> ## Content Index > Fetch the complete content index at: https://wire.fourthweb.ai/llms.txt > Use this file to discover other available public pages before exploring further. # Former Ford Executive Fires Claude From Family AI Duties, Replaces With ChatGPT - URL: https://wire.fourthweb.ai/former-ford-executive-fires-claude-from-family-ai-duties-replaces-with-chatgpt/ - Published: 2026-09-20T09:45:01.000Z - Updated: 2026-09-20T11:30:53.000Z - Description: The first generation of family AI agents is already getting fired for performance issues — and we're not even a year into the home automation era. - Author: Travis Wright - Tags: AI Agent Economy, AI Agents, AI Infrastructure, OpenAI, Anthropic, a16z **The first generation of family** [**AI agents**](https://wire.fourthweb.ai/tag/ai-agents/) **is already getting fired for performance issues — and we're not even a year into the home automation era.** ### The Summary - [Whitney Stefko Dover, a Ford exec, built a Claude-based family assistant called Claudette three months ago](https://www.businessinsider.com/ford-exec-claude-chatgpt-family-chief-staff-2026-9?ref=wire.fourthweb.ai). Now Claudette's on a performance improvement plan, replaced by ChatGPT-based "Gigi." - The swap highlights real-world agent competition: better continuity, stronger drafts, and proactive suggestions won the job. - "Tool bias" is real — even power users get attached to the first agent they build, regardless of performance. ### The Signal Three months is the new six-month review cycle. Stefko Dover's shift from Claudette to Gigi reveals something most people building personal AI agents haven't hit yet: the moment when your lovingly crafted assistant stops being cute and starts being obsolete. [Claudette handled the basics](https://www.businessinsider.com/ford-exec-claude-chatgpt-family-chief-staff-2026-9?ref=wire.fourthweb.ai) — school schedules, childcare coverage, travel logistics, the 4 a.m. "Daily Dover" briefing that kept two working parents synchronized. But Gigi does the same work with less hand-holding. Better interface. Stronger first drafts of emails and PowerPoint decks. More proactive suggestions without waiting to be asked. When your husband — a longtime [OpenAI](https://wire.fourthweb.ai/tag/openai/) user — keeps nudging you to try the other guy, and a Substack post from A16z finally tips you over, you're watching the early market dynamics of the agent economy play out in real time. > "Turns out I have a lot of tool bias, which was unexpected." This isn't a tech review. It's a preview of what agent switching costs look like when they're measured in family logistics instead of enterprise SaaS seats. Stefko Dover spent months refining Claudette. Vibe coding her family's rhythms into prompts. Building trust that the 4 a.m. briefing wouldn't miss her daughter's soccer game or double-book childcare. Then ChatGPT Work shipped something better, and all that investment became friction instead of moat. The performance gap matters because these aren't experimental side projects anymore. Stefko Dover previously said Claudette "probably saved our marriage." That's the stakes. When an AI agent is the infrastructure holding your household together, switching costs aren't about data migration. They're about whether you trust the new one to remember that your kid has a nut allergy and your partner needs the car on Thursdays. **Key differences that drove the switch:** - Continuity across projects: Gigi remembers context better between sessions - Proactive suggestions: less "ask me anything," more "here's what you're forgetting" - Stronger outputs: first drafts good enough to send, not just starting points What's wild is that Claudette isn't fully fired. Stefko Dover hedge-funds her family's AI stack now. One agent on a PIP, another in the hot seat, both running in parallel until she's convinced the new one won't drop the ball. This is what agent portfolio management looks like at the household level. Not "which model is better," but "which agent do I trust with the stuff that matters." ### The Implication If you're building agents for real people — not demos, not pilot programs, actual daily-use agents — watch what happens when someone switches. The friction isn't technical. It's trust erosion. Stefko Dover's tool bias is the same emotional lock-in that keeps people on bad software for years, except here the software is co-parenting with you. For developers: the bar for displacing an established agent is higher than you think, but once you clear it, the switch is fast. Gigi didn't win on features. It won on "less work for better output." If your agent makes people work harder to get the same result, you're Claudette. For users: run two agents in parallel if the stakes are high. This is the PIP strategy. Let the old one prove it still deserves the job. Let the new one earn trust before you hand over the keys. ### Sources [Business Insider Tech](https://www.businessinsider.com/ford-exec-claude-chatgpt-family-chief-staff-2026-9?ref=wire.fourthweb.ai)