ChatGPT launched as a consumer tool in November 2022. Nearly four years later, here’s how generative AI tools have changed my working life (or not).
It’s been over a year since I did a post about the ways in which I use generative AI.
So, I thought, time to see what’s stayed the same, and what’s changed.
What has not changed at all is a ground rule: I don’t work as a “human-in-the-loop”. I deploy fancy new tech tools if they save me time or give me insights I otherwise wouldn’t have had. Round here, at the Safe Hands office, the phrase is “AI-in-the-loop”.
Let’s get straight to it.
How things stood in July 2025
I wrote a detailed blog post about what I was doing with generative AI (the details were offered in the hope of making those tools more accessible to other people).
Read that here: The diary of an AI user: the bad and the good – Safe Hands
Now, a year later, this is the executive summary:
- I used AI for frivolous things like recipe generation or plant identification; both of those are still in play.
- I was not using AI tools for generating social media posts based on my blog posts (they just don’t sound like me). Still not doing that (but see below about email subject lines.)
- I was doing some early experimentation with using inbuilt AI tools (the Gemini sidebar in Google Docs for instance). Things were underwhelming.
- I was largely ignoring the AI mode on my phone, skimming any given answer and then looking at the sources, choosing to look at the ones that seemed as if they would work for me. That’s still the case.
- I was an enthusiastic user of NotebookLM. It’s now called Gemini Notebook, and it’s still my most-used AI tool.
- I’m still using AI for things like summarising, consolidating texts, thought experiments, project plans and marketing ideas.
- Then, I was not using AI for writing. Still not doing that.
- I am still using my own custom-built tools (Gemini’s Gems, Custom GPTs and Claude Projects).
What’s new since then?
Tools I use: The big trend over the last year has been in an expansion of the tools I use. Back in the olden days I was pretty much confined to Gemini and ChatGPT. In the interim, Safe Hands has been asked to train a cohort of people to use Claude – and that meant I had to go on a steep learning curve.
Favourite tool: By far the most useful tool I’ve added to my arsenal in the last year is the adoption of one particular AI meeting notetaker: Granola AI. I didn’t know until I looked at it properly how much time and distraction was bundled up in keeping track of what happens in meetings. There are a lot of AI notetakers on the market, but Granola is so good that I have opted to pay for it (having started on the free plan). Pluses include that it doesn’t insert itself as a persona into meetings; that it does really good summaries; and that you can interrogate all your notes inside the Granola app. This one tool alone keeps me sane.
Just plain useful: This is the hidden way in which AI has improved. Over the last few months, I’ve been increasingly leaning on Gemini in Google Drive and on Copilot in Office applications. This is one of the ways in which AI really improves my life. Some of the things I’ve done successfully:
- I had a spreadsheet for a set of training cohorts from a client that was organised by name rather than training day. I asked Gemini to rearrange it so that I could see who was to be in the training day by day. Voila – done, no mistakes made.
- I had a document with hyperlinked URLs (a lot of them) which I needed to extract so I could put them into a Notebook. Get me the URLs, I said to Gemini. Perhaps a minute later, I had a page of them, ready to be copied and pasted into a notebook.
- I had a messy set of data from some work I did with Claude – I asked Copilot to turn it into a table in a Word document. Done, perhaps 30 seconds later.
- I could not come up with an idea for a picture to illustrate this very post. I said this to Copilot: “Suggest an image I can use to illustrate this. Don’t make the image i just want concepts”. (Grammar doesn’t matter when talking to AI!). It came up with idea of using an image to do with tool-using, and I went and did my own searching on Unsplash.
One neat little trick: As part of my Claude learning curve, I delved a bit into the business of making what’s called a Skill. Essentially this is a way of storing a set of instructions that can be invoked when you want to do a particular task. I made one that helps me take these blog posts and generate subject lines for the email newsletter which gets sent out the next day. I don’t always use exactly what Claude gives me, but it does always give me a set of ideas to work with.
The big new learning curve: I’ve started seriously experimenting with using local LLMs (which are instances of generative AI that live only on your computer). They’re the tool of choice for privacy protection and I’ve had some interesting results when comparing the tool I chose with results from the big models. It’s too early to sum all that up, but this is my shiny new toy.
Final words
I ended last year’s post with a list of lessons I’d learned in the course of my journey with generative AI. Those still stand but I’d say that the overarching thing I’ve embedded in my practice is the importance of boundaries. I’ve been around this tech enough now to know that there are many things I just don’t want to hand over to it: my own writing and thinking have a value that the machine will never deliver.
Main picture: Brett Jordan, Unsplash
Previous Sensible Guide articles
The diary of an AI user: the bad and the good – (The 2025 post) Generative AI can be overwhelming. But over time, it’s possible to find ways to make it work for you. Here’s how AI use works in practice for me…
The AI tool everybody should be using – NotebookLM – Research assistant and note-taker rolled into one – allow me to introduce you to NotebookLM.
Productivity – what is it, can AI help? – Being busy is not the same as being productive. That’s a cliche – but sometimes cliches are useful.
Why – and how – to ensure privacy when using GenAI (it’s in the anonymising!) – Generative AI tools are not bastions of privacy and security. Here’s what you need to know about anonymising documents to work carefully and securely.
If any of this is relevant to your work…
You might find one of these useful: