How to Build Your Own AI Assistant: A Beginner’s Guide for Non-Technical People
By Susan Sly
If you have ever looked at the headlines about AI agents and thought, “That’s for engineers, not for me,” I want to change your mind in the next ten minutes. You do not need to write a single line of code to build your own AI assistant. You need a clear problem, a good tool, and a little willingness to feel awkward for an afternoon.
I know this because I watched it happen in real time — with nearly 500 people who were certain they couldn’t do it.
Quick answer: How do beginners build an AI assistant?
A beginner builds an AI assistant in five steps: (1) pick one repetitive, time-draining task; (2) choose a no-code tool you already have access to, such as ChatGPT, Claude, or Microsoft Copilot; (3) give the assistant a clear role and your context (preferences, voice, constraints); (4) test it on real work and correct it like you would onboard a new hire; and (5) expand to a second assistant only once the first one is reliable. The entire process can take an afternoon, and no programming is required.
Now let me show you what that looks like in practice.
My afternoon with 500 reluctant agent-builders
Not long ago, I was in Athens delivering a keynote for a Fortune 500 company. The morning was the talk. The afternoon was the part I was most curious about: a hackathon for roughly 500 employees from a specific division in the company — almost none of them technical. These are precise, risk-aware professionals whose entire job is to be careful, which made them exactly the kind of audience you might expect to resist building something with AI. The assignment was deceptively simple: using Microsoft Copilot, each person would build an AI agent to solve one problem they faced in their actual job.
When I scanned the room at the start, I saw a wave of quiet panic. Folded arms. The specific facial expression of a capable adult who suddenly feels like a beginner again. A few people clearly wanted to be anywhere else.
Then something shifted. One team built an agent to triage a specific universal problem for many companies – social media compliance . Another team automated the first draft of a routine compliance memo they had dreaded for years. Yes another team created an assistant to scan lengthy contracts and flag the clauses that needed a closer look — work that used to eat an entire morning. The panic turned into leaning forward. By the end of the afternoon, the room had a completely different energy — the satisfaction of people who had just discovered they were far more capable than they thought.
That is the real story of AI assistants for beginners. The barrier was never technical skill. It was the belief that you needed it.
What is an AI assistant — and how is it different from an AI agent?
An AI assistant is a tool that helps you complete tasks through conversation — answering questions, drafting writing, organizing information, and offering recommendations. You stay in the driver’s seat and direct each step.
An AI agent goes one level further: you give it a goal and some boundaries, and it can take a sequence of actions on its own to reach that goal, checking in with you along the way.
For most beginners, the line blurs, and that is fine. You will likely start with an assistant and gradually give it more responsibility until it starts to feel like an agent. The point is not the label. The point is offloading work that drains your time and attention so you can focus on what only you can do.
The proof that non-technical people can do this: Jesse Genet
If you need one example to convince you, it is Jesse Genet. The Cut profiled how Genet — a former startup founder who had never written a line of code and had never even opened a developer terminal — built an entire household of specialized AI agents to run her family and homeschool her children, all while caring for a baby in a carrier. You can read the piece here: The Cut — Jesse Genet on AI agents at home.
What makes her approach so instructive is that she treats each agent like an employee. Every one has a defined role — a curriculum planner, a finance manager, a content creator, a coding helper — a written “personality” and scope, and an onboarding process. She does not ask a single agent to do everything. She builds a small team of specialists, each excellent at one thing.
That framing — hire specialists, don’t build one overworked generalist — is the most useful mental model I know for beginners. Hold on to it.
My own AI assistants: Henry and my email team
I do not just teach this. I live inside it. Here are the two assistants I rely on every single day.
Henry: my pocket chief of staff (built on ChatGPT)
I treat ChatGPT on my phone as a personal assistant, and I have named him Henry. With the current generation of ChatGPT — now on the GPT-5.5 model with its upgraded memory — Henry carries context across all of our past conversations automatically. I no longer have to re-explain who I am, how I travel, or what I need. He simply remembers.
Here is what that looks like in real life:
- He knows I am gluten free and filters every restaurant and meal recommendation accordingly.
- In Athens, he picked excellent restaurants that fit both my dietary needs and my packed schedule.
- A pharmacist recommended a Greek probiotic, and Henry read the label from a photo and explained the ingredients to me on the spot.
- He plans my workouts around my speaking and travel calendar, so movement still fits even on brutal travel days.
- And in one of the more human moments I’ve had with a piece of software, he reminded me to slow down and savor the moment in Athens — because he knew I had just come through a long bout of Norovirus after speaking in Davos, and that I tend to push too hard.
Short of actually booking my flights and hotels, Henry does almost everything an executive assistant would: coaching, tips, schedule management, organizing my scattered thoughts, and taking the guesswork out of planning. He is not a novelty. He is infrastructure.
My email assistant (built on Claude)
My second assistant lives in my inbox, and I built it on Claude. It prioritizes my inbox, drafts replies in my voice, reviews attachments so I understand what’s inside before I open them, and flags the correspondence that truly needs me. Email used to be a tax on my attention. Now it is a managed system, and the assistant handles the first pass so I handle only what matters.
Notice the pattern: two assistants, two clear jobs. A pocket chief of staff and an inbox manager. Specialists, not a single overloaded generalist.
How to build your own AI assistant in 5 steps
Step 1: Pick one problem worth solving
Do not start with the technology. Start with friction. Ask yourself: What task do I do repeatedly that I dread, that drains an hour I’ll never get back, or that I keep putting off? Inbox triage, meeting prep, first-draft writing, research summaries, and travel planning are all excellent starting points. One problem. That is the whole assignment.
Step 2: Choose a beginner-friendly, no-code tool
You almost certainly already have access to at least one of these:
- ChatGPT — outstanding general-purpose assistant with strong memory; ideal for a personal “chief of staff” like Henry.
- Claude — excellent for writing in your voice, reviewing documents and attachments, and thoughtful long-form work.
- Microsoft Copilot — built into the Microsoft 365 tools many companies already use, which is exactly why it worked so well for a corporate hackathon.
You do not need all three. Pick the one closest to where your work already lives.
Step 3: Give it a role and your context
This is the step beginners skip — and it is the difference between a generic chatbot and an assistant that feels like yours. Tell it plainly who it is and who you are. For example:
“You are my executive assistant. I am a keynote speaker who travels constantly and is gluten free. I value concise, direct communication. When I share my schedule, help me protect time for workouts and recovery.”
Then let it remember. Turn on memory, save your key preferences, and let the context compound over time. The more it knows, the more useful it becomes.
Step 4: Test on real work and correct it like a new hire
Your first results will be about 80 percent right. That is normal. Do not abandon the assistant — coach it. Tell it what landed and what didn’t. “Too formal.” “You missed the deadline in the attachment.” “This is my actual voice — match it.” Every correction makes the next output sharper. You are onboarding an employee, not flipping a switch.
Step 5: Expand only once the first one is reliable
Resist the urge to build five assistants on day one. Get one genuinely dependable first. Once it is, add a second specialist for a different problem — exactly as Jesse Genet built her team one role at a time. Reliability first, then expansion.
Beginner AI assistant tools at a glance
| Tool | Best for | Why beginners love it |
| ChatGPT (GPT-5.5) | A personal all-rounder / chief of staff | Strong cross-chat memory; reads photos and documents; conversational |
| Claude | Writing in your voice, reviewing attachments | Nuanced drafting; excellent with documents and tone |
| Microsoft Copilot | Work tasks inside Microsoft 365 | Lives where corporate work already happens; no new login |
Common beginner mistakes to avoid
- Building a generalist that does everything. Specialists outperform. Give each assistant one clear job.
- Skipping the context. An assistant with no knowledge of you produces generic results. Tell it who you are.
- Quitting after the first imperfect output. The magic is in the correction loop, not the first try.
- Waiting until you “understand AI” to begin. You learn this by doing, not by studying. The 500 people in Athens proved it in a single afternoon.
What’s next for me — and a challenge for you
I am heading on a trip to Kenya, and I am taking a page directly from Jesse Genet’s playbook. As I launch a new podcast, work with multiple consulting clients, and continue to grow The Pause Technologies and Amsara, I am setting up more agents — a small, deliberate team of specialists, each owning a piece of the work that used to live entirely in my head.
That is the invitation I want to leave you with. You do not need permission, a computer science degree, or the perfect plan. You need one problem and one afternoon. The panic I saw in that room in Athens was real — and so was the satisfaction that replaced it.
Pick your one problem this week. Build your first assistant. Then come tell me what it does.
Frequently asked questions
Do I need to know how to code to build an AI assistant? No. Modern tools like ChatGPT, Claude, and Microsoft Copilot are entirely no-code. You build and direct your assistant through plain conversation. Jesse Genet built a whole household of agents without ever writing a line of code.
What is the easiest AI assistant for a complete beginner? Start with whichever tool is closest to your existing work. If you live in Microsoft 365, use Copilot. If you want a personal all-purpose assistant, ChatGPT is the most beginner-friendly. For writing and document review, Claude is excellent.
How long does it take to set up an AI assistant? A useful first version can be built in an afternoon. Nearly 500 non-technical employees each built a working agent in a single hackathon session. Refining it to fit your voice and workflow happens over the following days as you correct it.
What’s the difference between an AI assistant and an AI agent? An assistant helps you complete tasks step by step while you direct it. An agent takes a goal and carries out a sequence of actions more independently, checking in along the way. Beginners typically start with an assistant and grow into agents.
Can an AI assistant replace a human executive assistant? It can handle a large share of the work — scheduling support, drafting, research, reminders, organizing information, and recommendations. My assistant Henry does almost everything an EA would, short of actually booking my flights and hotels. It augments your capacity rather than replacing human judgment.
Is it safe to give an AI assistant personal information? Share preferences and context that make the assistant useful, but review each tool’s privacy and memory settings, avoid storing highly sensitive data such as passwords or financial credentials, and use private or temporary chat modes when appropriate.
Key takeaways
- The barrier to building an AI assistant is belief, not technical skill.
- Start with one problem, one no-code tool, and clear context about who you are.
- Treat each assistant like an employee with a single, defined role — and onboard a team of specialists over time, the way Jesse Genet did.
- Correct it like a new hire. The first output is rarely perfect; the value is in the feedback loop.
- You can build your first assistant in an afternoon. The hardest part is starting.
Susan Sly is a keynote speaker, entrepreneur, and technologist who was voted one of the Top 7 Female AI Thought Leaders of 2026. She is building the next chapter of her work — including a new podcast, The Pause Technologies, and Amsara — alongside a growing team of AI assistants.
