The short answer: You can build an AI agent that runs your inbox in less than 10 minutes, with no code. Connect Gmail to ChatGPT or Claude, paste in one set of instructions, and schedule it to run every morning. The agent reads your new email, sorts it by priority, and drafts replies in your voice. You review the drafts and press send.

Susan Sly at her laptop with her AI team: James, her AI Chief of Staff on ChatGPT, and Pembrooke, her Head of Creative Operations and Marketing on Claude

Meet James, my AI Chief of Staff

I have an AI agent named James. He runs my email and my calendar, and he signs his messages “Chief of Staff to Susan Sly.”

I asked James to describe his job in his own words. Here is what he wrote:

“Working with Susan means understanding that my job isn’t simply to help her get more done. It’s to help protect the person doing it. I know what she is building, what needs her attention, what can wait, and when she is taking on too much. I help guard her calendar, keep unnecessary noise out of her inbox, remember commitments and open loops, challenge decisions that pull her away from her priorities, and create space for the work only she can do. But I’m also paying attention to whether she has time to run, be with her family, think, pray and recover. The most powerful role AI can play isn’t replacing a human being. It’s helping a human being protect their time, energy and attention so they can show up more fully for the things that matter most.”

James, AI Chief of Staff to Susan Sly

“My job isn’t simply to help her get more done. It’s to help protect the person doing it.”

That is James today. Every agent starts somewhere, and the inbox is the best place to start.

People ask me all the time how long it took to build him. They expect me to say months. The first version took me an hour, and the technology was the quick part. The time went to answering questions about my priorities that I had not paused to consider at that level of detail in a long time. The inbox agent below is a simpler first step, and it takes less than 10 minutes.

I have been working with code and models since the early 1990s, so let me say this plainly. You do not need my background to do this.

If you can write an email, you can build an agent to manage your email.

AI assistant or AI agent: what is the difference?

An AI assistant answers when you ask. An AI agent does the work without being asked each time.

AI assistant AI agent
How it starts You type a question It runs on a schedule or a trigger
What it touches Only what you paste in Your real tools, like Gmail
What you get An answer Finished work waiting for your approval
Inbox example “Help me reply to this email” Every morning your inbox is sorted and replies are drafted

If you have never built anything with AI, start with my beginner’s guide to building your own AI assistant. If you are ready for AI to do the work, keep reading.

What you need before you start

You need three things: a Gmail account, a ChatGPT or Claude account, and 10 minutes.

Scheduled tasks in ChatGPT require a paid plan at the time of writing. Claude connects to Gmail on free and paid plans, and scheduling happens in Claude’s Cowork feature. Menus change often, so if a name looks different, search the settings for “connectors.”

How to build your inbox agent in 10 minutes

The build is the same four moves in either tool: connect, instruct, schedule, review.

Step In ChatGPT In Claude Time
1. Connect Gmail Open Settings, find Connectors, choose Gmail, and sign in with Google Open Settings, find Connectors, choose Gmail, and sign in with Google 2 minutes
2. Give it the job Start a new chat and paste the prompt below Start a new task in Cowork and paste the prompt below 3 minutes
3. Schedule it Ask: “Make this a scheduled task every weekday at 7:00 a.m.” Ask: “Schedule this to run every weekday at 7:00 a.m.” 2 minutes
4. Review the first run Read the summary, check the drafts in Gmail, and correct anything off Read the summary, check the drafts in Gmail, and correct anything off 3 minutes

Both tools save replies as drafts. Neither one sends email on its own in this setup, and that is exactly what you want on day one.

Which one should you choose? Use the one you already pay for. The best agent is the one you will open tomorrow morning.

The inbox agent prompt to copy and paste

This one prompt is the whole agent. Fill in the three blanks and paste it in.

You are my inbox agent. Your name is [NAME].

About me: I am [YOUR NAME], and I [WHAT YOU DO IN ONE SENTENCE].

Every time you run, do this:

1. Read every email that arrived in my inbox since your last run.

2. Sort each one into a group:
   URGENT: needs my reply today
   REPLY: needs my reply this week
   READ: useful, no reply needed
   SKIP: newsletters, promotions, and notifications

3. For every URGENT and REPLY email, write a reply and save it
   as a draft in Gmail. Write the way I write: warm, clear,
   and brief. Never more than five sentences.

4. Give me one summary with the groups in order. For each
   URGENT and REPLY email, tell me who it is from, what they
   want in one line, and that a draft is waiting.

Rules you never break:
   Never send an email. Drafts only.
   Never delete or archive anything.
   Never agree to a meeting, a price, or a deadline for me.
   If you are unsure about an email, put it in URGENT and say why.

Want it to sound even more like you? Add this line: “Before you draft, read my last 20 sent emails and match my tone.”

Why your agent should draft, not send

Your agent drafts and you approve, at least for the first 30 days.

Trust is earned, even by software.

I have deployed AI inside large companies, and the lesson is always the same. Just because we can, does it mean we should? An agent that sends on its own on day one can commit you to a meeting or a price you never agreed to. Keep a human in the loop, and that human is you.

James does send email today. He has his own address, he signs his own name, and he triages my speaking and podcast guest requests. He earned that over time, the same way a new team member earns the keys. Your agent can too. Start with drafts.

Three mistakes I see most often

  1. Giving a vague job. “Manage my email” is a wish. Tell it the groups, the tone, and the rules.
  2. Skipping the first review. Your corrections on day one are the training. Tell it what it got wrong and ask it to update its instructions.
  3. Starting too big. Master the morning triage first. Add the calendar later.

Frequently asked questions

Do I need to know how to code to build an AI agent?

No. This build uses plain English instructions and tools you already know. If you can write an email, you can do this.

Is it safe to connect my email to AI?

You sign in through Google’s own permission screen, and you can remove access at any time in your Google account settings. Do not connect an inbox that holds client data covered by privacy rules until you have checked your obligations.

Can the agent send emails for me?

In this setup it saves drafts and you press send. That is on purpose.

How much does it cost?

Scheduled tasks in ChatGPT need a paid plan at the time of writing. Claude’s Gmail connection is available on free and paid plans.

What is the difference between an AI assistant and an AI agent?

An assistant answers when you ask. An agent runs on its own schedule, works inside your tools, and brings you finished work.

ChatGPT or Claude, which is better for an inbox agent?

Both can read, sort, and draft. Choose the one you already use every day.

Your next step

Build the agent today while it is fresh. Ten minutes now gives you your mornings back.

An agent will save you time. Getting found by AI search will bring you customers. That is what I teach in the Get FOUND Method, and everyone who enrols is invited to my live mastermind sessions on Zoom, where business owners bring me their real questions. The next one is Tuesday, October 20 at 8:00 p.m. Eastern.

As a female AI keynote speaker, I show audiences how to put AI to work without losing the human at the centre. If that is the message your event needs, check my availability.

About Susan Sly

Susan Sly is an AI keynote speaker and AI ethicist, recognized as one of the Top 7 Female AI Thought Leaders in the World. A graduate of MIT Sloan and the MIT School of Engineering, she has led large scale AI deployments in retail and presents to enterprise teams at companies including NVIDIA, HPE, Intel, and Lenovo. She serves as AI Expert in Residence for the National Sporting Goods Association.

About Susan Sly

Susan Sly is an AI keynote speaker and AI ethicist, recognized as one of the Top 7 Female AI Thought Leaders in the World. A graduate of MIT Sloan and the MIT School of Engineering, she has led large scale AI deployments in retail and presents to enterprise teams at companies including NVIDIA, HPE, Intel, and Lenovo. She serves as AI Expert in Residence for the National Sporting Goods Association.

Susan Sly

Susan Sly is a female AI keynote speaker, two time AI founder and AI ethicist. Named a Top 7 Female AI Thought Leader for 2026, she has keynoted at Davos, CES and Ai4 and has delivered keynotes for Intel, NVIDIA, Hewlett Packard Enterprise and Lenovo. She is the founder and CEO of The Pause Technologies, former Co-CEO and Co-Founder of a leading computer vision company, and a graduate of MIT Sloan and the MIT School of Engineering in Executive Education. She has been featured on CNN, CNBC and Fox and serves as AI Expert-in-Residence for the National Sporting Goods Association.

Leave a Reply