The person helping with AI is doing work too
An AI workshop can leave families or colleagues carrying the follow-up work. Programme budgets and measures should make that work visible.
OpenAI announced free ChatGPT workshops with Older Adults Technology Services from AARP on September 16, designed for 1,000 older adults across ten US cities. Practical use and online safety were part of the programme. For an executive funding an initiative like this, I would put one more item in the budget: the help people will need after they leave.
That help takes someone’s time. A workshop may make it easier to start using AI while leaving a family member, colleague or service team to check the results. If the sponsor only counts attendance and successful exercises, that work can disappear from the programme’s account of what it achieved.
A separate August preprint by Hyehyun Chu and colleagues offers a reason to examine this. Their study of six older and seven younger adults found that family support could solve an immediate problem without leaving a method the learner could reuse. It is a small qualitative study, not an evaluation of the OpenAI workshops or evidence about the wider workforce.
The useful distinction is between help that teaches and help that has to remain available. A facilitator can teach someone how to compare an AI answer with a reliable source. Some decisions will still require a qualified person. Reducing every request for help would be the wrong ambition; knowing when to ask can be evidence of sound judgement.
For an employer, consider a team member who becomes the unofficial person everyone asks to check AI-generated work. The business might report faster drafting while that colleague absorbs the verification. If this happens, another general AI course will not settle how much checking the role requires, who should do it, or what other work they should stop doing. Those are management decisions.
I would ask the learning leader and the operating manager to examine a few actual requests together. Someone who cannot recognize an unreliable answer may need guided practice. Someone who recognizes the problem but cannot access the approved information needs the access problem fixed. A question that requires professional judgement needs a dependable referral route. The response should follow the reason help was needed.
Before expanding the programme, agree who receives those requests and allow time for that work. Then revisit a small group of participants after they have used AI on their own. Record what they handled, where they sought help and what happened next. Fewer questions alone would tell me little; people may have learned, stopped using the tool, or become reluctant to ask.
The sponsor can fund useful learning now. It should also be clear who will carry the work the session cannot remove.
Correction, October 8: The earlier version attributed the family-support findings to OpenAI. They come from the separate Chu and colleagues study cited here.
The signal I’m watching
Whether AI learning programmes identify who handles follow-up questions and distinguish a learning need from an access problem or a decision requiring qualified help.
What would strengthen this signal
Participant follow-up and support records showing which requests persist, who handles them, and whether revised learning or work arrangements resolve the underlying difficulty.