Improving policy implementation with ChatGPT Edu

I am implementing a departmental Information Asset Management policy. Broadly this involves identifying Information Asset Owners (IAO), training them, then getting them to identify and record the information of enduring value for which they are responsible (called "information assets").

The information asset identification phase is particularly challenging as the concept of an information asset is fairly unintuitive, and very, very broad. Not only is it not always obvious what "of enduring value means", but there is a huge variety of information of enduring value held within the university, it exists in a huge variety of forms, and the same collection of information can be defined as multiple different, valid information asset schemas.

The most effective way of identifying what information assets a given Information Asset Owner holds is to sit with them anywhere between 15-60 minutes talking through what information they work with and suggesting different ways of capturing that as information assets. The problem is that this is incredibly labour intensive and there is only one of me doing this work.

I created a GPT under my ChatGPT Edu account with the departmental Information Asset Management Policy, the Information Asset Management Training and a suite of useful examples and a set of instructions telling the GPT to advise staff on how to identify their information assets.

Information Asset Owners can access the GPT, add a bunch of context about the work they do and get recommendations about what information assets they might hold, and providing the basic initial write-up for them. The Information Asset Owners can chat through suggested assets, reject or refine them and leave with a list which hopefully is at least decent quality and makes sense to them.

The Information Asset Owners can then submit their identified assets for me to review and if I think there are improvements, I will often use the self-same GPT to help me phrase my recommendations.

There are three main benefits to this GenAI-supported process:

1. The submitted first draft information asset lists are better quality than if they just worked on them on their own.
2. The Information Asset Owners understand their information assets better than if they just worked on them on their own.
3. I save a fair amount of time as I am not providing as much direct 1:1 support and feedback.

My advice to anyone starting out with GenAI tools is: identify a work problem you are facing, even if you are not wholly clear on what it amounts to. Dump a good amount of context about the problem - what you are trying to achieve, what is happening around it, what seems to be stopping you (even if none of these are very clear) - then ask the chat "Assess how confident you are that you understand my problem in % terms. Ask me questions, one at a time, about this problem until your confidence in your understanding is 95%, then create a short summary reflecting that understand".

Starting a longer, more difficult piece of work with a clearly defined problem statement like this is incredibly useful.