Midlife MavericksLast week, I wrote about the skills that matter in a world where AI is rewriting the rules. This week, I want to go deeper. I want to show you what I actually do with these tools. The path and the day-to-day.
I started using ChatGPT in June 2024. Free version. Zero technical experience. I spent 25 years in finance, 15 of them in investment banking, and the most technical thing I did was an Excel spreadsheet.
What pushed me to try it was something very specific. I was preparing a funding application for a local playground. I needed demographic data I couldn’t find anywhere. Days on Google, nothing. I opened ChatGPT almost out of desperation, asked the question, and within minutes I had what I needed. I wrote a three-minute speech based on that data. I secured £37,000 in funding.
I thought: OK, there’s something here.
June 2024 to today
This didn’t happen overnight. There was a progression, with many hours of learning in between.
June 2024. Free ChatGPT. First experiments. I started writing content on Threads around the same time.
January 2025. Switched to the paid version. Started training ChatGPT with my voice. Built my first custom GPTs.
Summer 2025. The workflows I built were saving me 20 to 30 hours a week. I kept building from there.
Today. I’m migrating to Claude, rebuilding my GPTs into skills and projects. I have agents running in Notion and Make. I use automations across multiple platforms. And I’m still learning.
The tools and what they actually do
I’m not going to give you a list of tools with star ratings. I’ll tell you what I use, why, and what changed.
Perplexity — Research without hallucinations
This is my starting point for any research. It doesn’t invent information, it’s fast, and it gives me sources. When I need to investigate a new topic, a market, a trend, I start here. Always.
Fathom — The gold is in the full transcript
I record all my meetings in Fathom. I have a database with every single one. Most people don’t do this. I use the entire transcript, not the summaries or action points.
Summaries are fine. But what matters is what people say between the lines. The language they use. The concerns they don’t verbalise directly. The patterns that repeat from meeting to meeting. All of that is in the full transcript, and most people ignore it.
It changed how I work with my consulting clients. I can analyse how a fund manager presents in a meeting with investors, identify where they lose the thread, where they don’t answer a question directly, and help them improve. If you can measure it, you can manage it.
ChatGPT — Bots, projects and actions
I use ChatGPT in three ways:
Custom GPTs for repetitive work. I have bots for client proposals, for my newsletter, for content repurposing, for presentation analysis. Each one saves me 2 to 4 hours a week.
Projects for ongoing work. When I’m working on a project that develops over weeks, I use the projects feature. The different chats within the project communicate with each other and maintain context.
Actions for automated daily tasks. Every day at 7am I receive a summary of relevant news for my industry, directly in my email.
Voice training — 60 pages of “me”
I have a brand document with over 60 pages that trains AI to write like me. It’s not just the tone. It’s the different tones for different audiences, different types of content, different contexts. The way I write a newsletter is different from how I write a client proposal, which is different from how I write a Threads post.
This applies to everything: articles, newsletter, client communication, proposals, work review, content production. The result is that I write one newsletter a week and repurpose it into 60 pieces of content. In two hours.
Voice as a capture system
Using voice changed everything. Not dictation. Capture.
When I leave a client meeting where I couldn’t record, I immediately do a voice brain dump. Three minutes. I organise those thoughts into notes and follow-ups afterwards.
When I have an idea on the street, in the car, walking, I record it. Could be a work idea, a content idea, doesn’t matter. The important thing is it leaves my head and goes into the right database. It doesn’t take up mental space.
My most recent addition is Wispr Flow, which lets me do voice-to-text anywhere. Takes the weight off starting to write from scratch. Money well spent.
Notion — The centre of everything
All my documents, all my content, all my meeting recordings, my CRM, everything is in Notion. AI is only as good as the context you give it.
Most people have the tools, but the information is scattered. A bit in email, another bit in Google Drive, another in their head. If AI can’t see your world, it can’t help you.
Putting everything in one place and having my agents running inside Notion is what changed my day-to-day the most. My Notion agent is called Eugenia.
I use a series of other agents that automatically add tasks to my task list. Ideas pulled from client calls get added to my ideas database. And my favourite one: I have an agent that counts how many “swear words” I say per meeting, because I’m trying to do a language detox. (To each their own, at least I’m honest.)
AI as tech support and systems builder
I use AI to troubleshoot technical problems, build systems and set up automations. When I need to learn something new, whether it’s configuring an automation in Zapier or Make or solving a problem in Power Automate, I ask for a step-by-step tutorial. It’s like having an IT technician who never gets annoyed and is never on holiday.
Every automation and system I describe in this post was built with AI’s help. I didn’t hire a tech team. I learned and built, tool by tool.
Perplexity as a learning coach
Whenever I want to learn something new, I start in Perplexity. I ask it to make me a plan: what I need to learn, what the best sources are, and what the sequence should be. Then I keep that conversation open.
Every time I complete a section, I tell it: “Finished part one, log it.” If I only have 30 minutes one day, I ask: “I only have 30 minutes today. What should I do from this list?” And it adapts.
Instead of consuming information randomly, I have a plan that adapts to the time I have.
Notebook LM — Source of truth and learning tool
I use Notebook LM for two things.
For research, it’s my source of truth. When I gather several documents on a topic, I dump them all into Notebook LM. It organises and makes sense of all the information, using only those sources. No web search, no external noise. Just my documents.
For learning, it’s where I consolidate everything. If I have YouTube video tutorials, I drop the links straight into Notebook LM and it teaches me the content. Could be a specific tutorial or a video with an interesting idea that I don’t have time to watch, because time is short right now. After going through everything, I make study notes. It ensures I actually absorbed what I needed, and didn’t just skim over it.
Power Automate + Microsoft 365
I operate on Microsoft 365 and I’ve automated a significant part of my inbox management with Power Automate. It’s not perfect. It’s not fully done. But it’s already removed an enormous amount of repetitive tasks.
Zapier and Make — The bridges
I use Zapier and Make to run automations between platforms. From Kit to Outlook, from Outlook to Kit, from Substack to Kit, cleaning up personal emails outside the Microsoft space. Each one of these automations seems small. Together, they free up hours.
The time audit before automating
Before building any automation, I did something I recommend to everyone: a time audit.
For three days, I used ChatGPT as a stopwatch. I told it in real time what I was doing. “8:32, email. 8:36, finished the email. 8:37, opened the CRM.” Every task, every minute. At lunchtime, I’d ask for a summary of the morning. ChatGPT kept track of everything and we’d continue in the afternoon. Next day, same thing.
After three days, I had a very clear, and very uncomfortable, picture of where my time was actually going. What repeated. What was unnecessary.
It’s tempting to want to automate everything. But some things shouldn’t be automated. They should simply be eliminated. If a task shouldn’t exist, automating it gives it a life it doesn’t deserve.
I thought carefully about each automation. I hired someone to help me with the first one, and then built the rest myself.
What all of this changes
I gained time. But what I really gained was presence and a clear head to think.
I started at 48, with zero technical experience. I learned by doing, not reading. I built my own systems, my own automations, my own agents. No team of engineers, just curiosity and the tools I described in this post.
The tools change every few months. What matters is to never stop learning or be afraid to experiment. That responsibility is yours alone.
Comment below. What’s the one thing stopping you from experimenting with AI? I read every response.
This article was first published in Portuguese in my weekly column Oh pá, não me lixem! for Executiva.
