Relay: A Delivery Engine Concept for AI-Native Service Companies
A design-led MVP concept for service companies that sell finished outcomes, not seats. Relay scopes the result, lets agents run the workflow, routes work through human review gates, ships the deliverable, and shows the margin on every run.
Run your service like software.
The Challenge
Service companies increasingly want AI to do production work, but the operating model is still stitched together from tickets, spreadsheets, chats, and manual status meetings. That makes every outcome hard to scope, hard to price, and hard to audit. Agents can draft, reconcile, research, and prepare deliverables, but without gates, evidence, connector health, and margin visibility, teams either slow everything down with manual checks or ship work they cannot confidently explain. The concept set out to solve one thing: how do you run a service business as repeatable software while keeping humans in control of the moments that matter?
Our Solution
We designed and built Relay as a full product story across a landing page and operator dashboard. The landing page explains the operating loop: scope the outcome, run the agent workflow, review at the gates, ship the work, and price each run by delivered value. The dashboard turns that story into an operating surface. It includes a workflow library with prices and target margins, a delivery board for every active outcome, run timelines with evidence attached, connector health across tools like Slack, Notion, Linear, Stripe, QuickBooks, GitHub, and Salesforce, plus a margin cockpit that shows where profit is created or lost. Every run moves through explicit scope, production, review, and delivery stages, so agents can do the repeatable work while operators approve the high-risk transitions.
