Y Combinator's Requests for Startups · Summer 2026

AI-Native Service Companies

Requested by Gustaf Alströmer · View in YC · Start with an MVP

Build a production-ready AI-native service company that delivers a professional service using AI agents as the primary workforce, with humans in oversight roles only. Core vision: Create a company that provides a high-value professional service — legal document review — where AI agents handle 90%+ of the actual work, dramatically undercutting traditional service providers on price while maintaining or exceeding quality — where AI agents handle 90%+ of the actual work, dramatically undercutting traditional service providers on price while maintaining or exceeding quality. The system must support the full service delivery loop: 1. Receive client briefs and requirements 2. Decompose work into agent-executable tasks 3. Execute tasks autonomously with quality checks 4. Human review triggers for edge cases and low-confidence outputs 5. Deliver results to clients 6. Collect feedback and improve agent performance Core capabilities: Client intake and work management: - Structured intake forms that capture service requirements - Automatic work decomposition into agent task queues - Real-time progress tracking dashboard for clients - SLA monitoring and alerting - Client communication portal with async updates - Revision request handling AI agent execution layer: - Specialized agents for each service domain task type - Confidence scoring on every output - Automatic escalation to human review below threshold - Parallel processing of independent task streams - Quality assurance agents that review other agents' outputs - Audit trail of all agent decisions and reasoning Human oversight interface: - Review queue for low-confidence items - Side-by-side comparison of agent output vs. expected standard - One-click approve/reject/edit workflow - Feedback capture that trains future agent behavior - Workload management across human reviewers Client delivery and billing: - Formatted output delivery matching client specifications - Usage-based billing tied to actual work units delivered - Quality guarantee tracking and SLA reporting - Client satisfaction scoring - Automatic invoicing and payment processing Analytics and improvement: - Agent performance metrics by task type - Human intervention rate trending down over time - Cost per unit of work delivered - Client retention and expansion metrics - Competitive benchmarking vs. traditional service providers Build a working demo covering: 1. Submit a service request and see it decomposed into agent tasks 2. Watch agents process tasks with confidence scores displayed 3. Review and approve a low-confidence item in the human queue 4. Download completed deliverable in client-ready format 5. View the cost breakdown showing AI vs. human labor ratio

Builds a working MVP of an AI-native service company platform. Includes a client intake form, AI agent task queue that processes submitted work, a human review interface for low-confidence outputs, and a client dashboard showing job status and delivered results.