Presentation: Agents, Architecture, & Amnesia: Becoming AI-Native Without Losing Our Minds
Nobody is talking about architectural debt. Tracy Bannon just showed why AI agents without governance become a nightmare.
Why it matters
As teams rush to deploy autonomous agents, the risk isn't capability—it's loss of control and institutional knowledge. A framework for 'Minimum Viable Governance' is becoming table-stakes for AI-native orgs.
The key facts
10 to knowFramework: Minimum Viable Governance (identity, delegation, ADRs)
Core thesis: 'Architectural Amnesia' as risk of reckless AI autonomy speed
Focus: Managing technical debt at machine speed across SDLC
Shift from bots to autonomous agents requires governance rethink
Published by InfoQ (professional dev/architecture audience)
Core tension: bots (supervised) → agents (autonomous) requires new governance models
Framework components: identity management, delegation controls, Architecture Decision Records (ADRs)
Risk framing: 'architectural amnesia' from prioritizing speed over auditability
Application scope: governance applied across full software development lifecycle (SDLC)
Governance model proposed: 'Minimum Viable Governance' (operationalizable, not theoretical)
Go to the source
InfoQ AI/MLinfoq.com
Publisher excerpt: Tracy Bannon shares a cautionary tale of "The Sorcerer’s Apprentice" to illustrate the risks of unbridled AI autonomy. She discusses the shift from bots to autonomous agents, explaining how reckless speed leads to "Architectural Amnesia." She provides a concrete framework for "Minimum Viable…