Expert Showcase: Building AI Workflows That Stick

Schedule

The time frames are only estimates and may vary according to how the class is progressing.

Introduction – Marily Nika (5 minutes)

• Marily welcomes you to the Expert Showcase.

Fireside Chat (30 minutes)

• Please check back for more information.

What Is an AI Workflow and What Makes One Stick? – Mark Kashef (30 minutes)

• The model is rarely the reason most AI workflows are quietly abandoned within a month. Either the work was never worth redesigning, or the system ended up depending entirely on the person who built it. Mark Kashef, cofounder and CEO of Prompt Advisers, gives you a practical test for spotting work worth redesigning, a clear line between a workflow and a one-off prompt, and a framework for designing something that survives contact with a real team. The examples are derived from his own business on these systems: what he hands to tools like Claude Code and Codex, what he deliberately keeps manual, and why that boundary matters far more than which tool you pick. In a half hour, you’ll be able to look at your own week, name the two or three workflows actually worth building, and say what finished looks like for each one.

Break (5 mins)

How to Build an AI Writing System That Learns How You Work – Katie Parrott (30 minutes)

• Most people write with AI one chat at a time, which means every draft starts from zero. A better alternative is to build a system that learns how you want work done, so you don’t have to explain yourself over and over again. Katie Parrott, staff writer at Every, has spent the past two years building a writing system that works this way. It interviews her before she drafts, challenges her arguments, reviews work against her own standards, and carries lessons from one piece into the next. Katie shares what she’s learned about designing AI workflows that understand not just what you’re trying to produce, but how you want the work to be done. You’ll get a clear model for what an AI writing system can look like, plus a practical starting point for building your own.

Building an Agent Workflow in Gemini Enterprise That You’ll Actually Use – Nicolette Stepakoff (30 minutes)

• Most people try an AI tool, love it for a week, then drift back to old habits. Nicolette Stepakoff, who leads AI adoption for Google Public Sector, has seen this pattern in some of the most change-resistant organizations around, including federal agencies and defense teams. In this session, she shows how to transform a useful AI experiment into a repeatable part of work. In a live demonstration, she’ll build a no-code agent in Gemini Enterprise that connects to your everyday work data and turns a one-off prompt into a reusable routine. You’ll see how to turn a use case into a routine you can use and trust every day.

Break (5 minutes)

Session to Come (30 minutes)

• Please check back for more information.

From Idea to Prototype: How Stripe Built an AI Workflow for Product Teams – Owen Williams (30 minutes)

• AI workflows stick when they fit the systems, standards, and loops people already use. Owen Williams, product design manager at Stripe, shares how he built Protodash, a custom internal prototyping system grounded in Stripe’s design system and product patterns, so teams can create realistic experiences instead of generic mockups. He shows how Protodash grew from a developer tool into a browser-based, v0-like tool that designers and PMs use to explore ideas fast, without a code editor, dramatically accelerating the product-development lifecycle at Stripe.

Building a Personal Product Management Workflow That Gets Better Over Time – Daniel Blum (30 minutes)

• AI productivity gains often fade because each new task starts from scratch. Daniel Blum, senior product manager at Melio, shows how he used Claude Cowork to build a self-improving AI work system for product management that brings together Notion, Slack, and email to support weekly planning, daily context, and focused follow-through. He shares how he designed the system to learn from his edits over time, so it became more useful without becoming another workflow to maintain. You’ll learn practical principles for building a personal AI system that supports better decisions instead of simply generating more output.

Closing Remarks – Marily Nika (5 minutes)

• Marily closes out today’s event.