How to Train ChatGPT to Work Like a Chief of Staff (Not a Clueless Intern)

The Single Best Thing I Did to Make AI Actually Useful

Jun 16, 2025

I've spent most of my career being the right-hand person—executive assistant, operations lead, strategic thought partner.

I made it my business to know everyone else's preferences, guarded their calendars like state secrets, and smoothed every bump so they could focus on big-leverage work.

But over the last 6 months I’ve been slowly building my own team of AI assistants and ‘right hands’ to help me get work done.

And honestly it feels luxurious.

After all these years, being SERVED in the same way I have served is just mind blowing.

But I still see so many people treating ChatGPT or other AI tools like a stranger at a bus stop:

Over-explaining everything.
Explaining things over and over again
Crossing their fingers hoping for a decent answer
Getting a politely irrelevant answer.
Repeat (over and over again)

It’s exhausting & time consuming. And AI is supposed to be saving us time and headspace!

The Personal Context AI Advantage

The fix is the same move every great leader makes: hire someone who can understand your context. They know your quirks, they understand your preferences and they are continually paying attention and learning.

When I was someone's right hand, I knew:

They think better on walks than in conference rooms
They need prep time before big presentations
They make bolder decisions in the morning
They hate unnecessary meetings on Fridays

This intimate knowledge let me anticipate needs, not just respond to requests. And that's precisely what your AI needs to transform from a glorified search engine into a true thinking partner.

The key is context.

I run two always-on co-pilots built on this principle:

Creative Co-Pilot—trained on my writing style, frameworks, and product suite. Ghostwriter, brainstorm buddy, and blunt editor rolled into one.
Chief of Staff—knows my goals, energy rhythms, key relationships, even that 3pm is coffee-then-deep-work time, not meeting time.

Because they hold my context:

I can ask them to time block my day for tomorrow and I know they won’t schedule deep work time in the afternoon.
I can ask them to write a CTA for this article & it understands what products I have in the mix and writes something that fits.
I can ask them to suggest some ideas for my next video and it already knows what my past videos were, what keeps coming up in the community and my content pillars and can make thoughtful suggestions.

The difference isn't just incremental—it's transformative. Context turns AI from a clever calculator into the kind of assistant I used to be for others.

What Your AI Chief of Staff Should Know About You

Building this kind of contextual intelligence means loading your AI with everything a human chief of staff would absorb over months of working together. Think about if you had to hire a human - what would they need to know? It’s exactly the same process for an AI Chief of Staff.

Your Big Picture

What you're working towards this quarter
Current projects & why they matter
How you prefer to make decisions

Your Role & Responsibilities:

What you're actually responsible for day-to-day
Where you add the most value
What's on your plate vs. what you delegate

Your Work Style:

When you do your best thinking
How you like meetings structured (or if you hate them entirely)
What kind of communication works with different people
Energy rhythms throughout the day/week

Your People:

Key colleagues & how you work with each one
Clients or stakeholders & their preferences
Who needs what kind of updates & when

Your Quirks:

Things that drive you nuts
Preferences that aren't obvious but matter
The difference between your Monday energy & Friday energy

Don't worry, this isn't as daunting as it seems. All you need to do is open ChatGPT (or Claude) turn on voice mode an start talking.

Start by saying: You’re going to build a Knowledge Doc all about me that I can give to a chief of staff to get up to speed on me. I’ll brain dump all my thoughts - your job is to clean it up into a comprehensive knowledge doc (don’t summarise, we need as much context as we can).

Then go through the list above and just talk through each things.

Honestly this took me about 20 minutes to do initially and it has been the BIGGEST change in the output I get.

Once you have your knowledge doc load it into a CustomGPT or Claude Project & voila you have your very own chief of staff.

And now imagine this… You open Claude and ask Claude to give you a brief of your email inbox - is there anything important. An email comes in from someone who wants to book a zoom meeting. But you’re chief of staff knows you hate unnecessary zoom meetings. It drafts an email for you politely suggesting if they want to send through a loom first.

This is the kind of stuff I do ALL day with my chief of staff.

Where Tana Takes This to Another Level

I have my chief of staff built into Claude (because I can connect to my Google Calendar, Docs & Emails). But I also have it built right into Tana where all my notes, tasks & projects live.

And this is honestly why I love using Tana as my second brain.

Not only can I have my Chief Of Staff where I work most of the time. But it can pull in context from my notes. So I can ask it to hep me run my weekly planning session and it knows what I have coming up, what current projects I have on, that I need to block off Friday to see my nephew run cross country and that I’ve also been procrastinating on that important project for weeks.

Now that’s powerful context.

Planning my month with my chief of staff in April

Other functions my Chief Of Staff is setup to do in Tana are:

Weekly Game Plan—Monday morning, AI chief of staff reviews commitments + energy forecast, then proposes a game plan for the week.
Project Check-ins—Mid-week, run "Review Progress" on any project node. The CoS assesses what's done, what's stuck, and suggests next moves based on the full project context.
End-of-Day Debrief—"Reflect on today" pulls from my time blocks, completed tasks, and any notes I jotted down to create a coherent summary and suggest tomorrow's priorities.

The common thread? Zero mega-prompts. I converse like I would with a human who’s worked beside me for years.

I’m convinced that this is probably one of the easiest and low hanging fruit things you could right now to improve how you work with AI. And all it takes is about 10-20 minutes brain dumping to ChatGPT to get something workable.

Seriously try it and see how much better you get along with your AI assistant.

Whenever You're Ready To Build A Personal Knowledge System That Actually Works — Here Are Couple Of Ways I Can Help:

Ready to ditch rigid note-taking systems & build your own commonplace book that captures sparks without the overwhelm? My course Knowledge Alchemy teaches you how to set up a sustainable note-taking practice that feels like exploration, not homework.
If you’re curious about building your own AI assistant that can help you think, write, and create the way you work, then check out my Creative Co-Pilot workshop.