strAIght talk: AI tips for Amazonians

Recorded with a live audience during Amazon's Bring Your Parents to Work Day — with her "80 going on 60 year old" dad in the front row — VP Candi Castleberry tells the story of the app she built in one evening to help her sisters care for their mom. She draws the line between AI and agentic AI, shares what she calls "dancing with AI," and lands on the one question that moved her from watching to building. Proof that curiosity is the only prerequisite — at any age.

Guest: Candi Castleberry, VP, Amazon eXperience & Upskilling (AXU)

Alexa for Shopping combines product expertise with personal context to help customers find, compare, and buy what they need. A Senior Manager on the team explains the AI behind the experience and where it's headed next.

Guest: Brandon Knapp, Sr. Manager Product Manager Technical, Alexa for Shopping

What if your documents got VP-level feedback before the actual review? On this episode of strAIght talk, hear how one builder turned a decade of leadership feedback into an AI-powered review tool now used by over a thousand colleagues each week.

Guest: Isabela Huggins, Sr. Manager, Ops & Enablement, Global Logistics

Beth Galetti, SVP of People Experience and Technology, hadn't written code in 25 years. Then she built a live website with AI on a transatlantic flight and challenged every person in her organization to build something in two hours or less. Over 1,000 did.

Guest: Beth Galetti, SVP, People Experience and Technology, Amazon

The Working Backwards AI tool was built to be your toughest thinking partner for product ideation. The tool walks you through five customer questions, generates personas who react to your idea, and brings in domain experts to test your thinking. Over half of users say their final idea was different from what they started with.

Guest: Phil Nachbar, Principal Product Manager, Working Backwards

Telling your team to "experiment with AI" without a structure usually leads to scattered pilots that fizzle out. One leader built a model that loads both ends: a dedicated Tiger team reimagining one high-value workflow while the broader org gets protected time to skill up and share what's working. On this episode, our guest walks through the full framework, how the two sides feed each other, and why getting to a disappointment fast is more valuable than waiting months for a polished demo. His biggest tip: pick one workflow where the pain is obvious, give a single owner two weeks, and ship something working.

Guest: Ishit Vachhrajani, Director of Tech, AI, and Executive in Residence, AWS Marketing

Reviewing a document against a standard used to mean line-by-line cross-referencing — toggling between the policy and the draft, flagging deviations manually. One team built an agent that takes the first pass, handing back an annotated Word doc with comment bubbles on anything that needs attention. On this episode, the builder shares the full setup, the exact prompts, and one trick that pulls standards from any internal page without a connector. His biggest tip: start with the review task that comes up the most — volume drives the biggest savings.

Guest: Sean Callahan, Sr Program Manager, Legal, AWS Infrastructure

What if the goal isn't getting through your tasks faster, but using AI to handle them more efficiently so you can focus on higher-impact work? In this episode of strAIght talk, one Amazonian shares how he rethought his daily work and freed himself to lead differently.

Guest: Daryl Hammett, Director, Delivery & AI Platforms, AWS ProServe

In this episode of strAIght talk, the Amazon Luna team breaks down the four-part "harness" framework they used to turn a simple chatbot into Courtroom Chaos, a live AI-native game with celebrity judges. You'll walk away with a repeatable approach for building AI experiences that feel intentional, not generic. Whether you're prototyping your first AI feature or trying to make an existing one stick, start here.

Guests: Pen Szeto, Head of AI and Sergey Rudenko, AI Technologist - Amazon Luna

"Which AI tool should I actually use?" Leo Ohannesian, a product manager on the Amazon Quick team, comes back to strAIght talk with a candid answer: pick one, go deep, and let it learn you. On this episode, he breaks down how Quick's desktop app builds a knowledge graph from your connected tools, learns your working style through procedural memory, and keeps your context fresh through an activity feed that reads your sources every 15 minutes. His biggest tip: open your feed, customize how it prioritizes your work, then ask Quick, "What's a repetitive behavior I've had that could just run for me?"

Guest: Leo Ohannesian, Principal Product Manager, Amazon Quick

In this episode, an Amazonian shares how he got tired of watching the best offsite conversations happen by accident, so he built an Amazon Quick app that makes them happen on purpose. The MIXX takes pre-work survey responses and delivers a complete meeting prep packet: who should sit together, where the group already agrees, and where to spend your limited time.

Guest: Mark Boccia, Sr. Mgr, Learning Design & Delivery, Global AI Accelerator / Global Ops L&D

We went back through every episode of strAIght talk and pulled together the four most essential AI techniques every Amazonian should know. Each one builds on the last, giving you a complete, repeatable approach to getting better results from AI. Whether you're brand new to AI or looking to sharpen your approach, start here.

In this episode, a VP who leads HR for Amazon's global operations shares his personal AI journey: from basic prompts to a daily framework that changed how he runs meetings, helped him lose 20 pounds, and even helped someone in his community build a business plan. His one rule: stay the human in charge.

Guest: J. Ofori Agboka, VP People Experience & Technology (PXT) - Global Operations, Amazon

In this episode, Morgan shares how her talent development team used AI not just to code a new product, but to fundamentally rethink what onboarding should look like. What started as a debate with AI led to a six-week rebuild that shifted the focus from task completion to customer delivery.

Guest: Morgan Grimes, Sr Manager, Product Management, Amazon eXperience & Upskilling

One team in Global Tax Data used AI to reimagine how they work together — from automating month-end data validation that used to take days, to building an agent that translates code into plain English for compliance reviews. On this episode, members from Amazon’s Global Tax Data team share how "AI Solution Jams" (teach, demo, create) helped them scale AI experimentation across six countries and land three ideas in a finance-wide hackathon. Their biggest tip: it's about the habit, not the heroics. Start with the most repetitive thing in front of you and build from there.

Guests: Chris Jocson, Parmi Makh, and Raksha B from Global Tax Data team.

​Carving out hours for exam prep feels impossible when you're juggling deliverables, launches, and meetings. On this episode, Sery Cybulski shares how she used AI to pass multiple AWS certifications with just 30 minutes a day. Her approach: tell AI exactly what you don't know, let it build a tailored study plan, and when you get questions wrong, feed them back so it adapts to your gaps. No cramming required.

​Guest: Sery Cybulski, Senior Program Manager, AWS Support Engineering

One team at AWS used AI at every step to build Kate Spade’s AI "gift concierge" — a shopping assistant that asks customers what they need and narrows options through conversation. On this episode, David Dorf shares how they went from concept to live product in 12 weeks. His biggest tip: you don't need a perfect idea. Keep a backlog, present a rough one to AI, and let it research and refine until you pick up momentum.

Guest: David Dorf, Global Head of Retail Industry Solutions, AWS

"I built an agent... now what?" Orlando O'Neill built five custom AI agents for his marketing work. On this episode, he shares how he makes them reliable: asking agents to diagnose their own mistakes, fixing issues permanently with steering files, and running weekly reviews to improve them over time. His biggest tip: when your agent gets something wrong, don't scrap it. Ask it why, and fix it together.

Guest: Orlando O'Neill, Senior Product Marketing Manager, Alexa Marketing

Jackie Pizaña wanted AI to do more than summarize documents. Using an internal guided workflow called Cortex Kickstart, she built four AI agents that track commitments, flag escalations, prep coaching materials, and orient her week. Her biggest tip: lean into one tool, learn it well, and let your confidence build from there.

Guest: Jackie Pizana, Senior Manager, Operations Field Talent Acquisition

Eight weeks, no roadmap, and a 50/50 shot. Vedran Brozovic shares how he used AI to find the right data, generate technical requirements, and build an hourly forecasting system in weeks—not months. His biggest tip: ask AI questions like you'd ask a colleague, not a search engine.

Guest: Vedran Brozovic, Product Manager, Automated Profitability Management