·project

The Agent Suite

A 2023 effort around voice AI agents and business automations for customer communication, lead qualification, appointment booking, and operations.

Problem

Many small and mid-sized businesses still treat customer communication as a manual queue: missed calls, slow follow-up, inconsistent qualification, and brittle handoffs between CRM, scheduling, payment, and operations tools. The Agent Suite lives in the practical end of AI agents: systems that answer, route, book, qualify, and follow up so a business can capture more demand without turning every workflow into another hiring problem.

Solution

The product was positioned around human-sounding voice AI agents and automation workflows for inbound calls, outbound outreach, customer service, appointment setting, and lead qualification. The core promise was not just "chat with an AI," but a reliable communication layer that plugs into the business systems a team already uses while preserving a personal customer experience.

How

  • Collaborators: Garrett Sheehan and Matt Paternostro.
  • Stack: Python, Supabase, Postgres, SQLModel, Pydantic, FastAPI, Stripe, TypeScript, React Native, and Expo.
  • Reference: The Agent Suite About page.

The build combined agent-facing product surfaces with the less visible infrastructure that makes automations useful in production: typed APIs, persistent customer and workflow state, payment plumbing, and mobile-ready interfaces for operators who need to monitor or adjust what the agents are doing.

Results

This was a 2023 effort. The useful read on it now is historical: it was about turning AI agents into revenue, support, and operations capacity for real businesses, not a currently active flagship project.

Lessons

Business automation works best when the agent is not treated as a novelty. The interesting engineering problem is the whole loop: natural conversation, structured state, integrations, escalation paths, billing, observability, and a product interface that makes the system legible to the humans who remain responsible for outcomes.

Archive Signals

The later Claude and Devin posts point back to the same operational thesis: agents become valuable when they sit inside funnels, workflows, products, and review loops.

Neighborhood

Related

From Arxiv Reading to ML Systems TasteFrom Arxiv Reading to M...ComputatrumComputatrumThe APIThe APISoftware Engineering After AgentsSoftware Engineering Af...Block Sparse Attention With Block RetrievalBlock Sparse Attention ...A navigable mindA navigable mindpersonal sitepersonal siteMLN-DashboardMLN-DashboardCopyright CalculatorCopyright CalculatorAI systems engineeringAI systems engineeringFull-stack product engineeringFull-stack product engineer...The Agent Suite