Solo founders
How to automate your job with AI (and still keep it safe)
Short answer
Turn meetings into notes, notes into action items, and action items into tasks an AI agent does in your browser. You review the result at the end of the day. Use only tools and data practices your employer permits, and turn off model training on your accounts where you can.
If you work at a computer, AI can already do a lot of the tasks in your job. Most bosses do not yet know how good it is. Tomi, founder of SupaKid Studios, says that gives you a window to use AI to get your time back, for health, for a business, or for your dog and kids.
This is the workflow he uses and describes. Read the safety section first.
Before anything: use only what your employer permits
This is the most important step. Your employer owns its data and sets its rules. Before you connect any tool to work email, chat, a CRM or meetings:
- Read your company's AI and data policy, or ask IT or your manager.
- Use only tools they allow, and only for the kinds of information they allow.
- Tell people before you record a meeting, and get any consent your workplace or region requires.
- Do not paste confidential or customer data into a tool that is not approved.
If your company says no, stop. This guide is not a reason to break your rules. If you are the boss or you freelance, you set the rules yourself, but the same care applies to your clients' data.
The daily loop
Tomi uses three kinds of tools: a meeting note taker, an AI browser that can act on websites, and a chat assistant that can use a computer. The names matter less than the roles.
- Capture the meeting. Use a note taker your workplace allows. Tomi uses Wispr Flow's notetaker, or the notes your company already uses. Tell participants when you record.
- Pull the action items. After the meeting, tell your assistant to read the notes and list the to-dos and the things you said you would do.
- Hand off the work it can do. For example, if a task board needs updating, an AI browser such as Aside opens the site and updates it. Other tasks: making documents, editing slides, writing code.
- Step away. Tomi's version: go outside, go to the gym.
- Review at the end of the day. Check the work, test the code, finish or merge it. Plan for a real review, not a glance. The time depends on the job and what the agent actually finished.
Make it write like you
For messages, you can have the agent read your past messages and meeting transcripts to learn how you write and talk. Only do this if your policy allows the agent to read those messages. Then review every draft before it sends. Nothing should go out under your name that you have not read.
Protect your data
Two things to check in the tools you use:
- Training settings. In ChatGPT data controls, you can turn off Improve the model for everyone. In Claude's privacy settings, chats are used for training only if you allow it. Turning training off does not delete saved chats, and feedback you send can still be used for training.
- What the agent can see. Give it access only to the sites and files its job needs. Log in to work tools yourself and keep sensitive accounts out of reach.
How we do it at SupaKid Studios
Tomi built this loop for his own work: a meeting ends, the action items go to an agent, the agent works in the browser, and he reviews the result later. He keeps the review step because the agent can be wrong, and a half-finished task looks finished until someone checks it.
We apply the same idea with clients, always with their approval of anything that goes out. Blockwise, a small accounting firm, was categorizing transactions by hand after letting their bookkeeper go. We showed them how to hand their QuickBooks bank feed to an AI bookkeeping agent. It is the same pattern: one repeat job, one agent, a person checking.
For the bigger picture of what setup fits your day, read which AI setup you need.
What goes wrong
- Skipping permission. The fastest way to lose a job is to feed company data to an unapproved tool.
- No review. Agents do tasks that look done but are not.
- Too many tasks at once. Start with two or three repeat tasks.
- Letting the agent send. Keep a person approving outbound messages.
- Context chaos. If you bounce between many agents all day, read how to stop context switching with AI.
Start small today
Pick one meeting this week. Capture notes with an approved tool, list the action items, and give the agent one task. Review what came back.
If you want help setting this up for your own work or your team, SupaKid Studios offers coaching. A 30-minute call is enough to map your week and pick the first thing to automate.
Questions people ask
Can AI really do my job?
If you work at a computer, AI can already do a lot of the tasks in many jobs: updating boards, drafting documents and slides, and working on code. You still need to review the output.
Is it okay to use AI at work?
Only if your employer permits the tools and the way you use them. Check your company policy before you connect any work account or share work information.
What do I need to start?
A way to capture meeting notes, an AI browser agent that can use your work websites, and a chat assistant that can take on tasks. Use ones your employer allows.
Does turning off training protect my data?
It stops your chats from being used to improve models. It does not delete saved chats, and feedback you send can still be used for training.