Skip to content

Ryan McMillan — Technology & AI Leader

Practical AI systems
for regulated teams.

I lead technology at Corporate Central Credit Union, putting AI to work for regulated financial institutions — securely, usefully, in production. And I build Delega, the task infrastructure AI agents need to hand off real work.

20+ years in IT · Credit union technology · AI automation · Creator of Delega

Beneath the surface

Agents can do the work.
They need somewhere to hand it off.

delega · event logshared across agents
  • 09:14:02task.created#0142 "summarize exam findings" — by ryan
  • 09:14:07task.claimed#0142 → agent:claude
  • 09:16:31task.delegated#0142.1 → agent:codex · context attached
  • 09:21:54webhook.firedtask.completed → #0142.1
  • 09:22:10task.completed#0142 — queued for human sign-off
That layer is Delega

Focus

The work connecting both worlds

Credit union technology on one side, the agent future on the other — and the same discipline applied to each.

  • 01

    Practical AI for credit unions

    Secure, useful AI adoption inside regulated financial institutions — past the pilot, into production, with governance that holds up to examiners.

  • 02

    Agent infrastructure

    Delega: task handoffs, delegation chains, and work queues that let AI agents do real, accountable work across sessions, tools, and machines.

  • 03

    Technology leadership

    Modernizing systems, improving governance, and turning complexity into tools people actually use.

What I'm building

Delega — accountable coordination for AI agents

I built and operated a hosted coordination layer for agents working across sessions, tools, and machines. The commercial thesis failed. The engineering thesis — that semi-autonomous tools need provenance, bounded authority, recoverable ownership, and explicit human control — got stronger.

  • Coordination

    Atomic leases, recoverable ownership, and delegation lineage across agents.

  • Provenance

    Untrusted ingress stays identifiable through automation-created children.

  • Human control

    A narrow, single-use answer path distinguishes human decisions from agent claims.

  • Bounded systems

    Roles, leases, cascade caps, and secret redaction constrain model authority.

Also building: Finch AI speech-to-text for Windows · PitchTracker youth baseball pitch counts and player safety

Principles

How I work

  • 01

    Boring is a feature.

    The best AI projects aren't flashy demos — they're the automations that make someone's Tuesday afternoon less miserable.

  • 02

    A pilot is 10% of the journey.

    The other 90% is operations, governance, and change management. Budget for production or don't start.

  • 03

    Trust is the product.

    In a regulated institution, security and member trust aren't constraints on the work. They are the work.

  • 04

    Agents do work. People own outcomes.

    Automation should widen human judgment, not route around it. Every handoff ends in accountability.

  • 05

    Build the layer beneath.

    Durable systems beat clever prompts. When the demo era ends, infrastructure is what's left.

Ryan McMillan

Contact

Let's talk.

Open to speaking engagements and conversations about AI in regulated financial institutions — and about what agents need to do real work. The fastest way to reach me is a message on LinkedIn or X.

Milwaukee, Wisconsin · Credit union fintech by day · Baseball coach · Father of four