Using AI for Life Admin: Week 1 of My Personal Assistant Experiment

Testing ChatGPT Pro and Codex as a personal admin layer, not just business productivity tools.

After attending a recent Codex hackathon, I was gifted a month of ChatGPT Pro.

I had already been thinking about whether to trial Pro, but the higher monthly cost had made me pause. I knew I could use it for business, research, writing and AI learning, but I wanted to be intentional about what I tested.

My default assumption was that I should use it for something obviously productive: client systems, business workflows, research, content, or automation experiments. But the more I thought about it, the more I realised that the area I actually wanted to test was not business productivity at all.

It was life admin.

Not in a dramatic "AI will run my life" way, but in the ordinary and practical sense of whether AI could help me sort through the mental clutter of daily life. Could it help me remember the small but important things? Could it help me create a better system for personal admin, family admin, reminders, appointments, lists, and all the loose ends that are easy to carry around invisibly?

So I decided to use this month as a personal OS experiment: not AI for doing more work, but AI for carrying less.

Why this felt different from my usual AI use

I have been using ChatGPT for around three and a half years and Codex for around eighteen months. Most of that usage has sat somewhere between business, work and learning.

Even when I have used AI personally, I have usually associated it with work-adjacent things: writing, research, systems, client ideas, automation, strategy, or learning new tools.

Setting up ChatGPT Pro felt different because it was an individual account. It gave me a separate space that was not immediately tied to business delivery.

That separation mattered more than I expected. My business account has a different purpose: it is where work projects, client thinking, business workflows and more formal systems belong. This Pro account felt more like a dedicated space for personal life admin, personal projects and experiments.

I also have a separate Mac mini that I use for AI experiments, which added another useful boundary. It meant this was not just a different tab or a different login, but a different environment entirely. It gave me somewhere to test ideas without everything blending back into business work.

That has been one of the first useful insights from Week 1: sometimes the container matters.

Having a dedicated space changes the way I think. It gives me somewhere I can brain dump a to-do list, capture scattered thoughts, or explore a personal project without automatically asking, "How does this fit into my business?"

AI for work is everywhere. But what about life admin?

Most of the conversation around AI still seems to sit in the world of work. We hear about productivity, automation, workflows, content, coding, research, reporting and how much more people can get done.

That makes sense. In a business setting, the value of AI can often be easier to explain. Did it save time? Did it speed up a workflow? Did it reduce manual effort? Did it help produce something faster?

The personal layer is blurrier.

The admin of daily life is not always easy to measure, because it is rarely one neat task or one workflow. It is the reminders, appointments, meal planning, errands, forms, bookings, renewals, household tasks, family admin, follow-ups, decisions and loose ends that quietly accumulate in the background.

It is the mental list that never quite switches off.

While AI for business is everywhere, AI for messy life admin feels less talked about. Possibly because it is harder to define. Possibly because it feels less "productive." Possibly because it sits in that unpaid, invisible layer of life that many people are simply expected to manage.

But that is exactly why I am interested in it.

What I was actually trying to solve

The mental load of life's to-do lists has been a personal bugbear of mine for years.

I have tried plenty of productivity apps, task managers, calendars, reminders, notebooks and systems. Some have worked for a while. Most eventually became another place where unfinished tasks went to sit.

The problem is not that I need another to-do list.

The problem is sorting and sifting.

It is remembering the long list of small but important things. It is capturing random brain dumps and putting them into the right buckets. It is turning messy thoughts into next actions. It is separating tasks from decisions. It is knowing what actually needs attention next, rather than just having another place to store the overwhelm.

That is the part that creates the mental load.

So for Week 1, I decided to give ChatGPT Pro and Codex the biggest challenge I had: not "help me be more productive," but "help me structure the admin I carry around in my head."

This is where the idea of a personal OS comes in.

For me, a personal OS is not one perfect app or one magical dashboard. It is more like a support layer. A way of capturing, sorting and reviewing the things that otherwise float around mentally.

A place for life admin. A place for personal projects. A place for lists that are not business lists. A place to work out what needs doing, what needs deciding, and what can wait.

What I tested in Week 1

Week 1 was not about building a perfect system. It was about understanding the shape of the problem.

I started by testing the separation itself. What happens when I use AI in a dedicated personal space rather than inside my business context?

I used ChatGPT Pro to think through what a personal assistant layer might need to capture. I looked at categories like reminders, recurring tasks, appointments, household admin, family admin, small errands, things to buy, things to book, and decisions I need to make.

I also started thinking about where Codex might fit.

ChatGPT is useful for conversation, reflection, sorting and structuring. Codex opens up the possibility of turning some of those ideas into a more structured system over time: files, workflows, prompts, lightweight tools, or a personal OS setup that is more persistent than a chat thread.

But I deliberately kept the first week simple. Before building anything, I wanted to understand what I actually needed.

And what I needed was not more complexity.

I needed a clearer place to put things.

The value question

One of the most interesting parts of this experiment is the unspoken question of value.

In business, it is fairly straightforward to justify AI. If a tool saves time, improves output, speeds up a process or reduces manual work, the value is easier to see.

But how do we value stress relieved?

How do we value fewer loose ends?

How do we value not having to carry so much in our heads?

That is the part I am most interested in right now. I already know AI can help me produce more. I have seen that in work, writing, research and systems thinking.

But for this experiment, I am less interested in whether AI can help me produce more. I am interested in whether it can help me carry less.

That feels like a different question.

And possibly a more useful one.

Boundaries and privacy

Personal AI also brings a different set of considerations.

Using AI to support work tasks is one thing. Using it to help with household admin, family admin and personal systems is another layer entirely.

There are obvious privacy and security questions. What information am I comfortable sharing? What should stay out of the system entirely? What should be kept as non-confidential examples only? What should AI help me sort, but not action?

For now, I am keeping the experiment bounded.

I am not handing over my inbox, calendar, messages or private information and hoping for the best. I am not giving AI access to everything. I am comfortable using it to help me structure non-confidential admin, organise ideas, clarify next actions and think through systems.

But I am not expecting it to make decisions for me. And I am not expecting it to take actions on my behalf.

In many ways, I would not expect a new human assistant to immediately know what to decide or do either. They would need context, boundaries, instructions and trust built over time.

AI is no different.

At this stage, the value is in helping me organise, clarify and prepare. Not in handing over responsibility.

What I learned in Week 1

The biggest insight from Week 1 is that AI for personal admin is not simply a smaller version of AI for business.

It is a different layer.

Business AI often starts with the question: how can this save time or increase output?

Personal AI starts with a softer but equally important question: how can this reduce the invisible load?

For me, the most useful part so far has not been automation. It has been having a dedicated space to think through life admin without immediately turning it into business work.

A space to brain dump, sort, and see the shape of the mess.

That alone has been useful.

The next step is to see whether this can move from reflection into rhythm. Can I create a weekly personal admin system that I would actually use? Can AI help me review what is outstanding, group tasks into useful buckets, and turn scattered thoughts into next actions?

That is what I want to test next.

For now, my Week 1 takeaway is simple: AI for work is everywhere, but maybe one of the most useful questions is not only how AI can help us do more.

Maybe it is whether AI can help us hold less in our heads.

Note: This piece was drafted from my handwritten notes and co-edited with ChatGPT.