In mid-April, Amplitude hosted its second AI Week. We stopped business as usual and dedicated an entire week to improving our AI skills. But unlike last year’s AI Week, we roped in the entire company.

That’s right, this learning week wasn’t just for EPD: marketing, finance, sales, and even leadership joined in on becoming AI-first. Together, we learned new AI tools, questioned existing processes, and tested out new ways to work.

The goals:

• Feel more comfortable and confident with AI. We know AI can help us build a better Amplitude for our customers. We want everyone to come along on that journey.
• One concrete improvement. Everyone needed to find at least one way to level up their workflow and use AI to make their day-to-day work smoother.
• Working apps, agents, and workflows that address wider issues, multiply each other’s efforts, and solve problems that weren’t possible to solve six months ago.

How we achieved those goals was up to us. To support and encourage our open-ended learning, we also had AI guest speakers, in-office events, and unfettered access to AI tools.

So Monday morning, 9 AM, we got to work becoming BFFs with Claude. Here’s how it went and what we built.

Making AI Week a reality: Alignment, support, inspiration

Just like with our first AI Week, our bottoms-up work was made possible by top-down alignment. Leadership had a shared vision of the week’s importance, and they made sure that people actually stopped work as normal. CEO Spenser Skates even built his own AI demo.

Beyond that alignment, this week also took the critical support of our IT team, workplace team, and a core team of organizers. They’re our undisputed MVPs.

• The IT team got marketers into GitHub, helped salespeople open a terminal for the first time, distributed API keys and tool access, and contained the anxiety during Claude outages. In total, they enabled 600+ employees and fielded 40+ official support tickets.
• The workplace team made the week feel special in person. Along with seamless support for the highest occupancy rates in our offices ever, they got us breakfast burritos, boba teas, ramen lunch, happy hour drinks, and Friday morning donuts.
• The organizer team was a cross-functional core that centralized our efforts. They planned out the week, hosted team meetings, and built and maintained an online hub for AI Week info. They even built a demo repository and recording tool.

And for inspiration, we were incredibly lucky to hear from guest speakers about their own AI transformations. Special thanks to Akshay Kothari of Notion, Jeanne DeWitt Grosser of Vercel, Brian Scanlan of Intercom, Ron Gill of Benchling, Sanjay Jeyakumar of 1Sphere AI, David Paffenholz of Juicebox, Claire Vo of ChatPRD, and Sajith Wickramasekara of Benchling.

So what did we build?

After four days of vibe coding, team workshops, and “Thinking…” we shared our projects within our functional orgs, picked the best-of-the-best, and then demoed to the whole company. Here are the top-voted projects among more than 300 demos.

GTM

• The Deal Navigator. This engine pulls together internal data, web information, expert workflows, and even a commission calculator to help reps make prospect-ready, fully customized decks and one-pagers in minutes. Built by Paul Fox, Grace Evans, Nick Pratt, Benjamin Young, Becca Monroe, Christopher Murphy, and Aaron Feldman.
• Signal. A sales intelligence dashboard that consolidates live customer and prospect data from Salesforce, Slack, Granola, LinkedIn, Amplitude, and more to help AEs stay on top of renewals, risks, and upsell opportunities. Built by Stan Lysenko.
• Demos & Value, Reimagined. Five interconnected builds that give every SE what they need for every customer interaction: live call intelligence, automated demo environment generation, a deal-specific playbook engine, a customer-facing POV site tracked in Amplitude, and an agent-to-agent orchestration layer. Built by Ian Coe, Aaron Gottesfeld, Brian Ivanovick, Evandro Pacolla, Jay Michael, Jeff Dubois, Michael Clair, Michael Pakter, and Ted Sfikas.
• The Pit Stop. A customer-facing implementation tracker that gives both our team and the customer a single place to manage the onboarding journey. Yes, it has a racecar progress bar. Built by Kelly Casanova, Gabriel Galarza, Vandana Kumar, Rose Edu, and Kavitha Rajagopal.
• AEr Tower: Next Best Action Engine. This command center pulls prioritized, AI-driven recommendations across prospecting, customer health, and deal execution to take territory planning from a one-time cycle to an always-on motion. Built by Christian Newth, Matt Kahan, Corey Gibbel, Simon Levison, Erin Del Mundo, Anna Marrujo, and Eric Carlson.

EPD

• How do I talk to customers? A custom Amplitude agent (connected to all the tools where you do work) for PMs that can recruit users for customer research, do all the prep notes, and organize everything in Notion. It even puts the customer calls on your calendar. Built by Jacob Newman and Jules Lee.
• Amplitude Data Audit. A data modeling tool that runs against a customer’s org and pulls their data details, auto-detects high-signal data modeling issues, and creates a prioritized list of recommendations to share with customers. Audit smarter, not harder. Built by Himaja Jangle.
• AI Slop Janitors. A fleet of 6 agents built to block new security risks, hunt for hidden critical issues, and auto-fix security code issues at AI speed. Winner of most unhinged demo video with AI Spenser Skates over lofi beats. Built by Lily Chau.
• Design Agent. Like an internal Lovable, this platform instantly turns ideas or screenshots into polished HTML designs or marketing assets that look and feel like Amplitude. Read the Twitter article about it. Built by Will Newton.

Marketing

• Data Monster, Evolved. Our mascot Data Monster embarks on a journey into a new dimension: the 3rd one. Leveraging hand-drawn art and genAI tools across Leonardio.ai, Figma Weave, and Midjourney, this design system generates custom 3D Data Monster images and animations that are on-model and adorable. Built by Kirk Bell.
• Scout. An AI-powered experience scouting tool for anyone who plans events. Type in any city, and it searches the web in real time to surface Michelin-starred restaurants, iconic cultural spaces, trending retail activations, and upcoming events worth building a program around. Built by Chelsea Colby.
• Just Ask: Prompt Library Improvements and Promotion. This refined version of the Amplitude AI Agent Prompt Library (built with Cursor and Sanity MCP) has enhanced search, filtering by industry and use case, and a direct integration to Amplitude. Plus, a comprehensive social campaign built with Figma AI. Built by Colin Craft.

People + Places

• PathFinder. This people operating system synthesizes Workday and other HR platforms to answer employee questions, take actions on your behalf, and loop in a PBP when needed. Understand equity, figure out benefits, and even transfer teams, without getting stuck in a queue. Built by Shariff Mohammad, Julia Lee, Rachel Courtney, Jessica Ng, Wallace Oldham, Brittney Lee, Ryan Orr, and Matt Wallace.
• The Performance Review Concierge. A custom-built skill and Workday MCP that lets employees run their entire performance review cycle in Claude. It mines your work automatically, drafts reviews and feedback in your voice, and submits directly to Workday. Built by Nick Morello, Vin Honrao, Sheena Bal, Miranda Yu, Fatima Moudoud, Jeanette Britton, Trisha Lewis, and Heather Brennan.

G&A

• Deal Desk Quoting Agent. This agent builds pricing proposals for the sales team fast. All it takes is five to six pieces of data, and it creates an itemized, year-over-year quote. The team is excited to keep building this out with deal intelligence layers, customer call data, and directly connect it to Salesforce. Built by Katie Helie and Jessica Adona.
• Overage Anticipation and Alerting. Built on top of our existing Snowflake consumption data set, this tool gives AEs visibility into when customers will hit overages so they can proactively reach out to them with positive, forward-looking upsell conversations. Built by Angie Lau, Jack Perejuan, Kareem Powell, Kyle Helstad, and Mykieu Tran.

People’s Choice Awards: The most-loved demos from the AI Week Hub

• Sensei Sales DataMonster. A real-time AI sales coach that listens to your live calls and surfaces the right battlecard, proof point, or talk track the moment you need it, powered by the Amplitude playbook. Built by Chris Murphy.
• POV Builder. This agent guides AEs through a nine-step framework to generate a formatted executive POV statement, podcast narrative, and customer-facing microsite. Built by Simon Khaligi.
• Pixelmatch. Compare front-end implementation to what your designer friend put inside Figma so you nail your next visual bug bash. Built by Amy Lin.

Measuring success: Calluses, not merit badges

There’s a trend for companies to signal how AI-forward they are by bragging about how much they spend on AI. However, we don’t think tokenmaxxing is the best gauge for actual results. Instead, we looked at the calluses our employees got from the work they put in:

• 300+ demos recorded and uploaded
• 300+ skills built and added to our GitHub skills repository
• 97.5% of all Ampliteers using Claude (even ones on vacation)
• 550+ builders using Lovable
• 1,400+ enthusiastic emoji reactions on our dedicated AI Week Slack channel
• Yes, we did 2x our baseline token spend

Taken together, we can safely say we hit our goals for the week. We’re more comfortable and confident with AI, we built concrete improvements, and we addressed wider issues.

It’s still early to say what the ongoing impact of this week will be. But across the board, we’re excited to lean into the AI future, embrace what’s possible, and carry forward our learning as we build an even better AI Analytics Platform.