Principal Enterprise Architect, Adobe · AI Systems · Bengaluru, India
The hard part of enterprise AI isn't building agents. It's trusting them.
I design the architecture and governance that make agentic AI accountable enough to run a real enterprise — grounded in 25 years of systems that had to survive production.
See the framework →The methodology
The Agent Governance Maturity Model
Agentic AI is easy to demo and brutal to scale. Left ungoverned, every new agent adds sprawl no one can account for — decisions you can't trace, owners you can't name. This model is the way out: five stages from "shadow AI" to auditable, self-evolving governance, measured across six dimensions so you know which rung you're actually on.
Governance on paper
Policies exist; nothing's enforced.
Reactive
You find out after the incident.
Defined
Standards written and followed.
Observable
Behaviour visible in real time.
Evolving
Governance keeps pace with the agents.
Shaped through Architecting the Intelligent Enterprise — a closed-door roundtable with 18 senior enterprise architects. This model came out of practitioners, not a podium.
Assessed across six dimensionsAgent Inventory Completeness · Decision Authority Clarity · Runtime Observability · Policy Enforcement Gap · Incident Response Readiness · Governance Evolution Speed
Measure where you stand →Four fronts of one argument
Governance, industrialisation, the architect's role, and sovereignty aren't four interests — they're four faces of the same problem: making agentic AI accountable enough to run at enterprise scale.
Agent Governance
Making agentic systems accountable, not just capable.
02Industrialising Agentic AI
Moving agents from demo to dependable infrastructure.
03The Evolving Architect Role
What the enterprise architect becomes in an agentic enterprise.
04Sovereign AI
Data localization, sovereignty, and an India-based practitioner's view.
The State of Enterprise Agent Governance — Q4 2026
The first of a quarterly brief. A practitioner's synthesis of where enterprises actually stand on autonomous agents — the three shifts defining the quarter, the failure modes that stall deployments between demo and production, and a deep dive on the blind API hand-off whose fix is the integration patterns architects already trust. Read against OWASP and the NIST AI RMF.