Summary
The AI Quality Assistant helps college staff review quality. It analyses your achievement data, compares one period with another, checks learner targets against SMART criteria and reviews schemes of work. It works only from the files you upload or paste, and sets out strengths and areas for improvement from your own evidence. This guide takes you from a first upload to a finished analysis you can use in quality reviews.
Before you start
- Your data as a spreadsheet or CSV. This matters most. A standard achievement report works best. The assistant recognises the usual columns: starts, leavers, achievement, retention and pass rates, national averages and your gap to them, and absence. Without national benchmark columns, it cannot compare you with national figures.
- At least three years of figures, for a trend. The three-year trend needs the current year and two earlier ones.
- Both data sets, for a comparison. For example, last year’s report and this year’s.
- The targets or scheme of work to review. Upload the file or paste the text. Both work equally well.
How it works
These are deliberate. Knowing them up front saves second-guessing.
Some buttons ask before they start. Data Comparison always asks you to confirm that all your files are uploaded, then waits for your reply. SMART Target Reviewer and Scheme of Work Reviewer ask for your document if you have not supplied one. Answer, and the work continues from your reply.
It only reports what your data supports. You may get fewer than five strengths, or a trend section that is one line long. If your data does not cover something, it says so rather than filling the gap. It adds nothing you have not supplied.
It never names individuals. Even where your data or targets include names, the response uses learner IDs, roles, cohorts or course codes instead.
Try it now
- Open the AI Quality Assistant from the Assistants list on the left-hand menu.
- Paste three or four learner targets into the chat box, each with a learner ID.
- Select SMART Target Reviewer.
You should get one table with a row for each target. Each row is marked yes or no against Specific, Measurable, Achievable, Relevant and Time-bound, with a rewritten target where one falls short.
You will know it worked when every target that falls short has a reason and a rewritten SMART version you could give back to the tutor.
The four workflows
Data Analysis
- Use it when: you want a quality review of one data set, such as this year’s achievement report.
- You get: a title that sums up what the data shows, then Key Strengths (up to five, each with an example from the data), a Three Year Trend and Areas for Improvement. Each section ends with a table of course code, measure, your value, the national benchmark and the gap.
- Note: upload your data first, then select the button. With fewer than three years of data, the trend section just names the years your data covers.
Data Comparison
- Use it when: you want to see what has changed between periods, for example last year against this year.
- You get: a statement of which file it treats as the earlier period, a table of changes for each measure, key strengths, key insights, and areas for improvement with their evidence.
- Note: it always asks you to confirm that all your files are uploaded before it starts. It compares all the files in one go, and can take more than two.
SMART Target Reviewer
- Use it when: you want to check tutor-set targets for learners.
- You get: one table covering every target, marked against each SMART criterion, with a reason and a rewritten target where one falls short, or “None” where no change is needed.
- Note: if you have not uploaded or pasted any targets, it asks for them first.
Scheme of Work Reviewer
- Use it when: you want a scheme of work checked before it is used.
- You get: a section-by-section review against 13 required sections, including intent, units covered, learning outcomes, assessment details, the weekly breakdown and independent study. Each section is marked present, missing or partial, with suggested text where something is missing or weak. It ends with the most important gaps to fix first.
- Note: it checks against a general standard, not your college’s own template. A section your template does not use, such as career pathway badges, will show as missing. Treat these as prompts to consider, not failures.
Walkthrough: reviewing your achievement data
- Export your achievement report from your college systems. The assistant cannot reach those systems, so everything it analyses comes from the file you give it.
- Upload the report using the Upload button in the top menu bar, or drag it into the chat. Include three years of figures if you have them, so you get a trend.
- Select Data Analysis. The analysis starts straight away.
- Check the strengths against what you know. Each one cites a specific figure, so you can check it quickly.
- Read the areas for improvement with their evidence. These are your starting points for improvement planning.
- Export the tables. Use CSV for the summary tables, or Word to edit the text.
- Next year, run Data Comparison with both reports to see what has changed.
Getting better results
- Use a standard achievement report where you can. The assistant recognises its columns, including national benchmarks.
- Upload every file for a comparison before you answer its question. It compares everything at once.
- Include learner IDs with targets. The results table uses them to identify each row.
- Check the sourced references under each answer to see which files were used.
- Adapt the suggested text. Rewritten targets and scheme of work text are drafts.
- When an analysis misses something you know is in the data, tell us which file and column it was in. That tells us whether the problem is in the file or in the assistant, and it is the most useful feedback you can give.
What it will not do
- Use anything but the files you upload or paste. It cannot reach your college systems or the web, and it adds no outside knowledge.
- Name individuals in any response.
- Pad its results. If the data supports three strengths, you get three.
- Make decisions for you. It analyses and suggests. You decide on targets, schemes of work and quality judgements.
- Review against your college’s own scheme of work template. It uses a general standard.
Your data
- KnowledgeFlow runs in your organisation’s own Azure account. Data does not leave it.
- Prompts and responses are stored in line with your organisation’s data retention policy.
- Uploaded files are accessible only to you, in your browser.
- No user data is used to train the underlying AI models.
Common questions
I selected Data Comparison and it asked me a question. Is it stuck? No. It needs you to confirm all your files are uploaded. Reply, and it starts the comparison.
Why did I only get three strengths? Your data clearly supports three. It lists only strengths the evidence backs, and does not pad to five.
The Three Year Trend section is one line. Is something wrong? Your data holds fewer than three years. Upload a report with two earlier years as well to get the trend.
Why are there no names in my SMART target table? The assistant never names individuals. It uses the learner ID or other reference from your data.
The scheme of work review says a section is missing that we do not use. It checks against a general standard, not your template. Decide whether the section matters for your course.
Next steps
Upload this year’s achievement report, run Data Analysis, and check its strengths against what you already know about your provision.