I thought I was already ahead with AI… until I watched my husband use it like a CTO

For a while now, I have been playing around with AI.

Not in a dramatic way. Not in a “drop everything and become an AI expert overnight” way.

More in the way many of us in business have. I’ve used ChatGPT since the very beginning. It knows a lot about me, my business, my thinking, my clients needs and pain points, my goals and strategies. It has been part of my workflow for a long time.

But what changed things for me was not ChatGPT itself.

It was watching my husband.

My husband is a CTO in a software company in France, and for the past six months he has been deeply immersed in AI. It actually started earlier than that, when I shared with him what I had learned in the INSEAD AI course I took over a year ago, that got him reflecting on how AI could become part of his role as a CTO. Then about six months ago, he really jumped on it with Claude Code, and what he has been producing in his company since then has been quite incredible.

He has code running overnight. He wakes up in the middle of the night to check progress. He gets frustrated when he runs out of credits. He starts again. And by the morning, he has often produced what would normally be two or three weeks of output without AI in the past.

For months, I listened to him talk about this all the time. I watched him obsess over it, and bring this topic up during every single dinner conversation (kids present or not). And honestly, I thought he was being a bit obsessed with AI, and specifically with Claude.

Not because I doubted the value. More because I did not relate to his passion for code. Of course, I am a woman in tech and I have been running my SaaS business for 15 years. But I am not a developer. I have never built the tech myself or ever aspired to it.

Still, over time, something started to shift in me.

I realized that while I had been using AI mostly through conversations, he was using it to create leverage.

And at some point, I hit a very clear internal turning point. I thought: I need to stop just using AI as a smart chat partner and start learning how to integrate it into my business in a deeper way.

The question I kept asking myself was, “What would it look like if I had an AI team working for me 24/7?”

So I decided to see how I could build an actual team of AI assistants inside my business. Not just for content or brainstorming, but to run tasks, save me time, support execution, and help me build what I would call a one-human business supported by a team of AI agents.

That mattered to me for two reasons.

1. Because I want that for myself. I want more leverage, more speed, more structure, and more time freedom when running my business.

2. Because I wanted to offer the same model to my clients, after I’ve tested it on myself. I work with women in tech who are building businesses, products, and offers. I believe very strongly that the AI-empowered business is where we are going. And I do not believe in teaching this from theory. I believe I need to build it in my own business first. Only then I can teach others.

That is what happened this past weekend.

The moment I started building my AI team and realized I was doing it wrong

I decided to explore this hot “AI Agents” topic properly.

Not through another course. Not through Instagram content (which I get plenty in my feed and have trouble trusting). But by actually building something inside my business.

So I downloaded Claude and started playing with it.

I also knew that Claude had the Cowork feature with all the integrations, and because I am not a software developer, Claude Code felt too complicated for me as a starting point. So I started with the integrations. I looked at all the apps I was already using that could be connected. And once I saw what was possible, my mind immediately started going in a thousand directions.

Very quickly, I realized this was going to be a rabbit hole.

The more you explore, the more you can go in different directions. The opportunities feel limitless. And that is exactly why you can end up wasting a lot of time.

So I went back to my good old friend ChatGPT, and we started breaking things down together.

Instead of asking, “What can AI do?” I asked, “What are the super important roles in my business where AI agents could actually help me go faster, optimize my time, create more value for my clients, and help me reach my goals?”

There were many.

Because I am ambitious, I could easily think of ten different roles across the business.

But I decided not to do what many of us do, which is try to build everything at once.

I chose to focus on one role first. Exactly what I teach my clients: the rule of one.

So I started building my AI-powered Executive Assistant whom I called Claudia.

So I worked with ChatGPT to create an executive assistant role. Then I copy-pasted that into Claude. And then I explained to Claude what the system needed to do as if I were briefing a human assistant.

Very logical. Very clear. Very human.

And then what happened next was also very clear.

It started running the tasks, and right away, it ran out of credits.

Then it told me to wait a couple of hours until it could work again.

Huge frustration.

Exactly the same frustration I had watched my husband experience again and again.

And that was the moment I understood something important.

Just playing with AI does not help.

You need the right architecture.

And you need to keep credits in mind while you design it.

The real problem was not AI. It was my architecture.

There is another important part of this story.

I did not start this exploration by pretending I knew everything.

Because I do not.

I am a business owner, coach, and SaaS founder. I am not a developer. So my first thought was: Who can I ask for advice? Who can guide me on where to even start?

The first obvious person was my husband.

He told me he loves Claude Code. He also mentioned Claude Cowork, even though he had not used it before. That gave me one angle.

Then I wanted a second opinion, so I went to the CTO of my own tech business Le VPN, my co-founder, who has been integrating AI into our business since the early days of generative AI. I know he has already built many things into our business. He is a big fan of OpenAI, and he also suggested OpenClaw plus Claude Code. We discussed his vision, different ways to integrate things, and how later we could always move toward a more advanced setup with a stronger subscription and more sophisticated processes.

And this is where I hit one of the most useful lessons of the whole weekend.

The paid subscription matters. But it is not magic.

Here in Europe, the basic paid Claude subscription starts around 18 euros a month. And what I can tell you is this: you really cannot do that much with it if your architecture is messy. After just one weekend of experimenting, I was already at 86 percent of my weekly credit limit. Sometimes I would do something for 30 minutes and then run out of credits and have to wait for two hours.

This suddenly gave me a new lens on a conversation I had with one of my clients the week before.

She had been trying to build her MVP on her own using Lovable. And she kept telling me she was constantly running out of credits. She was surprised when she heard that inside the next intake of the Weatherproof Business Launchpad, we are doing a no-code MVP sprint using Lovable and building MVPs with the front end and the back end on a free version with very limited credits.

At the time, I did not really know how to answer her concern.

Now I do.

Her problem was not really the credits.

Her problem was the lack of lean architecture.

This is something developers understand because they have had to optimize around constraints for years. They know they need a proper structure first. They know not to overcomplicate things. As business people, we often do not think this way naturally.

But now I see it much more clearly.

You do not have to create the architecture yourself from scratch. But you do have to know the problem exists. Then you can ask the right questions and let AI help you create the architecture.

That was one of my biggest lessons from this weekend of experimenting.

Building my first AI agent… during my kids’ naps this weekend

Once I understood that the main issue was architecture, I looked at the next problem.

My data was everywhere.

• On my computer.

• In Google Drive.

• In Notion.

• In email.

• And in other places too (and I don’t mention post-its!).

That means every time the AI needs to browse all of that, read everything, make sense of it, and bring it together, it uses a huge amount of AI brain power which translates into exhausting my credits. So even with a paid plan, you can get blocked very quickly.

At first, I thought maybe I should just bring everything into Notion. Copy the data there and centralize it that way.

But after going back and forth with ChatGPT and interogating it on the best way to build this based on everything it already knows about my business after 3-years of daily chatting with me, I realized that it would not really be scalable for what I’m trying to build. And I hate building things twice.

My business is already becoming quite complex. There will be more data, more interactions, more moving parts, more integrations, and potentially more agents later.

So the best solution ChatGPT suggested for me was Airtable.

For simpler businesses, and for many of my clients, I still think Notion is an excellent place to start. It is easy. It is accessible. You can do a lot with the free version.

But for me, I chose Airtable, even though it is a paid solution.

What happened next was actually one of my favorite parts.

Because ChatGPT already knows so much about me and my business, it became very easy to use it to generate the right prompts for Airtable AI. It helped me translate my existing Excel trackers, my current ways of monitoring data, and the systems I was already using into a much more structured database.

It basically dictated to me the exact prompts to give to Airtable to build it, and then adjusted them based on the responses from Airtable that I copied back into our chat. It took about 30 minutes and zero Claude credits consumed.

So over Saturday and Sunday, not in some glamorous uninterrupted sprint, but in real life, with kids, with breaks, with two walks my husband took with them so I could get an hour and a half at home to work on the commands, I built the first version of my CEO dashboard.

It included all the different goals I have in my business.

• My tasks.

• My weekly goals.

• Monthly goals.

• Quarterly goals.

• Yearly goals.

• How everything aligns.

• My routine blocks.

• My podcast guest-speaking pipeline.

• My live events pipeline.

• My calendar.

• My priorities.

• The tracking of everything.

Then I connected that Airtable base to Claude Cowork and integrated it into the executive assistant agent I had created there.

By Monday morning, I had my AI executive assistant Claudia working for me.

Meet Claudia, my AI executive assistant, Who Changed More Than My Workflow. She Changed How I Think.

By Monday morning, I received my first morning brief by email.

It landed in my inbox at 7 a.m., as scheduled.

And it contained everything I needed to start the day.

• My upcoming schedule.

• The different things that had come up overnight.

• My priorities for the week.

• How those priorities aligned with my longer-term priorities.

• Things to keep in mind.

• Suggestions for how to manage my energy during the day.

• What type of routine blocks I could add into my calendar if I had free time.

That alone already felt useful.

But I did not stop there.

I also asked Claudia to help me at the end of the day.

The idea was that throughout the day, I would give her short updates. Voice notes. Quick thoughts. New ideas. Strategic reflections. Things I wanted to remember. Things I wanted to do in the short term and medium term.

We discussed options and agreed on how we would work together.

I would pop in with brief notes during the day and tell her what was going on.

She would update her memory.

Then she would update all the relevant tables, sort tasks where they belonged, move things into the right pipeline, and keep everything current.

Then when I close the day, I send a message saying, “closing the day.”

She gives me a summary of what was accomplished, the progress made, updates everything again, and, most importantly, cheers me on my wins.

Then on Friday night, when I close the week, she sends me a full recap of everything achieved that week, how that connects to my goals, and how much progress I made. I receive that by email, and it is already reflected in the tables and dashboards.

There was one more thing I asked her to do.

At 6 a.m. every morning, she sends me a short world news digest.

Just a few key points from politics, economy, technology, business, and especially AI.

That matters because four years ago I made a conscious decision to stop reading and watching the news. I used to fall into a rabbit hole. My husband became the one who gave me a short digest of what was happening in the world, and then if I needed to look deeper, I would.

Now I have automated that to Claudia.

And this matters much more than convenience.

I do not want to scroll.

I do not want to see upsetting images or videos that shift my emotional state before I begin my day.

My energy is one of my highest priorities. Without energy, I cannot create value or help my clients. And I have a lot I want to create and give to the world.

So now that brief comes at 6 a.m. every morning.

And on that Monday morning, after receiving both briefs, I did what I often do after yoga. I had creative ideas. So I put my computer next to me and sent voice notes to Claudia as the ideas came into my head. She took notes, classified everything, and organized it into projects and priorities.

That felt amazing.

And it is still an experiment. At the moment I am writing this, it has been about 48 hours since I started. I am literally waiting for Claude to reset again because I have used my credits and have to wait until 11 a.m. before I can continue. I may end up needing a stronger subscription eventually. But I actually like experimenting first on a basic plan before spending more, because I do not yet know exactly where this is going.

What I do know is this.

The executive assistant role is only the beginning.

She saves time, yes. But she is not doing the actual work of the business. She is helping me do the work better.

The next step for me is to build a bigger team of AI agents.

And more than anything, this experiment is changing the way I think about building.

Now, whenever I do something, I ask myself: What kind of digital asset can I create here?

What database, what access point, what structured source of information can I build so that one day an AI agent can take over this role or support it inside the company?

This is exactly the kind of thinking I now want to help my clients build as well.

Inside the next intake of the Weatherproof Business Launchpad, I want us not only to validate business ideas and make sure the business works. I also want us to start thinking about how to design an AI-powered business.

Not overnight. Not in a hype driven way. But by gradually building digital assets and systems that make later delegation and automation possible.

I am doing this experiment because I want to go faster in my own business.

But I am also doing it because I want to help my clients build this kind of business in the future.

And I believe I can only teach what I have made work myself.

So this is the journey I am on now.

And I am only at the beginning.

There is a lot of noise around AI right now, especially on Instagram. Many people talking about it are, frankly, not qualified. That is exactly why I want to be careful.

I want to learn from qualified sources. I want to seek advice from actual software developers who are integrating these tools into real businesses at a much more sophisticated level than I am. I want to build in public. I want to experiment honestly. I want to share what is real.

And I also believe I can catch up quickly.

Not by pretending to be a software engineer. But by being very clear on the business goals I have, building the right systems around them, getting AI support in the actual build part, and then helping my clients do the same.

If this topic resonates with you, I would genuinely love to know.

Because I am seriously considering making this a real component of the Weatherproof Business Launchpad in the future.

Not just how to validate and build a business idea. But how to build a business that is structurally ready for AI, automation, leverage, and scale.

That is the experiment I am running now.

Because I don’t believe the future belongs to those who do more.

It belongs to those who build with clarity, structure, and intention. And do it in a way that protects their time, their energy, and the life they are building alongside it.