How I AI: Brian Greenbaum's 3-Step Playbook for Driving Company-Wide AI Adoption

In this episode of How I AI, Pendo product designer Brian Greenbaum explains how a personal Cursor prototype led to a cross-functional AI program with hands-on sessions, an open Slack channel, policy guidance, and adoption surveys.

Brian Greenbaum, a product designer at Pendo, built a practical answer. He created a step-by-step program that got product and design teams actually using AI, not just talking about it. It started while he was on paternity leave, experimenting with an AI coding tool.

Brian covers the message that earned leadership support, the live workshop exercise he used with the team, and the governance work that made approved tools and data-sharing rules easier to understand.

start with a concrete personal result

build something worth showing

While on paternity leave, Brian started experimenting with Cursor, an AI code editor. He had an idea for a music app: scan a QR code on a physical card to play an album on Spotify, basically a modern record player. He's not a developer, but:

"I pulled up Cursor and within a couple hours I had a working prototype. I was creating QR codes, PDFs, doing all this really cool stuff."

At Pendo, Brian works on analytics features where realistic, data-heavy prototypes are hard to communicate in Figma alone. He realized tools like Cursor could help him build high-fidelity prototypes with real interactions and fake data, which made ideas much easier to explain than static mockups.

make a specific case to leadership

Still on leave, Brian drafted a Slack message to his manager, their manager, the CPO, and a handful of AI-curious coworkers. He made a straightforward business case:

1. Internal efficiency: The product team could "leverage AI tools to get more done in fewer hours, improve decision making, and communicate ideas more effectively."
2. External positioning: By becoming proficient in AI, Pendo could better serve customers going through similar transformations.

The CPO immediately asked him to present at the next all-hands. That gave him the buy-in to turn the idea into a formal initiative.

create time and space for practice

combine scheduled sessions with an open channel

Brian identified time as the biggest barrier to adoption. Most people already believed AI mattered. They just didn't have time to experiment. His solution was simple: put dedicated AI time on the calendar and create an ongoing place for people to share what they were learning.

• Bi-weekly "Product AI" sessions: Interactive meetings designed to get people's hands dirty, not just listen to presentations.
• Public Slack channel: A hub for sharing articles, experiments, and questions, what Brian calls "radical many-to-many sharing."

keep the sessions hands-on

For his kickoff, Brian had everyone actually use AI, live. He designed a simple exercise using bolt.new:

1. Same prompt, different results: Everyone pasted the same prompt to create a to-do app.

The surprising part was how different the outputs were. Everyone used the same prompt, but the AI generated wildly different apps. Some failed entirely, which turned into a useful lesson about iteration and debugging.

2. Creative exploration: Brian encouraged people to "go wild" with modifiers like "add a retro 8-bit pixel art theme" or "make it look like MySpace from 2007." People laughed, experimented, and saw the creative potential.

The hands-on format lowered the intimidation factor fast. The Slack channel kept experiments and ideas flowing between sessions. Designers even started using Midjourney to create animated UI characters, the kind of detail that usually gets cut because it takes too much time.

measure adoption and clarify policy

How to Pitch a Company-Wide AI Adoption Initiative to Leadership: Win support for an AI-adoption initiative by pairing a concrete personal proof point with a specific business problem, a bounded experiment, and a clear request for sponsorship.

track sentiment and policy awareness

As part of a company OKR around AI adoption, Brian's group ran a baseline survey focused on:

• Personal sentiment toward AI's impact
• Familiarity with company AI policies
• Awareness of which tools were available

They ran the survey again at the end of the quarter. The biggest gains were around policy awareness and understanding which tools employees could safely use.

create a clear path for approved use

The survey exposed a predictable problem: people were using personal ChatGPT accounts and guessing about what data was safe to share. Brian's team partnered with legal, security, IT, and finance to build an AI Knowledge Center that included:

• An alphabetized table of approved AI tools
• Clear data-sharing guidelines for each (e.g., "Internal Data Only," "No PII")
• Security and legal status
• A process for requesting access or new tool evaluations

That replaced a lot of uncertainty with a clear path. People could experiment openly instead of quietly working around policy gaps.

What changed

Brian later built a prototype MCP server against Pendo's public APIs and demonstrated natural-language generation of interactive product-usage dashboards in Claude.

The demonstration attracted the CTO's attention and helped accelerate internal exploration of agents and MCP-related work. The article does not treat the prototype itself as a production deployment.

The three-part playbook

Brian's approach has three parts:

1. Start with a result you understand well enough to explain and defend.
2. Create recurring hands-on time plus an asynchronous place for questions and examples.
3. Publish approved tools, data rules, access paths, and the measures you will use to judge progress.

How to Create an AI Governance Framework and Tool Library for Your Company: Create one maintained source of truth for approved AI tools, data rules, access, and experiments, using a baseline survey to focus governance on the questions employees actually have.

How to Run an Engaging Hands-On AI Workshop for Your Team: Teach a team to use AI by having everyone build the same small app, compare how the outputs differ, repair failures together, and then push the working result in a personal creative direction.

Measure Baseline Sentiment and Awareness Draft a 7-question anonymous baseline survey covering AI tool use, frequency, confidence, perceived value, policy awareness, uncertainty about data handling, and one blocked use case. Use neutral wording and include "not sure" where appropriate.

Collaborate with Cross-Functional Stakeholders Draft a concise email to invite representatives from our Legal, Security, IT, and Finance departments to a new cross-functional working group. The purpose of this group is to create [company name]'s first official AI governance policy and tool guidelines. The email should explain the goal is to enable safe and effective AI use, state the expected time commitment (e.g., one 60-minute meeting per week for the next month), and ask them to nominate a representative from their team. Keep the tone collaborative and proactive.

Create a Process for New Tool Requests New AI tool request Use case and expected value: [ ] Users and duration: [ ] Data types involved: [ ] Required integrations and permissions: [ ] Vendor and model: [ ] Retention or training settings: [ ] Cost and owner: [ ] Requested decision date: [ ]

The kickoff worked because Brian made participation unavoidable and low stakes. Everyone built from the same bolt.new prompt, compared the different outputs in the room, and learned by changing a working artifact instead of listening to a presentation about AI. That shared exercise gave the later governance and tool-library work a concrete reason to exist.

The program paired enthusiasm with operating structure. Workshops gave people practice, Slack made learning visible, and the knowledge center reduced uncertainty about what employees could use safely.

Create one maintained source of truth for approved AI tools, data rules, access, and experiments, using a baseline survey to focus governance on the questions employees actually have.