At Pear, we usually invest before there’s a product—sometimes before there’s even a clear idea. We back exceptional people. But this time, we flipped the script. Instead of asking founders what they’re building, we asked some of the world’s best founders and operators what they wish someone would build. Below are requests for startups from leaders at Databricks, Uber, Oscar Health, Snowflake, Gusto, and DoorDash.
If you’re building something in one of these spaces, we’d love to hear from you. Applications for PearX W27 are now open.
An agent that interviews your company
By Ali Ghodsi
A company’s most valuable knowledge is often stored in employees’ heads and daily routines rather than in searchable databases. It would be great to have an AI agent capture the reasoning behind how the work actually gets done by interviewing employees and translating their insights into permissioned context. This would transform unwritten, tribal knowledge into a durable asset that makes other enterprise agents smarter and acts as a personalized coach for employees.
The AI operating system for work
Every company is racing to deploy AI across functions like finance, HR, legal, customer support, operations, and engineering. Today, most of these efforts are built as isolated workflows, making them difficult to scale, measure, and share across the organization. We want founders building the platform that helps companies discover, deploy, govern, and continuously improve AI-powered workflows. The long-term opportunity isn’t another AI assistant—it’s the operating system that enables every business function to adopt AI quickly, capture best practices, and turn successful workflows into repeatable capabilities across the entire company.
A true personal coach that systematically makes me learn from myself
When I get upset about myself because I didn’t muster up the focus today to finish writing something, or have an idea that’s gnawing on me that I’m not doing anything with, I sometimes remember that I have certain behavioral patterns that have always been with me. But when my kids were born, I started wearing seatbelts in taxi cabs in New York, a simple behavioral change that hopefully made a difference but required a real trigger. So it is possible to change behavior. I want something that tracks what I do automatically, and coaches me in the moment to make me feel good about getting better.
Autonomous governance sandboxes for enterprise AI
Enterprises are racing to enable employees to build and experiment with AI agents for productivity gains. However, security teams are hitting the brakes due to concerns around data leakage, compliance, and “shadow AI” risks. Right now, setting up a safe environment to test an LLM against corporate data is a highly manual, weeks-long IT process involving data masking and rigid provisioning. I would love to see founders start thinking about how to solve this through autonomous security platforms that can instantly spin up secure “shadow sandboxes.” These platforms would connect to enterprise data layers, automatically mirror production data with high-fidelity synthetic equivalents, and continuously audit prompt and response logs in real time. By giving every employee an instant, zero-friction virtual playground with compliance built in by default, this infrastructure would transform security from the biggest barrier to enterprise AI adoption into the foundation that makes it possible.
Lovable for kids
By Tomer London
The next generation should learn to build with AI before they learn to code. Today’s AI app builders make it possible for anyone to create software, but they’re designed for adults. Kids need a fundamentally different experience, one that’s engaging, educational, and safe.
I am excited to meet a founder who is building an AI-native platform that enables children to create apps, games, websites, and interactive experiences simply by describing what they want to build. Along the way, the platform should teach the underlying technical concepts—from logic and design to coding fundamentals—so kids gradually understand how software works rather than treating AI as a black box. Strong parental controls, age-appropriate guardrails, and tools for parents and educators should be built in from day one. The opportunity isn’t just teaching kids to code—it’s empowering an entire generation to become creators with AI.
Enterprise AI evaluation infrastructure
By Andy Fang
I’d love to see startups that make it easy for enterprises to define benchmarks for critical workflows and automate evaluation—while giving companies full control over the models and evaluation harnesses they use.
Another area I’m excited about: open-weight American AI models for a wide range of enterprise and consumer use cases.