Skip to main content
BEN STONE

About

Working philosophy

I came to AI through operations, not hype.

I have a background in finance, accounting, operations, and process improvement — and I build agentic systems that also reach beyond finance into commerce, verification, browser agents, and live assistance.

I naturally see organizations as systems: information comes in, decisions are made, work is routed, exceptions occur, controls are applied, and outputs move somewhere else.

AI introduces an entirely new kind of worker into that system. My focus is figuring out how to use it responsibly and practically.

I work across the gap between the business problem and technical implementation. I can map a workflow with the people performing it, design an agentic architecture around it, work directly with repositories and APIs, build and test the implementation, diagnose where it breaks, and iterate until the system becomes useful.

I am particularly comfortable when the process is still messy. Sometimes the hardest automation problem isn't automation at all. It is determining what the process should actually be.

The useful part is the overlap.

I operate in the translation layer between business ambiguity and technical systems — fluent enough in operations to know what actually matters, and hands-on enough in AI engineering to build it.

Business

  • Operations
  • Process improvement
  • Controls
  • Commerce & workflows
  • Finance (domain depth)
  • Reporting

System design

  • Workflow mapping
  • Automation architecture
  • Data flow
  • Failure modes
  • Human controls

AI engineering

  • LLMs
  • Agents
  • Orchestration
  • Tool use
  • APIs
  • RAG / context
  • Verification
  • Open protocols
  • Observability
  • Deployment

Operating Principles

How I decide what to build, and how to build it.

Start with the work, not the model.
The newest model does not matter if the workflow is wrong.
Give agents jobs, not personalities.
Clear responsibilities and contracts are more useful than elaborate personas.
Evidence beats confidence.
A system should be able to show why its answer deserves to be trusted.
Automate decisions carefully.
The more consequential the action, the stronger the verification and approval boundary should be.
Build for whoever comes next.
Documentation, observable behavior, tests, ownership, and failure recovery belong in the product.
Ship the loop.
A small workflow operating end to end is more valuable than a massive architecture that is 80% connected.

Resume

Download

AI Automation portfolio résumé — focused on agentic systems, automation leadership, and production AI operations work.

Download PDF

Technology

What I actually work with.

AI / Agentic

Claude · OpenAI · Agentic workflows · MCP · Tool use · RAG / context systems · Multi-agent orchestration

Engineering

TypeScript · Python · Next.js · APIs · PostgreSQL · GitHub

Infrastructure

Azure · Vercel · Containers · Background jobs · Observability

Business Systems

Accounting platforms · Email / documents · CRM workflows · Operational databases · Reporting systems