VIRTUAL EVENT: December 3, 2026, 12:00 PM EST
Four months with an AI chief of staff, and what it taught me about the future of work
Raviv Levi, CData's Chief Product and Technology Officer handed his inbox, calendar and meetings to an AI chief of staff. Here's what it took to make it useful, and what it got wrong along the way.
In May, Raviv gave a Claude-based agent a one-hour interview about how he works and it said it had everything it needed. Then, the next morning, it asked his name.
The model was never the problem, memory and context were. This session covers four months of building the onboarding plan he never had for his agent, with the mistakes left in to help guide your own build.
What you’ll see
How agent memory lives in files, chat history builds context, and builds memory for the next session.
Using one gateway for every read and write, each one attributed and approval-gated.
How failures become written rules, and the agent runs its own retros
The four moves to start tomorrow
Who this session is for
Data, IT and AI leaders deciding how to give agents governed access to business systems
Executives and managers looking to build a personal AI agent
Hands-on builders who want the mechanics: files, scheduled tasks and approval gates
Register to see how CData is using our own tools and systems. Every read and write the agent does is routed through Connect AI acting as one gateway to multiple Salesforce instances, Jira, Confluence, Teams, Office365 and more.
