Nobody hands you a manual when you start building an AI-powered way of working. You try one tool, then another, then you're watching a YouTube video at 11pm about a workflow you don't need yet, and suddenly you have nine browser tabs open and nothing actually done.
That was me, not that long ago.
I've been creating a series of images showing the kinds of tools a solopreneur or small business owner might use, everything from ideas and research through to products and sales. It looks tidy when it's laid out in categories. Logical. Like someone sat down and planned the whole thing.
I didn't.
My stack evolved the way most real things do. Messily, responsively, and one problem at a time. Over the past year or so, it's shifted in ways I didn't anticipate. What I thought would be my primary tool has started to share the job. New tools have crept in through curiosity. Others are on the list but haven't earned their place yet. And I'm prepared that I may need to shift again in the near future.
Here's the honest version.
Where it started: ChatGPT and learning to think out loud
ChatGPT was my entry point. Like most people, it was the first AI tool I used seriously, and I spent a good few months building up context there: saving prompts, understanding what it could do, learning how to ask better questions.
I still use it regularly. And one of the reasons it stays is practical: lots of people I work with or talk to are familiar with ChatGPT, and some actions are actually quicker and easier in Chat, like my bulk Pinterest Pin generator. The ability to share a GPT or point someone toward a workflow they can replicate themselves matters when you're trying to be useful to other small business owners.
But over the past few months, I've been moving more and more of my actual work to Claude. I find it stronger for detailed writing, for thinking through strategy, and for the kind of longer-form work that underpins what I'm doing with The Curation Corner. It's become my default for anything that requires real depth.
The two tools do different jobs now. That's fine. That's probably how it should be.
Research and thinking: Perplexity and Gemini
Perplexity is where I go when I need to actually verify something. Not just get a plausible-sounding answer, but find a source I can trust. It's my research layer, and it's earned a permanent place in the stack. For example: find reports on the EU AI Act, or analyse the most successful images from luxury cosmetics campaigns across the UK and ROI in 2025 and summarise the key themes.
Gemini acts as my second opinion. When I'm not sure about a piece of content or a direction, I'll run it by Gemini to see if I get a different perspective. It's not my primary tool, but that second-opinion function is more useful than I expected.
Capturing ideas: Otter.ai and the voice memo habit
This one changed how I work more than almost anything else. I record voice memos and meetings, Otter transcribes them, and those transcripts become the raw material for content, notes, plans.
For a busy mother running multiple things at once, the ability to speak out loud while walking the dog or sitting in the car, and have my thoughts waiting for me in text form later, is not a small thing. It's how I capture ideas that used to just disappear. This is how most of my blogs start. I'll ramble on to Otter on a particular topic, then take the transcript and drop it into ChatGPT, and more often now Claude (because they know me so well), and get them to help organise my thoughts into a structure I then revise until I'm happy to publish.
Organisation: Notion as a prompt library (for now)
I haven't built Notion out fully yet. What I actually use it for right now is a prompt database: a place where I save my most useful prompts, tagged and organised so I can find them when I need them. Not a full knowledge management system. Not an elaborate OS. A library of things that work, filed so I can retrieve them.
It's on my list to develop further. But I'd rather tell you what I actually use it for than what I intend to use it for.
Content processing: NotebookLM
I use NotebookLM to process information I don't have time to read in full. Long reports, research documents, dense reads. NotebookLM turns them into podcast-style audio and structured summaries. It's quietly become one of the most useful tools in the stack for keeping up with things without losing hours to them.
Brand and visuals: Nano Banana and Canva
This is the category I'm most experimental about. Image generation is still something I'm figuring out.
Nano Banana is a Google-driven image generation tool. You can access it through Gemini or via a paid subscription to OpenArt.ai or Higgsfield. I use OpenArt.ai as it has access to multiple video and editing tools. It's produced surprisingly strong results for the cost, and it's become my go-to for generating brand-aligned imagery.
And Canva holds everything together: resizing, formatting, adapting visuals across platforms. It's not glamorous, but it's where hours go from wasted to saved. I find some of Canva's AI tools handy for speed, like the background remover or finding a basic stock image. It's a fancy workhorse for templates, whether it's social formats, search formats, or digital products.
You'll notice Midjourney isn't in my active stack. It's on my radar (I know people who use it brilliantly), but I haven't built it into a regular workflow yet. It'll come.
Automation: the honest gap
Zapier and Make are both tools I know I should be using more. I have dabbled a little with Make. The time savings available through proper automation are real. But I haven't prioritised building those workflows yet, and I'd rather be honest about that than pretend I have it all wired up. Also, from what I'm learning, most of what I need from a small business perspective can be done through Chat and Claude. More of that to come later.
Products and sales: Canva and Beacons
Beacons is where my digital products live currently. It got me up and running and experimenting fast. It's simple, visual, and doesn't require a full e-commerce setup to get something in front of people. For where I am right now, it works.
Canva appears again here. It's the connective tissue across more of this stack than I realised until I wrote it all down.
What this year has actually taught me
The right stack of tools, your right stack, will save you hours every week. Not because you have every tool, but because the tools you do have are ones you actually understand and use consistently.
I'm not running all of this at full capacity. My automation layer is thin. Some weeks I lean on two or three tools and ignore the rest. The stack is still evolving, the AI world is constantly evolving, and a year from now it will definitely look different again.
Start with one problem. Find the tool that solves it. Use it until it becomes second nature. Then add the next one.
That's the whole strategy.
Now I want to hear from you
What does your current stack look like, even if it's just one or two tools? Have you made any unexpected switches recently, or found something that changed how you work?