Let’s get one thing straight from the beginning, this article isn’t about using AI to spy on people, understand their whereabouts or habits or how they interact with each other and society.
It is about watching how people change in their early experience of AI as we currently know it - generative AI, talking to Claude, ChatGPT, Gemini etc and getting it to do a few things for them.
I wouldn’t put myself in the bleeding edge of adopters as I always keep a security outlook on the go, but I do keep myself aware of emerging trends and jump in when I feel the time it right.
But people who you wouldn’t label geeks or nerds are now also starting to realise the capabilities, and not just asking it to summarise emails or compose end of season thank you team messages.
Expected capability level.
Just as some people are more fluent in Microsoft® Excel than others, there will be a range of skills on display as people get used to AI in their lives. Equally there are levels of confidence that go with it. I have crudely made a chart to explain my thinking.
The fascinating thing to watch isn't someone learning how to use AI. It's the moment their mental model of what AI can do changes.
That gives you some particularly interesting people-watching observations:
- High confidence + low capability: “AI can do anything.” — potentially the most interesting group to observe because confidence can outrun competence.
- Low confidence + low capability: “I wouldn't know what to ask it.” — may use AI much less than they could.
- High confidence + high capability: “Give me the problem; I'll work out how AI can help.” — genuine fluency.
- Low confidence + high capability: “It probably can do this, but let's check.” — technically capable, but appropriately cautious.
Confidence can be quite easy to spot… see how they type, hunt-n-peck or fingers racing over the keys - whilst also hitting backspace a bit.
Are their mouse clicks deliberate of gentle?
Another giveaway is if they utilise any IT skills at home, setting up whole home wifi or self hosting a media server on Network Attached Storage (NAS). These people will be used to google copy/paste solutions whilst getting an understanding of what is really happening.
People who appear to have low confidence also appear to develop it slower too. The cautious aspiration of new technologies providing them with a safety blanket.
Skills, capability, experience, knowledge etc… exposes itself in what people do with the tools.
Image generation: easy to start on phones. Initially for sharing funny headshots but turns into developing ideas for posters and presentations. Prompt use is simple and they might appear frustrated at asking for small tweaks.
Document and and email summaries come next, followed by the a little probing research question or two. I asked it for a suitable meeting point between locations give attendees movements and modes of transport. A test where I knew the answer myself but what it to prove it’s worth.
Agent vlookup().
The magic moment happens when they start looking into agents or bots. Streamlining a workflow they do often, and making it not just automated, but adding additional steps too. It is the equivalent of someone learning the vlookup() function in Excel.
Suddenly they start getting notifications when a web page they are monitoring has been updated.
Summaries of posts and reasoning between different documents helps them understand multiple sides of an argument.
Things they have struggled with start to be delegated producing long responses, which we all know should be read to fully understand the output.
I think the next level, is the point when the person starts spotting a mistake or two. Something doesn’t quite fit in the response.
So more detail is added to the chat window, giving the agent a better chance at reasoning back.
Circular conversations start to swirl like small tornadoes, before the human in the loop realises there needs to be a hard stop, deep breaths and a bit of reflection on what has happened so far.
Watching people move through these stages shouldn’t just be for entertainment. We should learn from the experience they are having.
Without a great user experience, there is very little that keeps people interested. It if is too cumbersome, or complicated with special wording or codes, then it won’t appeal to the masses.
Eventually there will be standard way, and the more in-depth way, and just like the peoples confidence and capability level, they will fit in somewhere.
So, as I move through the quadrants myself, I continue to watch other people's experiences. The frustration, the defiance, the small wins — but above all, the growing self-belief that comes from achieving something they didn't think they could.
No comments yet