🦆 Duckbill: AI Life Admin

The best consumer founders know the game. 30K+ of them read Consumer Startups every week.

Stay ahead. Get the playbook behind today's breakout startups.

Read time: 3 mins 42 seconds

Meghan Joyce, a rising executive at Uber, stared at her personal to-do list. 

It wasn't the high-stakes work projects that felt crushing. It was the endless, draining tasks of life admin: disputing a return, scheduling a doctor's appointment, finding a new guitar teacher for her kid.

Her work life ran like a Swiss watch, but her home life, as she'd later describe it, was a "goat rodeo." 

She saw a massive, unserved need born from a societal shift, but the technology to solve it affordably didn't seem to exist. 

Should she accept this friction as the price of a demanding career, or risk everything to build the solution herself?

She chose the latter and built Duckbill, the AI executive assistant for your personal life.

Today, Duckbill has raised $33M and recorded tens of millions of real world interactions across thousands of users. Meghan has an ambition to completely democratize human to human services.

For decades, American households ran on invisible labor. 

A stay-at-home parent managed the vendors, booked the appointments, and disputed the returns. It was a 40-60 hour per week job. Someone was doing it.

Then, the workforce changed. Most households shifted to either single working adults or dual working couples. The life admin got compressed into the evenings and weekends of already burnt-out professionals.

Meghan saw this everywhere. The conversation was always the same: "How are you doing this? Because I can't figure out how to do this big job and also make life work."

She recognized the opportunity at Uber, but the technology didn’t exist to solve it affordably. 

The personal assistant market had been tried before. Expensive human-only services worked beautifully if you found the right person, but they didn't scale. Tech-first solutions were cheap but lacked the quality and reliability you needed for something as personal as your life admin.

She needed a catalyst.

While at Oscar Health, Meghan witnessed something that changed her thinking. She saw early-stage GenAI handling some of the most sensitive human interactions imaginable, for new parents, cancer patients, people in emotionally vulnerable moments dealing with medical and billing complexity.

It wasn’t perfect BUT it was working. 

The insight wasn't that AI could replace humans. It was that AI could finally make a hybrid model work, where humans and AI complement each other at scale.

Meghan resisted the urge to overengineer the MVP.

Instead of building a “powerful AI” from day one, she focused on simulating the experience of one. She hired a seasoned executive assistant and put her in the loop with just enough lightweight tech infrastructure to make the system feel intelligent and responsive.

Behind the scenes, it was human labor doing most of the work. They were explicit about that. Users paid at cost for real work done by a real person.

This approach served two purposes:

The early signals were strong! 

Many people were using the app twice a week. There were many recurring tasks like monthly haircuts, annual PCP appointments, etc. They kept paying.

This strong engagement and willingness to pay gave Meghan the conviction to leave Oscar and go all-in.

Meghan knew she needed to move that scale from 90% human labor completing the task to 90% tech driven, with a 10% human safety net.

She spent the next three years solving the problem first with humans, rather than rushing the AI launch. Every task that came through Duckbill was completed by a human co-pilot. They record every interaction, workflow, and edge case.

By the time they launched AI features in October 2024, they had years of real-world task data, which allowed them to train models that actually worked.

While most AI companies are trying to replace humans with AI, Duckbill is orchestrating humans and AI to do something neither could do alone.

The problem with pure AI tools is that AI is great at the first 80-90% of a task like research and planning, but the last 10-20% is where AI falls short. You need a lot more nuance, persistence, and judgement to solve the last mile well. 

Duckbill’s team built what they call the AI Orchestrator, a system that sits between customers and "doers" (their human co-pilots) and intelligently routes tasks.

Here's how it works:

Triage: The orchestrator analyzes incoming tasks and determines the optimal path. Which tasks can be handled by AI alone? Which requires human involvement? Which needs a hybrid approach?

Deployment: It routes tasks to the right resource based on complexity, urgency, and parameters. AI handles the predictable, scalable parts (gathering information, filling out forms, initial outreach). Humans handle the unpredictable parts (negotiation, judgment calls, edge cases).

Quality Control: It monitors task completion and ensures quality standards are met before delivery. Every task is verified before it reaches the customer.

Think of your complex tasks like an F1 car where an army of pit crew professionals use world-class tech and know-how to get you on the road smoothly and quickly.

The orchestrator is the pit crew chief. Each team member (AI agent, human specialist) knows their role and executes.

Users come back 2x per week because they know that tasks will get done. If something goes wrong, there's a human who cares about getting it right.

Over time, as AI improves, the ratio will shift more toward AI, but the human safety net will remain to provide that personal touch and address edge cases.

Thanks to the recent model improvement, they are now close to the 90/10 AI-human split.

To date, Duckbill has raised over $33M from investors like Forerunner Ventures. The platform has supported tens of millions of personal interactions and is used by thousands of people who rely on it to manage their personal lives.

There are still many challenges ahead. The obvious one is competition. Big AI labs like OpenAI and Anthropic, along with a wave of AI agents like Manus, are eyeing this space and want to own the AI personal assistance category. Years of real-world task data and the hybrid human + AI model are key differentiators, but the question is for how long?

The harder problem is maintaining trust and quality at scale. When you ask someone to manage your personal life, the bar is extremely high. One bad experience can break the relationship. Because Duckbill includes humans in the loop, users would likely expect better outcomes than they do from ChatGPT or a generic AI agent. Meeting that expectation consistently will require operational excellence beyond just better models.

However, what I love about Duckbill is that it’s fundamentally a system of orchestrating human-to-human services and getting real-world tasks done, with AI acting as leverage rather than a replacement. 

As long as the world still runs on human labor and coordination, there is a large and durable opportunity here. I am excited to see how this space plays out. 

—

Duckbill is a good reminder that great products are really just great systems.

Meghan didn’t try to replace humans. She built an operating system around real work and scaled it carefully.

That’s how I run my startup Pikes too. We use Notion as our startup OS. Roadmap, CRM, docs, SOPs, all in one place. It keeps things sane when everything else is moving fast.

If you’re building a startup, Notion offers 6 months of their Business Plan completely free, including their powerful AI tools through the Notion for Startups program.

It's a total no-brainer for any serious founder.

See you next Tuesday,

Leo

Follow me on X and LinkedIn

📥️ Want to advertise in Consumer Startups? Learn more.