# Endstation

> Endstation is an AI automation consultancy based in Cleveland and Detroit that builds production AI systems for operations teams in insurance and healthcare. We decompose complex workflows into specialized micro-agents glued together with deterministic code — a framework we call the Ultimate Macro Machine (UMM). Our thesis: reliability beats intelligence in production AI, and the most durable systems are the ones where the model does as little of the work as possible.

We work with insurance carriers, MGAs, TPAs, and healthcare operators on workflows that involve high-volume document processing, compliance requirements, and human judgment. Engagements follow a Discover → Build → Run model: a paid workflow audit first, fixed-price builds, then ongoing monitoring and expansion. Clients own all code, documentation, and infrastructure we produce.

Endstation's engineering team has shipped production systems at the Federal Reserve, Microsoft (HoloLens, Minecraft), IBM Watson Health, KeyBank, and 2U, plus Department of Defense contract work. HIPAA-trained, with zero-data-retention architecture on Azure and AWS.

## Core framework

- [The Ultimate Macro Machine framework (UMM)](https://endstation.ai/umms): Endstation's architectural approach to production AI — decomposing workflows into specialized micro-agents with typed inputs and outputs, connected by deterministic code. The reference for how we build everything.
- [The AI automation maturity model, Levels 0–4](https://endstation.ai/ai-maturity-model): A workflow-level (not organization-level) maturity framework. Levels 0 (chat-and-paste) through 4 (production agentic workflows), scored by a single diagnostic: *when the AI is wrong, how do you find out?*
- [Why we don't build autonomous agents for production](https://endstation.ai/no-autonomous-agents): Endstation's position on why autonomous agents are architecturally incompatible with regulated operations workflows, with documented production incidents from Amazon Kiro, Replit, Google Antigravity, and others.

## Services and engagements

- [Our services — Discover. Build. Run.](https://endstation.ai/our-services): How Endstation engagements work. Three phases, fixed-price builds, clients own the code. Typical investments: Discovery from $7,500, Workflow Build from $25,000, Focused Pilot from $50,000, Platform Build from $150,000.
- [Homepage / who we serve](https://endstation.ai/): Top-level overview of what Endstation does and who we build for. Anchors on insurance carriers, MGAs, TPAs, and healthcare operations teams.

## Training and workshops

- [Applied AI training for business teams](https://endstation.ai/ai-automation-fundamentals): Endstation's workshop catalog. The flagship 4-hour Applied AI Workshop plus four half-day deep dives (AI Workflow Design, Agentic Engineering, AI Agent Evaluation, AI Security). Includes the VECTOR prompt engineering framework and provisioned access to the AI Workbench.
- [AI Workbench (chat.endstation.ai)](https://chat.endstation.ai): Endstation's LLM exploration environment. Unlike ChatGPT or Claude, the workbench exposes the system prompt, tool definitions, temperature, and model choice so users can see why outputs differ. Used in all workshops; free 5-message trial available.

## Team and credentials

- [Team](https://endstation.ai/team): Endstation's leadership and engineering team. Shipped at the Federal Reserve, Microsoft (HoloLens, Minecraft), IBM Watson Health, KeyBank, 2U, and on Department of Defense contracts.

## About

Endstation LLC is based in Cleveland Heights, Ohio. Founded by Josh Tang. Operating name: Endstation. Contact: info@endstation.ai.

## Optional

- [TAMASSearch — past project](https://endstation.ai/past-projects/tamassearch): Custom search and retrieval engagement.
- [Scooch — past project](https://endstation.ai/past-projects/scooch): Universal parking app.