Meet My AI Team

Meet My AI Team

A few years ago, if you had told me I would someday have an entire team working with me around the clock, I probably would have pictured employees, assistants, researchers, editors, and programmers.

Today, much of that team is AI.

And I have learned something important:

I don’t want one AI trying to do everything.

I get much better results when I give different AI tools specific jobs, just as I would with people in a real company.

So, let me introduce you to my current AI team.

Mel — CEO

Before we get to the AI, somebody still has to run the company. :-)

That’s me.

AI can research, analyze, write, organize, check files, transcribe books, and help execute projects.

But I still decide:

• What we work on
• What gets published
• What opportunities are worth pursuing
• What risks are acceptable
• What gets approved
• Where the business is going

AI is extraordinarily useful, but I remain the decision maker.

Now, meet the rest of the team.

ChatGPT — “Chief”

ChatGPT has become my AI Chief of Staff.

Chief is the AI I use for the big picture.

Its role includes:

• Business strategy
• Publishing strategy
• Decision analysis
• Project planning
• Reviewing numbers
• Tracking publishing progress
• Comparing options
• Creating procedures
• Spotting problems before I act
• Helping me maintain continuity across a large number of projects

Chief knows how the different parts of my publishing business fit together.

When something happens—a royalty increase, a publishing problem, a new opportunity, a change in priorities—this is usually where I come first.

It is less like asking a chatbot a question and more like having an executive adviser sitting beside me.

ChatGPT Work

Work is different.

This is where ChatGPT can move beyond simply discussing a project with me and start working through larger multi-step jobs.

I use Work for projects involving things such as:

• Research
• Files
• Analysis
• Organizing information
• Multi-step workflows
• Producing finished work products

Think of regular ChatGPT as the executive office.

Work is the operations department.

Codex

Codex is my technical specialist.

When we are dealing with files, spreadsheets, structured data, automation, or things that need precise computer work, Codex can become extremely valuable.

For example, we are building what I call my Publishing Command Center.

That includes catalog data, Amazon opportunity research, operating procedures, change logs, backups, validation, and controlled updates.

Chief and I can decide what needs to happen.

Codex can help perform the technical work.

That separation is important.

One AI helps make the decision.

Another helps execute it.

ChatGPT Workspace Agents

Workspace Agents represent another step toward having specialized AI workers.

Instead of every job coming back to the same general AI conversation, an agent can be given a particular assignment or area of responsibility.

That opens the door to something I find fascinating:

an AI company inside my company.

You could eventually have separate agents handling research, catalog analysis, financial monitoring, production checks, marketing research, and other recurring jobs.

We are still experimenting with exactly how far to take this.

ChatGPT for Excel

Spreadsheets are becoming increasingly important in my business.

I publish a lot of books, so there is a tremendous amount of information to keep organized:

Titles.

Authors.

ISBNs.

ASINs.

Formats.

Prices.

Royalties.

Amazon status.

Research opportunities.

Production status.

With ChatGPT working with Excel, AI can help analyze and understand that data instead of me manually digging through thousands of rows.

For a large publishing catalog, that has enormous potential.

Claude isn't just one tool in my operation either.

I use Claude in several different ways depending on what I need done.

Claude Chat

Claude Chat is the traditional Claude conversation.

I use it for:

• Writing
• Analysis
• Reviewing documents
• Thinking through projects
• Research planning
• Second opinions
• Working with long documents

Sometimes I deliberately give the same problem to both Claude and ChatGPT.

I don't necessarily want them to agree.

I want another set of AI eyes on the problem.

Claude on Chrome

Claude on Chrome moves AI closer to the actual browser.

Instead of only telling me what I could do on a website, it can help work with what is happening inside the browser.

That makes it useful for jobs involving web-based systems and repetitive browser work.

For someone running an online publishing business, much of the company already lives inside a browser.

Amazon.

Publishing platforms.

Research websites.

Business tools.

That makes browser-based AI particularly interesting.

Claude Cowork

Cowork is where Claude starts behaving more like a digital employee working alongside me.

I can give it a project involving files and computer work and have it work through the assignment rather than simply answering a question.

I've used this kind of capability for substantial publishing projects.

The key difference is:

Chat AI talks with you.

Agentic AI can work on the assignment.

That is a very big change.

And Claude is now adding the ability to run multiple tasks in parallel.

That could eventually mean giving several independent research assignments to Claude at the same time.

Instead of one researcher…

I may effectively have five.

Astra has a very specific role on my team.

Book production and transcription.

I work extensively with older public-domain books, including historical Catholic works.

Some exist only as scanned books.

Turning one of those scans into a clean, usable manuscript can require checking hundreds of pages.

Astra can work through those large transcription projects and help produce material that can eventually become new print and digital editions.

We recently completed a book of about 380 scanned pages this way.

This is exactly why I like specialization.

I don't need every AI to be the best book transcription system.

I need one member of the team that is very good at that job.

People sometimes ask which AI is “best.”

I increasingly think that is the wrong question.

Imagine asking:

“Who is the best employee?”

Best at what?

Accounting?

Research?

Programming?

Editing?

Management?

Sales?

The same applies to AI.

Instead of searching for one perfect AI, I am building a team.

ChatGPT Chief helps me think and manage.

ChatGPT Work handles substantial workflows.

Codex handles technical execution.

Workspace Agents can become specialized workers.

ChatGPT for Excel helps turn business data into useful intelligence.

Claude Chat gives me another powerful thinking partner.

Claude on Chrome works closer to the browser.

Claude Cowork tackles computer-based projects.

Astra handles specialized book-production work.

And I remain CEO.

The Most Interesting Part

I think we are only at the beginning.

My goal isn't simply to save a few minutes writing an email.

I want to build an operation where AI allows one person to accomplish what previously required an entire small company.

Research can run while I am working on something else.

Books can move through production.

Catalog opportunities can be analyzed.

Business data can be monitored.

Problems can be identified.

Projects can be checked.

And multiple AI systems can even check each other's work.

For an entrepreneur, that is where things become extremely interesting.

I have been in business for a long time.

I have seen computers change business.

I watched the Internet change business.

I watched Amazon change publishing.

And now I am watching AI change what one individual is capable of building.

I intend to find out just how far that can go.

The team is getting bigger. :-)

How many AI “team members” do you currently use—and have you started assigning different AI tools different jobs?

To Our Success!

Mel Waller