Every few months, Ethan Mollick puts out a blog post on using AI – covering what’s changed, what’s useful, and what’s confusing. If you haven’t come across him, he’s professor at Wharton, has a great blog called One Useful Thing, and wrote a book about living and working with AI, called Co-Intelligence.

The latest of these posts came out yesterday (19th Oct 25). In the past 24 hours, I’ve also received the State of AI report, and the AI ROI Benchmark report from Make AI Work.

The key thing I’m seeing across these is that AI adoption by people (individuals) is moving one heck of a lot faster than companies can adapt to. People can drop legacy systems and habits more quickly and more easily than organisations are able to.

One of the many things I like about Ethan Mollick’s posts is their practicality. He’s someone who lives and breathes AI every day, but still manages to look at the value of the various tools to people who don’t frolic in the AI info firehose.

This is the gold nugget of his latest ‘opinionated guide to using AI‘:

The use cases and proportions in this chart is a usage breakdown from ChatGPT. Ethan has overlaid his own commentary on this, advising what type of model to use in each case. Then he says:

If the chart suggests that a free model is good enough for what you use AI for, pick your favourite and use it without worrying about anything else in the guide.

Ethan Mollick – An Opinionated Guide to Using AI Right Now

As regards the other two research papers, which are more interested in businesses than individuals, yes I looked at them, then called in NotebookLM. The key takeaways are:

1. There is a critical need for skills and expertise (a major barrier to scaling) – exactly which skills or what expertise isn’t particularly well covered, but privacy and data governance are a couple of example areas
2. AI has moved beyond pure experimentation and is becoming a commercial driver requiring measurable returns
3. Organisations are running into all sorts of speed bumps as they try to move AI from small, single-case prototypes to large-scale, integrated systems

Here’s the NotebookLM video overview on these reports (I recommend watching at 1.5x speed) – it’s 8 and a half minutes long.

The video covers a few different areas and I liked the wider perspective it gave me. What do you think?