Now

What I am focused on this quarter.

• Building the Leverge AI Agent Orchestrator: letting an enterprise assemble and deploy its own agentic workflow from the 306-agent library in a few clicks, with the same named-approver and audit model the shipped agents already run under.
• Keeping Leverge AI honest in production: 306 agents across 13 business functions, and the quality metrics that tell us which ones to retire.
• Getting three enterprise agent deployments from pilot to a signed-off production runbook. The blocking issue in all three is the same: nobody has written down what the agent must refuse to do.
• Writing the first version of the enterprise agentic AI reference architecture as a public playbook, and pressure-testing it against our own delivery pods.
• Publishing the AI-DLC framework properly, with the review and release changes spelled out, instead of explaining it on whiteboards.
• Running the first “State of agentic AI in the enterprise” survey: fieldwork in October and November with technology leaders in India, the US and the Middle East, findings published here and on rightagent.ai in December, with the anonymised data.
• Reading: Atul Gawande’s The Checklist Manifesto, again, because a refusal list is a checklist with teeth; Martin Kleppmann’s Designing Data-Intensive Applications, because most agent failures I see are data failures wearing a model costume; and Donella Meadows’s Thinking in Systems, for the budget ceilings.

This page follows the /now movement. It changes when my focus changes, not on a schedule.