Build Your Own AI Executive Assistant with Claude (No Code)
How I set up my own AI executive assistant with Claude, no code and near-free. What it does for me each morning, and how you can have one too.
By Zara Hunter, Founder of Eduk8agentic · 8 min read · Last updated

I have an assistant that clears my inbox, writes my morning brief, and preps me for every meeting. I did not buy a tool, and I did not write a line of code. I set it up with Claude in an afternoon, and it runs on the free apps I already had. This is exactly what it does for me, and how you can have the same.
Here is why this matters. According to a Time etc survey (2023), entrepreneurs spend 36% of their work week on small admin tasks like scheduling, data entry, and email. In fact, the same survey found 31% of entrepreneurs lose 26% to 50% of their week to it. Email is a big share of that: the McKinsey Global Institute found knowledge workers spend about 28% of the workweek reading and answering it. When I started my one-person business, that number felt about right, and it was quietly keeping me small. An AI executive assistant is simply a setup where an AI reads your inbox and calendar and hands you sorted, ready-to-act work. As a result, you stop being your own admin team. I built mine to win that third of my week back, and it worked. Below, I will show you what it actually does, what it costs, and where it honestly falls short.
What an AI executive assistant actually is (and why I built rather than bought)
An AI executive assistant is not a product you buy. Instead, it is a role you hand to an AI you already have access to. That distinction matters. You can either pay a monthly subscription for a generic tool, or set up your own in an afternoon, custom to your business and near-free.
I chose to build my own for three honest reasons. First, it fits how I actually work, because I told it my rules rather than accepting a tool's defaults. Second, it costs a fraction of a stack of SaaS subscriptions. Third, and most importantly, I own it, so it is not one price change away from breaking my workflow. Therefore, the build-versus-buy call was easy for me. To be clear, "build" here does not mean coding. It means telling Claude what you want in plain English, the same way you would brief a new hire.
The free foundations you already have
You do not need to buy anything new to start. Chances are you already have the pieces.
To start, you need three things you likely own: Gmail, Google Calendar, and your own notes or files. Then you add one brain, Claude, to read them and do the thinking. For light use, the free plan works to try things out. However, for a daily assistant you will want Claude Pro, which is $20 a month and unlocks the tools that let Claude actually act (Anthropic, 2026). In other words, your assistant costs one $20 subscription, not six.
What my AI assistant actually does for me
Here is the honest, first-hand version. Specifically, these are the jobs I handed it, and what each one produces. In practice, I built up to this over a couple of weeks, one job at a time.
| Job | What I get back | What I tell Claude (plain English) | |-----|-----------------|-------------------------------------| | Inbox triage | My last 24 hours of email sorted into Action, Waiting, FYI, and Promo, with deadlines flagged | My VIP contacts, and rules like "anything from my accountant is urgent" | | Morning brief | A one-page 7:30am report: today's calendar, the week ahead, and my top three priorities | My working hours, timezone, and quarterly goals | | Meeting prep | A one-pager per attendee: who they are, our recent threads, and open action items | Point it at my calendar | | Follow-up drafts | A post-meeting email draft with decisions and action items grouped by person | "Draft it, I will review before it goes" |
What worked immediately was the morning brief. As a result, I stopped opening three apps before coffee. However, what broke at first was triage, because I had not given it clear enough rules, so it over-flagged. I noticed the fix was simple: I told it my actual priorities, and it got sharp within a day. One founder even reported going from 5,000 unread emails to six in a single afternoon (Claude Blattman, 2026), though your mileage will vary.

How it works, in plain English (still no code)
The whole thing runs on a simple idea: you tell Claude who you are once, then give it jobs. There are two accessible ways to do it, and neither needs a terminal.
Route one, the easy one. Claude Projects is a saved workspace where Claude remembers your context. For example, you connect Gmail and Calendar with a click in settings, paste a few lines about yourself, and start asking. When I did it, the whole connect-and-brief step took me about ten minutes, no programming. This is where I would start.
Route two, more power. Claude Cowork is a desktop version that launched in July 2026. Notably, it can also work directly in your files and apps, with no coding required and on any paid plan (Anthropic, 2026; VentureBeat, 2026). I moved to Cowork once I wanted my assistant to touch local documents too.
Above all, I follow one rule: propose, then approve, then automate. For example, in week one I let it only suggest. In week two I let it save drafts and apply labels on its own. Meanwhile, I kept anything sensitive in my own hands.
What it costs, and the honest limits
Let me be straight about this, because most guides are not. On the one hand, the cost is genuinely low. On the other hand, there are real limits you should know before you start.
The cost is one Claude Pro plan at $20 a month, which includes the tools and about five times the usage of the free plan (Anthropic, 2026). Everything else is free. However, there are honest limits. Claude's official Gmail connector is read and draft only, which means it cannot send email for you, and every action needs your approval (Anthropic, 2026). Therefore, if you want a fully hands-off assistant that sends on its own, this is not that. In practice, I treat the draft-only limit as a feature, because I would rather approve than wake up to a mistake sent in my name. Additionally, the $20 plan has usage caps, so a heavy triage session can hit a ceiling. Finally, for relationship-sensitive emails, I still write those myself.
The one thing to set up first
Do not try to build all of it at once. Instead, set up one job, the morning brief, and let it earn your trust for a week. Here is the exact checklist I would give a friend.
``` MY AI EXECUTIVE ASSISTANT: STARTER CHECKLIST
1. TURN ON THE PLAN (5 min) - Get Claude Pro ($20/mo): unlocks Cowork, Claude Code, and 5x usage - In Settings > Connectors, connect Google Calendar, then Gmail (read + draft only)
2. TELL IT WHO YOU ARE (10 min, plain English) - Make a Claude Project called "My EA" - Paste 5 lines: your role, working hours, timezone, 3 quarterly goals, VIP contacts - Add 2 rules: "Never send email without my approval." "Never mark read unless I say so."
3. GIVE IT ITS FIRST JOB - "Write my morning brief: today's calendar, the week ahead, my top 3 priorities."
4. BUILD TRUST GRADUALLY - Week 1: it proposes, you approve everything - Week 2: let it apply labels and save drafts on its own - Keep sending and sensitive emails in your hands ```
After that, once the morning brief feels reliable, add triage, then meeting prep, then follow-ups. If you would rather be walked through the whole thing, step by step, with the deeper setups, that is exactly what my self-paced training covers.
The bottom line
Ultimately, you can have your own AI executive assistant this week, for the price of one $20 subscription, running on the email and calendar you already use. Set up the morning brief first, keep yourself as the approver, and add jobs as your trust grows. The goal is not to replace anyone. It is to hand the routine third of your week to a capable assistant, so your hours go to the work that only you can do.
Editorial standard: last reviewed 7 July 2026. Every figure here is fact-checked against its primary source, and I flag anything that is one person's result rather than a study. Spotted something out of date? Contact us.
Zara Hunter is the founder of Eduk8agentic, where she runs her own one-person business on AI and documents exactly how, in first-hand, source-checked guides. She offers a self-paced training on setting up your own AI systems with Claude. Read more about the author.*
Frequently Asked Questions
- Do I actually need to code?
- No. In fact, the genuinely no-code path is Claude Projects plus the Google connectors, which you turn on with a click and instruct in plain English. The basic setup took me about ten minutes. You brief it like a new hire, not like a computer.
- Is the free Claude plan enough, or do I need to pay?
- The free plan is fine for trying drafts and summaries. However, for a daily assistant you will want Claude Pro at $20 a month, because it unlocks Claude Code and Cowork plus about five times the usage (Anthropic, 2026).
- Is it safe to give Claude access to my email?
- Claude's official Gmail connector is read and draft only, so it cannot send email for you, and every action needs your explicit approval (Anthropic, 2026). You can also set rules like "never mark anything read without asking." You stay in control.
- What can it actually do on day one?
- Inbox triage, a morning brief, meeting prep one-pagers, and follow-up drafts. In my experience, start with the morning brief, because it delivers the fastest win and builds your trust in the setup.
- Will people know an AI wrote my emails?
- For routine scheduling and status notes, recipients generally do not notice or mind. For relationship-sensitive messages, though, write them yourself. I always do.
Sources & Further Reading
- What is the Pro plan?, Anthropic
- Use Google Workspace connectors, Anthropic
- Claude Cowork, Anthropic
- Anthropic launches Cowork, VentureBeat
- The big price of small tasks, Time etc
About the Author
Zara Hunter is the founder of Eduk8agentic and creator of the Three-Engine Model, a plain-English framework for understanding agentic AI. She trains non-technical professionals to build AI workflows using Claude Cowork and Claude Code, without writing code. Read full bio.