Working with your AI assistant: a guide for your first months

You don't need to read this guide

Once your project is set up, open it in ChatGPT or Claude and ask for what you need. When you finish, say "wrap up." Your assistant walks you through the rest as you go, one step at a time.

This guide is for when you want the why behind something, or ideas for what to try next. Read the part you need and skip the rest. Or ask your assistant how to do something.

What you have

• Your Sidekick folder. Sidekick Orchestration built it for you and sent it with your welcome email. It holds instructions about you and your work, the safety rules, and a place for notes. It lives on your computer, wherever you saved it.
• A project in your own ChatGPT or Claude account. You create it and connect your Sidekick folder to it, usually with Chase in your first session. Your account stays yours. Sidekick Orchestration doesn't log in to it, and Chase only sees what you choose to share.
• Your assistant. That's ChatGPT or Claude, working inside that project and following the instructions in your folder. It isn't a separate app or a person.
• Chase. Chase Bernier at Sidekick Orchestration, for setup and support. Your welcome email says what's included for you.

Not set up yet? Use the setup card for your app near the end of this guide, or wait for your first session with Chase.

This is the guide Sidekick Orchestration clients receive with their setup. If you're setting up on your own, most of it still applies, and the Free AI Starter Kit gets you started.

The short version

Two things you do

1. Ask for what you need, in plain words, inside your project.
2. Say "wrap up" when you finish or stop for now.

Things you can say any time

One rule that never changes: your assistant prepares and you decide. It's instructed never to send anything. You review it, and you send it.

Where this leads: from a first useful result, to handing over bigger pieces of work, to deciding where AI fits in your business based on your own experience.

How it works, in one minute

Picture the smartest new hire you've ever had. They write and research faster than anyone. But every morning they arrive with no memory of yesterday.

A plain AI chat works like that. Your Sidekick folder fixes the memory part. It works like the new hire's notebook: your context, your preferences, your work in progress and the lessons you've asked it to keep. Your assistant reads it at the start and updates it as you work.

"Wrap up" is the end-of-day handover. It brings the notebook up to date, so the next session starts where this one stopped. You don't have to remember where you left off, or explain it again.

The safety rules live in that folder too. They were written once, and your assistant follows them every time: drafts only, use only what you've approved, and ask before anything new. They're instructions it follows, not locks, which is why you still review anything before it goes out.

Your first week

Day one. You set up your project, usually with Chase, and get a first result on your own work, such as what needs you in your inbox and a first reply drafted. When that first result arrives, your assistant explains the "wrap up" habit.

The next few days. Use it on your own work, a little each day. Good moments are first thing in the morning ("What needs me today?"), right after a meeting ("Here are my notes. Draft the follow-up."), and before you reply to anything that matters.

When something's off, say so. "Too long." "That's just FYI." "I'd never say that." It fixes the draft right away. To make a fix stick, start with "from now on."

By the end of the week, most people have one or two jobs they hand over without thinking about it. That's the goal for week one.

Common questions

These are the questions people ask in their first weeks.

Can I use it on my phone?

For quick things, yes. Your notes live in your Sidekick folder on your computer. The ChatGPT phone app can't open it. The Claude phone app can reach it only while the desktop app is open on your computer. The phone is good for quick questions, drafting a reply to an email you paste in, and checking on a chat you started on your computer. Do your main work on the computer, where your notes are. If you did something useful on your phone, paste it into your project later and say "save this."

The connection screen asked for a lot of permissions. Will it send email?

Your safety rules tell it never to send. It drafts, and you send. Connection screens from Microsoft and Google often list more than your assistant will use, sometimes including sending. If the screen lets you leave sending out, do. If it says an administrator must approve, stop there and let Chase know. He'll sort it out with your IT provider, if you have one. Meanwhile you can paste emails in and keep working.

Where are my notes, and what did it keep?

In your Sidekick folder, wherever you saved it on your computer. You never need to open it. Ask: "What have you saved about this?" If something is wrong, say "That's wrong, it's [correct version]" or "Don't keep that."

It seems to have forgotten, or it can't find my notes.

Check that you're in your project. A chat outside it can't see your notes. In ChatGPT, check that Work is on and set to work locally. In Claude, check that you're in the desktop app, inside your project. Then ask: "What did you save from our last session?"

New chat, or keep going in this one?

One chat per piece of work. To continue something, go back to its chat or ask it to pick up where you left off. For something new, start a new chat (in Claude, a new task) in your project. In ChatGPT, once it confirms the work is saved, you can archive the chat to keep your list tidy. Archiving doesn't delete anything.

It doesn't sound like me.

Say exactly what's wrong: "Too formal." "I never open with 'I hope this finds you well.'" Your assistant can also build a voice profile from writing you choose, such as sent emails or examples you paste, with your permission. Pick the ones that sound most like you. Leave out any that AI helped write.

It got a fact wrong.

Correct it and say where the right answer lives: "That's in last month's price list." If the same mistake keeps coming back, say "From now on, check [source] for this." It saves that and uses it on similar work.

I used ChatGPT or Claude before. Can I bring that over?

Yes, when you're ready. Ask: "Help me bring over anything useful from how I used AI before." It shows you what it would keep. Your old chats and settings stay as they are.

Is my information safe?

Your Sidekick folder includes safety rules from day one: drafts only, only the sources you approve, and nothing sensitive connected until you decide. Keep the few things that should never be there out of it, like social insurance numbers, banking details and staff records, unless you've agreed a plan for them. No setup removes every risk, which is why you review before anything goes out. For the full picture, read AI and sensitive information.

Five stages, from first result to handing off bigger work

You don't need to go through these in order or on a schedule. If your plan includes coaching, sessions follow the work on your desk. Use this list to see what's possible next.

The links point to articles and free resources written for anyone. If one includes a setup prompt, ask your assistant before using it in your project.

Stage 1: Get a first useful result

You'll be able to get one useful piece of work done on your own material, and trust it enough to use it.

Try saying: "What needs me in my inbox from the last few days? Draft the most important reply." No inbox connected yet? Paste a few emails in and ask the same thing.

What matters here: a result on your own work teaches more than any tour. Read what comes back, fix one thing, and send it yourself.

Go deeper: Starter Prompts, 20 ready-to-use prompts for email, meetings, decisions and writing.

Stage 2: Make it yours

You'll be able to get drafts that sound like you and follow your preferences without being told each time.

Try saying: "Help me set up my voice profile." Or: "From now on, keep customer emails under 150 words."

What matters here: when you've fixed the same thing twice, make it stick, and give the reason: "From now on, no bullet points in client emails, because they read like a form letter." The reason teaches more than the rule.

Settings: see Settings worth knowing, or ask about one: "Walk me through [setting]."

Go deeper: Make AI Writing Sound Like You.

Stage 3: Make it a habit

You'll be able to use your assistant at the same few moments each day without thinking about it, and pick up any piece of work where you left off.

Try saying: "Wrap up" at the end of each piece of work, and "Pick up where we left off on [topic]" when you come back.

What matters here: attach it to moments you already have, like the ones from your first week. A habit sticks when it rides on something you do anyway.

Go deeper: Turn meeting notes into a reviewed action package.

Stage 4: Hand off more

You'll be able to hand over bigger pieces of work, like a proposal, a customer quote or a board brief, and get back something close to done.

Try saying: "Ask me questions first, one at a time, until you know the goal, who it's for and what done looks like. Then play it back to me."

What matters here: in our work with clients, three habits make the biggest difference to how much of a job AI can take on.

1. Brief it fully. Most weak output comes from a detail the AI had to guess. Let it ask you.
2. Say what done looks like. A few yes-or-no questions the work has to pass: under 800 words, every number matches the spreadsheet, one clear next step. Ask your assistant to check the draft against them before it shows you.
3. Make your fixes stick. Tell it which corrections should apply from now on, so the same mistake doesn't come back.

You set this up once for a type of work, then ask: "Save this as how we do proposals." Your assistant saves it as a routine, so next time one sentence starts the whole thing, checks included.

Deciding what to hand off: some work is a good fit, some needs your judgment, and some shouldn't go to AI at all. The AI Workflow Triage Kit helps you sort it.

Go deeper: Why AI does more of the work for some people and Should AI Do This Work?

Stage 5: Lead with it

You'll be able to decide where AI fits in your organization from your own experience, stay current on what's changing, and bring others along.

Try saying: "Based on what we've done together, where else could this help in my business, and where should we keep it out?"

What matters here: in our experience, leaders who use AI themselves make better calls about it for their teams. Start with your own work, learn what helps, then decide what should spread. If you want to take this to your team, it's a good time to talk to Chase.

Go deeper: How Executives Can Become AI-Proficient in 28 Days, and The Sidekick Skim, a five-minute weekly brief on what's changing in AI for business operators.

How to tell if the work is good

You don't need to check everything line by line. Check the things only you can judge.

Your assistant can check: length, format, facts against a source you named, whether every requested piece is there, and whether it followed your preferences. Ask: "Check this against what we agreed before you show me."

Only you can judge whether it's right for this person, in this tone, right now, and worth sending at all.

Three quick questions before anything goes out:

1. Is every name, date and number right?
2. Would I be comfortable if this person knew AI drafted it?
3. Does it say what I actually think?

If the work keeps needing the same fix, make the fix stick instead of checking harder. For more, read Why your AI keeps getting things wrong.

Settings worth knowing

Menus change often. The easiest route is to ask your assistant about one setting at a time: "Walk me through [setting]." Official help pages are linked in the setup cards below.

Setup card: ChatGPT (Work mode)

If you use ChatGPT. Checked September 2026.

1. Install the ChatGPT desktop app and sign in.
2. Save the zip file attached to your welcome email somewhere easy to find, like Documents, and unzip it. That gives you your Sidekick folder.
3. Under Projects, create a local project and select your unzipped Sidekick folder. Don't upload the files. Keep the project's memory on the default setting.
4. Start a new chat in the project and switch from Chat to Work. Keep it set to work locally, so it can read and update your folder.
5. The first time, send the start message from your welcome email.
6. After your first result, paste the project instructions your assistant gives you into the project's settings. From then on, start a new chat in the project, check it's set to Work, and ask. If you skipped this step, begin new chats with "Read START_HERE.md."

Official help: Get started with Work (includes a short video) · Projects in ChatGPT · Archiving chats

Setup card: Claude (desktop app)

If you use Claude. Checked September 2026.

1. Install the Claude desktop app and sign in. Working with files and connected apps needs a paid plan.
2. Save the zip file attached to your welcome email somewhere easy to find, and unzip it. That gives you your Sidekick folder.
3. Create a project, choose "Use an existing folder," and pick your Sidekick folder.
4. The first time, send the start message from your welcome email.
5. When your assistant gives you project instructions, paste them into the project's instructions. After that, start a new task in the project and ask.

On your phone (Pro or Max plans): use Dispatch in the Claude app to start a task that runs on your computer, answer its questions and check the results. Your computer must be awake, online and running the Claude desktop app.

Official help: Claude Help Center

Getting help

1. Ask your assistant first. Say "Help" and describe what happened. It fixes most everyday problems itself.
2. If it can't, it writes a short note for you to send to Chase. Reply to your welcome email, or use the contact details in it. Your welcome email also explains your sessions, how to book them and how long your support lasts.
3. Access and permission problems with Microsoft or Google: tell Chase first. If you have an IT provider, he'll work with them.
4. Coaching sessions, if your plan includes them, cover what's going well, what isn't, where you're stuck and what you want next. Bring one task you're working on.

Words you might hear