Dave Killeen will tell you, on record, that his relationship with AI has crossed into unhealthy territory. He also says he has never felt more alive, that he is 50 and feels 30, and that he sometimes waits for his two small children to go to bed so he can start a coding session.

In between, he gives Axel Sooriah the most practical answer on this show yet to the question every product team is facing: your exec says "we need AI," and you have no idea what they actually want.

Dave is Field Chief Product Officer at Pendo, with 25 years in product across Mail Online, BBC iPlayer, Badoo, Bumble, and Productboard. He was made redundant, spent seven months unemployed going down AI rabbit holes, got dropped by a search firm for terrifying founders in interviews, and was eventually hired off the back of a YouTube rant about speech-to-text. He now builds Dex, an open-source personal AI operating system, on the side.

What you'll learn:

• Why the answer to top-down AI pressure is a KPI driver tree and a career ladder, not an AI strategy

• His argument for why the CPO seat turns over faster than any other in the C-suite, and what causes it

• Whether delegating thinking to AI erodes your judgment, and his case that it does the opposite

• The simplest possible starting point: one chat you never close, voice notes, and a Friday whinge

• How to go from that to running your own agent in about ten minutes

• What happens when your AI reads your color profile and starts telling you off on a Monday morning

• Why he open-sourced Dex, and the frontier-model pricing change that nearly killed it

• Why he is now shrinking a model onto the desktop and leaning into locally hosted open source

• The two mistakes he'd undo from 25 years: not aligning his exec, and over-protecting his team

Chapters:

00:00: Intro
00:57: Welcome to Product in Practice
02:09: What a Field Chief Product Officer actually does
03:22: From Mail Online to Bumble, then made redundant
05:42: The pressure product teams are under right now
07:41: Why your exec doesn't need AI, they need a driver tree
10:45: Does delegating thinking to AI erode judgment?
11:41: Is AI creative? Meet Dexter
14:38: Making intentional time to go down the rabbit hole
16:45: Start here: one chat, voice notes, and a Friday whinge
19:33: Why this whole thing feels emotional
20:53: Where the curiosity came from
22:43: Explain it to me like I'm five, and the rabbit hole problem
25:56: Crossing into unhealthy territory
28:01: What he would do differently
30:25: What Dex is, and why he open-sourced it
33:45: PMs have always been closeted builders
37:24: The pricing change that nearly killed Dex
38:53: Going local: open-source models on the desktop
40:22: One piece of advice: find the time
41:43: Building without analytics, and what changed
43:26: Where Pendo and Atlassian fit together
45:42: His biggest mistake in 25 years
48:34: Treasure chest: one book, one dish
52:08: Where to find Dave

Connect with Dave: LinkedIn: https://www.linkedin.com/in/davekilleen/

Connect with Axel: LinkedIn: https://www.linkedin.com/in/axelsooriah/

About Product in Practice:

Product in Practice is Atlassian's product podcast, where we dig into how AI is changing the craft of product management. Product leaders show you how they actually work: the tools, the frameworks, and the decisions behind the job. Hosted by Axel Sooriah.

Subscribe so you don't miss the next one.

#productmanagement #PM #atlassian

Rachel Wolan runs product, design, and data at Webflow, and she is trying to delete every status meeting on her calendar by the end of the year. In this episode she shows Axel Sooriah exactly how: the builder days that took her org from roughly 15% to 95% building with AI, the one-line field that replaced a recurring roadmap meeting, and a live walkthrough of the system she built in Claude Code on top of Jira Product Discovery to triage feedback, generate mini-specs, and tell her what changed without anyone writing an update.

This is a hands-on episode. The second half is screen share.

What you'll learn:

• Why AI fluency does not come from training on how models work, and what builder days do instead

• How Rachel took her design team from roughly 15% tool usage to sustained adoption after a single builder day

• The rule that changed how her team preps: no meeting with her without a prototype

• Why she believes PMs should own evals, and what happens to the ones who don't live at the model layer

• The difference between developing something and engineering it, and why a designer and PM should be able to ship to 20 users alone

• How she replaced a recurring status meeting with a five-minute field update in Jira Product Discovery

• Her live build: auto-triaging sales and partner feedback, pulling the source Gong call, and generating a mini-spec with three candidate solutions

• The agent she trained on her own product review transcripts so her team gets her questions when she isn't in the room

• Why she runs 5 to 10 Claude Code agents at once, and how she thinks about agent permissions as the real safety question

Chapters:

00:00: Intro

01:07: Welcome to Product in Practice

02:49: From "I use AI" to a team that ships AI products

06:59: The biggest workflow shift: fewer meetings

09:43: How design, product, and data are blurring

11:52: Building in orders of magnitude

13:00: Why scale surfaces corner cases, and what adaptive UX changes

14:28: How the PM role changes: own the evals

17:30: A model swap that moved conversion overnight

18:20: Golden evals and knowing when to tear up the roadmap 19:20: Why reading about AI won't get you there

21:06: The personal apps she built to learn faster

24:40: Hands-on: Jira Product Discovery as the system of record 29:10: Rebuilding triage across sales, partners, and Gong

32:29: Autonomous prototypes and the mini-spec

33:32: Pulling PRDs, tech specs, and launch readiness into one view

37:07: Where this stops, and the real cost of coordination

39:47: How it's built: Claude Code and 5 to 10 agents at once 41:06: Where to draw the line on agent permissions

42:54: One piece of advice: always be building

Connect with Rachel:

LinkedIn: https://www.linkedin.com/in/rachelwolan/ X: https://x.com/rachelwolan Website: https://www.rachelwolan.com/

Connect with Axel:

LinkedIn: https://www.linkedin.com/in/axelsooriah/ Website: https://linktr.ee/axelsooriah

About Rachel Wolan:

Rachel Wolan is Chief Product Officer at Webflow, where she leads product, design, and data.She has more than twenty years building product, engineering, design, and data teams. Before Webflow she was VP and GM of Dropbox's core business, serving over 700 million users worldwide, and led Dropbox's AI strategy including the acquisition of Command E. Earlier roles include LiveRamp, Euclid, and Talkdesk. She describes herself as a builder first and an executive second.

About Product in Practice:

Product in Practice is Atlassian's product podcast. Product leaders walk through how they actually work: the frameworks, tools, and decisions behind the craft, with a live demo in every episode. Hosted by Axel Sooriah.

Subscribe so you don't miss the next one.

#productmanagement #PM #atlassian

Crystal Widjaja, former Chief Product Officer at Kumu and a Reforge instructor, has built data and growth systems at huge scale. Most product managers still use AI for quick summaries and one-shot artifacts. Crystal has spent her "semi-retirement" building something more ambitious: a system of AI agents that handles everything from her daily prep to a "friendship CRM."

In this episode, she walks through the messy reality of her stack, built with Obsidian, Python, and Claude Code, and shows how PMs can get past the trap of polished GUI tools and start building infrastructure that scales their impact.

What we discuss

• Moving from surface-level prompting to mature agent infrastructure and "context engineering"

• Automating low-leverage work like meeting notes and status updates, so you can spend time on strategy and customers

• The AI advantage in Southeast Asia, where a culture of delegation has prepared leaders for working with agents

• Three things that won't change: vision setting, creative voice, and deep customer empathy

• The "Ralph Loop": chaining complex AI tasks together for autonomous execution

Key takeaways

PMs need to get comfortable with the CLI. Crystal argues that sticking to simplified AI interfaces stops you from understanding the system underneath. To improve your workflow and use AI fully, you have to be willing to open the terminal and work with the technical scaffolding yourself.

Context is the new prompt engineering. The first wave of AI was about learning how to prompt. The next wave is about context engineering. Output quality depends on what you feed the system, from meeting transcripts to project files, and on building pipelines that sync that context automatically.

Start with the smallest unit of work. Instead of aiming for a big-bang AI solution, automate one specific task you hate, like syncing meeting notes to your computer. Once that works, iterate and chain those units into a fuller agent system.

AI should make you more human, not just more productive. Crystal uses her AI chief of staff to run a "friendship CRM," tracking things like a friend's offhand mention of a marathon or a gift idea. Offloading the menial memory work frees up energy to be a more present, thoughtful friend.

Adopt a "semi-unemployed" mindset. The best way to stay relevant is to leave enough room in your schedule to play with new tools. Automating low-leverage tasks buys you the time to stay ahead and learn how agent orchestration is changing.

Chapters

• 00:00 - Introduction to Crystal Widjaja

• 02:08 - AI adoption and practice in Southeast Asia

• 04:24 - Demystifying AI through practical workflows

• 07:41 - AI maturity: engineering vs. product management

• 11:11 - Practical steps to build AI fluency

• 14:34 - Critical skills for the AI era

• 18:48 - From prompt engineering to context engineering

• 22:13 - Timeless product management fundamentals

• 23:59 - Demo: building an AI "chief of staff"

• 27:59 - Automating work with "Ralph Loops"

• 36:38 - Impact on personal productivity and relationships

• 39:46 - Hype, reality, and the future of AI in 2029

• 41:51 - Final habits for effective AI use

About Atlassian

Behind every great human achievement, there is a team. From medicine and space travel to disaster response and pizza deliveries, we help teams all over the planet advance humanity through the power of software. Our mission is to help unleash the potential of every team.