How a CEO and Mom Uses AI Agents to Cut Household Admin Time

Carrol Chang, CEO of Andela and a mother of three in New York City, described the weight of household coordination as crushing, even with a supportive workplace and partner. The tasks included calendar management, sick pickups, school and day care calls, holiday gifts, family travel, groceries, nanny payments, and meal planning.

Her practical takeaway is clear: identify the repetitive, rules-based parts of family life before trying to automate everything. The strongest starting points are tasks with predictable inputs, like emails, schedules, routes, and shopping lists.

Chang built an agent for her kids’ sports schedules, where games change days and locations across different boroughs. The agent maps commute options, adds buffer time, accounts for travel to and from games, and helps manage schedule conflicts between parents.

She said the agent created around 60 calendar invites for soccer games and related travel, including entries such as “Uber to game, 23 minutes.” For parents, the lesson is to make the desired output specific: calendar invites, route notes, timing buffers, and conflict flags.

Chang also created a school agent that reads emails, extracts relevant details, and adds calendar events when needed. Its priority is flagging school closures and ignoring events that do not apply to her children, such as Mandarin-track events when her children are in the Spanish track.

The agent also distinguishes listserv messages from direct emails. If an individual teacher emails Chang and her husband directly, the message is routed through because it needs their attention.

Chang’s family also uses an AI agent for meal planning, taking food allergies and household requirements into account. Instead of relying on a rotating list of 15 dishes every three weeks, the AI suggests new menus and shopping instructions, with shopping apps making the process easier.

Chang said she previously calculated household mental labor at about five to eight hours a week and now spends probably two hours a week doing the work and checking the agents. Her warning is important: AI can make mistakes, so the system should include regular review rather than full blind trust.