Indiana cybersecurity professional and AI educator Tristen O’Brien has built more than 150 AI agents to help manage work and family life, with about 70 currently in use. The tools handle everything from reminders and household planning to finances, email and his YouTube channel. They have freed up time for family and hobbies, but the growing number of agents has also created an information-management challenge.
O’Brien, 44, has worked in technology for more than 20 years and is pursuing a doctorate in artificial intelligence while running a YouTube channel. He recalls feeling that he could keep up with family, work and school at the same time, but that adding anything else meant something had to give. After years of learning Python with the aim of automating everyday chores, he began experimenting with generative AI following ChatGPT’s launch.
From a burrito-ordering bot to 150 agents
O’Brien says he is not a developer and builds agents by working with AI through what is often called “vibe coding.” He began exploring agents after encountering them at an AI conference two years ago. His first was a simple tool that could order him a Chipotle burrito, but it showed him that AI could carry out specific tasks.
He mainly uses Claude, alongside ChatGPT, Perplexity and Gemini. For designing agents, he follows a process he calls BIP: first, he writes down the idea and requirements in full (“Braindump”); then he has Claude ask questions to clarify anything vague (“Interview”); finally, he turns the result into instructions (“Prompt”). A request that starts at about 15 words can grow into a 500-word prompt after clarification, he says. Claude’s “/goal” command can keep an agent working until it reaches a target, but O’Brien stresses that instructions still need to be precise to avoid unexpected results.
His agents span five research and education tools, four for research and news, one for product development, 18 for finances, 27 for family tasks, 26 for email and communications, 42 for YouTube, and 27 for promotion and outreach. Not all run at once. He currently keeps about 70 active, including 12 for family and household tasks.
Four run daily, covering reminders for birthdays, softball activities, the family dog and school paperwork. Four more run weekly to help with shopping, meal planning, a Sunday family briefing and weekly check-ins. One agent handles monthly home-maintenance reminders, while three others are used as needed for tasks such as travel planning, packing and gift selection.
A family chatbot links calendars and texts
O’Brien and his wife have two children, and the family uses a custom AI chatbot he built with the messaging service Twilio and a Claude plugin. It can send messages to family members, help them vote on weekly dinners and show their calendars side by side to make scheduling easier.
Connecting agents to external services is what makes many of them useful, O’Brien says. Claude can take on tasks across apps when connected to tools such as Gmail, calendars and Twilio. He recommends that users get familiar with how these connectors are authorized and set up.
His AI subscriptions cost about $600 a month, with the services also supporting his YouTube channel and other work. If he used AI only for personal tasks, he estimates he would cut that spending and use a $200-a-month Claude Max subscription instead.
A management structure helps contain the workload
As the number of agents grew, O’Brien sometimes felt overwhelmed by digital updates and found it difficult to follow up on every tool. He responded by organizing the agents like a company: he calls himself the “chief human officer,” with an AI chief of staff coordinating several director-level agents, which in turn oversee AI worker agents. The chief of staff consolidates updates, giving him a single place to check progress.
He has also created an animated game world where the agents appear as characters moving around a village. Their outfits vary by rank, making the roles easier to identify. O’Brien can explore the village with the chief of staff or find an individual agent to check on a task. A button sends all the characters to an in-game meeting hall. The interface is a visual way to track the agents and their responsibilities; it does not mean every tool is running continuously.
O’Brien has also retired some agents. One he once used read a Google spreadsheet of friends’ and relatives’ birthdays, visited a greeting-card service two weeks before each birthday, and mailed a card with simulated handwriting and his address as the postmark address. He later decided that automatically sent cards felt impersonal and stopped using the feature.
Data boundaries and time for offline hobbies
O’Brien says he does not give AI sensitive information such as Social Security numbers. Using these tools still involves granting some access to data, he says, much as people do with Gmail and other services that can access personal information. His son, an artist, is less enthusiastic about AI, and O’Brien acknowledges that there are limits to its use in creative work and related industries.
For now, he says AI has been more helpful than burdensome, and it has not displaced his offline interests. The time saved lets him spend more time 3D printing, flying drones and cycling. His aim in building the agents, he says, was not to maximize their number, but to make room for the things he values while balancing family, work and school.