← All modules
Module 07 · Flagship

Safeguarding with AI

Children's and adults' tracks. Built on the AOP foundation from Module 6.

6/7
Section progress
~9 minutes · interactive

AI as a reflective tool

9 minutes including the iteration trail. Using AI as a sounding board for your own thinking, not an answer-giver.

The previous five sections have been about AI doing work on the file. This one is about AI doing work on you. The use case is your reflection, rather than your records.

The most valuable mode in safeguarding is iteration as a sounding board. You bring AI a case it has all the information about, and you use it to interrogate your own thinking. The aim is for AI to help you find what you already half-know but have not yet named, rather than to hand you an answer you can act on without thinking.

A note on the tools. Copilot, ChatGPT, Claude, Gemini are all imperfect, and none of them are professional supervisors. They are thinking partners that get sharper the more you iterate with them, and that is the use case this section is about.

An editorial illustration of a calm practitioner figure at a desk leaning toward an open notebook, with three thought-bubbles in cyan stacked above the notebook getting larger and more detailed from bottom to top. To the side, a small notebook-and-pen icon labelled AI sounding board.
Iteration in motion. The thought sharpens, not because AI got smarter, but because you told it more about what you actually wanted.

What "reflective use" actually means

It is the difference between two questions:

"Tell me what to do."

AI gives you an answer. The answer is generic, often wrong, and the moment you accept it you have stopped thinking.

"Identify gaps, inconsistencies, and themes."

AI surfaces what is in the records: the gaps, the inconsistencies, the themes and potential concerns that emerge across the file. You decide what to do with what comes back.

The second mode is what makes AI genuinely useful in the reflective stages of safeguarding work: pre-supervision, post-supervision, the day after a difficult home visit, the week after a strategy meeting that did not go the way you expected. You are asking AI to stress-test your thinking, rather than to think for you.

Try it: the iteration trail

Walk through five iterations of the same case. Each iteration adds context, role, or a sharper reflective ask. Watch what changes in the response.

Iteration 1 of 5
Your prompt
AI returns

Two things to keep: the iteration-4 prompt as a re-usable template across cases, and the iteration-5 closing prompt as the discipline that turns the conversation into something useful.

The sounding-board mindset

When AI is acting as a sounding board, three things keep the work safe:

  • Treat the output as questions, not answers. Whatever AI gives you, the next question is "is this true of the case I know?" If yes, act on it. If no, you have learned something about how AI is reading your file, which is also useful.
  • Bring the output to supervision, not to the document. A sounding-board conversation with AI is a draft for your own thinking, not a draft for the record. The thinking belongs in supervision and in the conversations with your colleagues. The record gets what you decide after that.
  • Use it to find your own blind spot. The most valuable prompt in the iteration trail is the one that asks AI to flag any risk-averse phrasing in your own notes that might be smoothing what you actually felt in the room. The blind spot is usually in the gap between what you wrote and what you knew.
AI is a mediocre thinking partner that sharpens the more you iterate with it.
The output is not what you put in the record; it is the thinking you take to supervision.
Reflection

Pick a case from the past month where something did not sit right and you have not yet named what. Try the iteration trail prompt template on it. What does AI surface that you had not articulated? What does it surface that you disagree with? Which of those is more useful?

← Previous Next →