresher worked, so you pull the next medication audit as evidence of its effect. A summary that proves a change made a difference carries more weight than one that lists activity.

Judgement and grade stay with the provider

The six-point scale is a judgement about quality, and it belongs to the person running the service. Write your own view first, in a sentence, before you look at any AI output: "I think staff team is good, not very good, because training is strong but dementia awareness is low." Then compare it with the evidence. If the evidence does not support your first view, change the view, not the evidence.

Ask the AI to challenge you, not to decide. A useful second prompt is: "Here is my draft summary and the evidence. List three questions an inspector might ask that this summary does not answer." The reply might include: what is the plan for the 29% out of date on dementia awareness, how do you know supervision improves practice, and how are agency staff inducted. You answer each from your own knowledge and records. The mock inspection tutorial uses the same idea at service level.

Using quality illustrations without copying them

Each quality indicator has illustrations of what very good and weak practice can look like. They are examples of what an inspector might see, not a checklist of phrases to repeat. Read them to understand the level you are aiming for, then describe what your own service does in your own words.

A tempting shortcut is to paste the illustration text into the AI and ask it to rewrite your summary to match. Do not. It produces a polished description of a service that may not be yours. Instead, use the illustration as a question: "Do we do something like this, and can we show it?" If yes, the evidence goes in the folder. If no, it goes in the action plan.

Three more worked checks from the same home

First, wellbeing: the home has 11 residents' meeting action lists and a you-said-we-did board. The AI summary counts actions raised and completed from the lists you give it: 12 raised, 8 done, 3 explained, 1 open. The manager looks at the open item before it goes into the self-evaluation. If you want to set up that board, the residents' meeting minutes tutorial shows how.

Second, the setting: the maintenance log shows 46 requests in six months, 41 closed within a week, and five waiting on a contractor. The summary reports both numbers and the reason for the five, which a good manager checks with the contractor before saying anything. Third, care and support planned: a sample of six care plan reviews shows all dated within the last month, with four mentioning the person's own words about their goals. The AI flags that two of the six do not, which turns into an action to improve the wording at the next review, with the deputy as owner.

The summary that said more than the file

A manager pastes a training matrix and a few headings and asks the AI to "write the staff team section, we think we are very good". The draft arrives with a line saying "staff report feeling well supported, as shown in recent surveys". There is no survey in the folder. The AI picked up the usual shape of such statements and produced one.

The manager catches it on the read-through and asks herself the question the prompt already asks: where is the record? She then runs a short staff pulse check, three questions on a paper slip, gets 17 replies and writes the line from that instead. The lesson is that a good-sounding sentence with no source is a gap you have filled with fiction. Use the [NEEDS EVIDENCE] marker and treat every such tag as a task.

Turning findings into an action plan that gets finished

The third question, what are we going to do now, is where many self-evaluations go vague. "Improve dementia training" is not an action. An action names a task, a person, a date and the evidence that will show it is done. Ask the AI to convert your own bullet findings into that shape, and tell it to leave owner and date blank if you have not given them.

All of this is made-up for illustration, but the shape is what matters. Notice the last column: it makes the action testable. At the next monthly slot you open the evidence and either tick it or carry it forward with a reason. The action list then feeds the next version of the summary, so the "what changed" story writes itself.

Bringing in the voices of people who use the service

A self-evaluation that only quotes records misses the point of the framework, which is how people experience the care. Gather feedback in ways that suit the people involved: a short conversation at a residents' meeting, a pictorial survey, a note from a relative, a comment in a thank-you card. Record the source and date of each, and get permission before quoting.

AI is useful for finding themes in a pile of comments, as long as you strip names first. Paste 25 short comments and ask for the topics that recur, with a count and a trimmed example for each, plus a separate list of anything that sounds like a concern needing follow-up. A sample result, in illustration: "Food and choice (9); staff kindness (8); activities in the evening (4); noise at night (3); one comment about waiting for help, flagged for the manager." You read the originals, not just the themes, and you act on the flagged comment through your normal concerns route before it appears in any summary.

When you write the evidence line, quote only what a person agreed to share, and make clear what it supports. "One relative wrote that her mother 'looks forward to the Thursday quiz'" is better evidence for activity than a general claim, and it is true to the source. Never ask the AI to invent a representative quote or to reword one so that it sounds more positive.

Writing honestly about a weaker area

Every service has weaker areas. An evaluation that names one, shows how you found it and says what you are doing is more credible than one that finds nothing wrong. Say the evidence shows dementia awareness at 71%: write that plainly, explain why (three new starters and two renewals missed during a busy month), and set out the plan.

The AI can soften such a passage if you let it. Watch for phrases like "minor gap" or "some opportunities for development" when the figure is 29% out of date. Ask for the plain version: "State the figure and the reason. Do not use words like minor, slight or opportunity." Then read it as an inspector would, asking whether it leaves them guessing. If your own judgement is that a key question needs a lower grade than you hoped, write the lower grade and the plan to raise it.

A realistic time budget for a small service

For a small service, the first full pass is the heaviest. Illustrative estimates for a manager with a deputy: half a day to build and tidy the evidence folder; about 45 minutes per key question to run the prompt, read the draft against the sources and write your own judgement; an hour to build the action plan; and an hour for a second person to read the whole thing. That is around two days spread over a month, against what many managers describe as a rushed week before inspection.

After that, the monthly slot of 30 minutes keeps it alive. If your numbers are different, adjust, but keep the shape: collect continuously, summarise one key question a month, review actions every month. Time saved by AI is in the drafting and tidying; the time you must keep is for reading the evidence and making the call.

Using the summary when an inspector visits

An inspector may ask how you know something is good. Your evidence folder and summary are the answer. Print the key question summaries, keep the folder easy to open and make sure the deputy can find any file in a minute. If they ask about something you have flagged as a weakness, you can say what you found, when and what you did, with the dates. If they point to something you had not seen, record it and add it to the log; a service that learns from an inspector's observation shows exactly the capacity for improvement the framework describes.

Be careful not to present the AI-drafted summary as if it were an independent check. It is your own account of your service, organised with help. The honest framing, if asked, is that you use a drafting tool to organise records and that every statement is checked by you against the source. That is the same standard you would apply to a summary written by a new deputy.

Keeping it current rather than writing it once a year

A self-evaluation written the week before an inspection is a snapshot. One that you update each month from your routine records is a live picture. Add a 30-minute slot to the monthly management meeting: drop the new audit, training export and meeting notes into the folder, run the same prompt for one key question per month, and review what changed since last time. Over six months you cover every key question, and each summary has a date.

The record of what changed and why is what makes the capacity-for-improvement question easy to answer. Keep a simple improvement log with columns for date, what we found, what we did, what difference it made and the evidence. Entries from the earlier example might read: "12 Aug, medication recording incident, refresher delivered, September audit found no recording errors in 60 checks". Pair this with the monthly audit checklist so the evidence arrives without a special effort.

Data, confidentiality and who checks the draft

Self-evaluation evidence often contains information about residents and staff. Keep names out, use summaries and counts, and paste only what the prompt needs. Use a business plan with a vendor statement that content is not used for training, and follow your own organisation's data-protection rules. If your provider has approved tools, use those; if not, ask before you paste anything.

A second person should read the final version against the evidence folder: the deputy, a director for the provider or a trusted peer from another service. They check that every figure traces to a file, that no grade is stronger than the evidence allows and that each action has an owner and a date. If you would like help building the folder and routine for your service, my AI implementation consultation covers it.

Self-evaluation questions from small Scottish care services

Can AI write the self-evaluation for me?

It can draft the evidence summary from material you supply, but the judgements and grades belong to you as the provider. Use it to organise evidence, spot gaps and tighten wording, then check every statement against a record. Never let it choose a grade or invent evidence.

Which quality framework should a small service use?

It depends on the service type. The Care Inspectorate publishes separate frameworks, for example for care homes for adults and older people, and for support services such as care at home. Use the framework for your registered service type and check the current version on its website.

What does the self-evaluation have to answer?

The Care Inspectorate's guidance frames self-evaluation around three questions: how are we doing, how do we know, and what are we going to do now. Your summary should show evidence for the first two and a dated, owned plan for the third.

How much evidence is enough for each indicator?

Enough that someone who does not work in your service can see what you do, how you know it works and what you changed. Three to five well-chosen pieces of evidence per indicator, each dated and traceable, beat a folder of forty documents nobody has read.

Further reads

• Care Home Medication Audit: A Monthly Template and AI Action Log — A source of dated audit evidence for the wellbeing questions.
• Care Home Mandatory Training Tracker: Renewal Dates and AI Overdue Alerts — Training completion evidence for the staff team question.
• Care Home Activities Calendar: Plan a Month With AI and Resident Input — Evidence of meaningful activity and resident choice.
• Care Home Policies and Procedures UK: Keeping Them Current With AI — Show that policies are reviewed and staff have read them.
• Care Staff Supervision Template: Notes and Action Logs With AI — Supervision records support leadership and staff evidence.
• Can AI Write a Nursery Self-Evaluation Form From Your Evidence? — The same evidence-first method for a nursery.
• AI Tools and AI Development: The Complete 2026 Guide — the AI hub, including every tutorial in the AI-for-business series.

Sources: Care Inspectorate, A quality framework for care homes for adults and older people (published 1 April 2022, pages updated 5 October 2026), its key question 3 and core assurances pages, the adults social care quality frameworks list and the self-evaluation tools and guidance page for care homes (updated 27 October 2026), all read 4 November 2026. Check the Care Inspectorate website for the current framework and tools for your service type.