Zach Bruhnke

Roadblock Title:

Why AI Architecture Has to Get More Boring

Time:

Wednesday -10:30 AM (Tower C)

Abstract:

Most AI agent demos fall apart the moment they encounter production infrastructure. In financial systems, the hard problems aren’t prompting or model selection; they’re retries triggering duplicate actions, stale state corrupting workflows, non-deterministic outputs breaking reconciliation, and distributed failures creating operational risk no LLM benchmark measures.

This session walks through real engineering roadblocks we encountered while integrating AI-driven workflows into financial infrastructure systems. We’ll cover where agentic patterns improved engineering velocity, where they created entirely new failure modes, and the architectural changes required to make these systems observable, recoverable, and safe to operate in production.

Attendees will leave with practical patterns for building AI systems on top of intentionally boring foundations: deterministic workflows, idempotent operations, explicit state management, human review boundaries, and infrastructure designed for failure instead of demos.

Bio:

Zach Bruhnke is the Co-Founder & CEO of MBI, a software company helping banks replace legacy infrastructure with modern, AI-native systems.

Zach has spent his career building financial technology companies from the ground up. He co-founded HMBradley, where he led the development of a full-stack digital banking platform. Recognizing that traditional institutions face structural limitations from outdated core systems and consultant-heavy transformation models, he launched MBI to provide banks with direct access to the software infrastructure needed to compete in the next generation of financial services.