LLenny's Podcast
← All episodes
Oji and Ezinne Udezue07 September 2025

How AI is reshaping the product role

7Frameworks
15Insights

Frameworks in this episode

Innovation4 steps

AI at the Core vs AI at the Edge

Decide whether to sprinkle AI onto existing code or rebuild the workflow with the LLM as the core

Read →
Strategy3 steps

Ethnographic Insight Over Stated Preference

Observe what customers do, not what they say — AI transcript-reading is not true customer insight

Read →
Communication3 steps

Opinionated Simplicity

Ship one compelling opinion instead of ten confusing options — you learn faster being wrong than being neutral

Read →
Self-Mastery3 steps

Passion-Project Learning (An Eval for Your Life)

Learn new tech by picking one passion project that forces you to touch everything you need to learn

Read →
Leadership3 steps

Strategy Activation via Crossing the Chasm

Treat internal strategy rollout as an adoption curve — over-communicate the why and manage early adopters vs laggards

Read →
Entrepreneurship3 steps

The Sharp Problem

Build for old, enduring needs re-imagined with new tech so the improvement is 3-10x undeniable

Read →
Leadership3 steps

The Shipyard Team

Organize product work as a six-capability pod of controlled chaos with sensory 'tendrils' to customers

Read →

Insights & moments

The myth-busts, hot takes, explainers, and tools worth keeping.

Myth Buster· 1

Myth Buster57:30

AI Won't Save Customer Research — Watch What People Do

Ezinne cautions that using AI to read all your customer interview transcripts is not real customer insight. What people say differs from what they do; ethnographic research — observing people in their real context — remains the gold standard. People don't always say what they mean, and sometimes lie to themselves about their reasons, so AI-summarized interviews give you junk.

  • What customers do matters more than what they say
  • AI reading all your interview transcripts is not true customer insight
  • Ethnographic research — observing people in context — is still the best method
  • Instrument to see how people actually finish tasks
  • People often don't say what they mean or realize they're lying about their reasons

“That isn't what it is. It's actually trying to figure out how to best walk in their shoes.”

Ezinne Udezue · 58:30

“people think AI has come to save it all, but no, it's going to give you junk because people don't always say what they mean”

Ezinne Udezue · 59:00
#customer research#ai#ethnography#product management

Hot Take· 4

Hot Take08:30

Don't Be the Bottleneck — Add Value Elsewhere

With build cycles accelerating, PMs are increasingly cast as the bottleneck. Aji's answer: don't crouch or worry about the label — adapt with new tools and skills, move faster on defining sharp problems, and since PMs see everything, add value in the go-to-market and other places they traditionally neglect.

  • The 20-year fixed ratio of PM/dev/GTM time is exploding as builds accelerate
  • One company goes from pitch call to prototype in four hours
  • Static PRD writing is now insufficient; cycle time is too fast
  • PMs should adapt with new tool sets and skill sets, not defend old process
  • Because PMs see everything, they should add value in under-served areas like go-to-market

“the thing that we want PMs to do is not be the crouch, not worry about people saying, "Oh, you're the bottleneck." No, don't be…”

Oji Udezue · 10:00
#product management#ai#velocity#go-to-market
Hot Take35:30

The One-Man SaaS Goes Away; 100 Million Developers Arrive

Aji's prediction: within a few years, high-agency people (not just PMs) will write the software that automates their own lives, so the one-man SaaS goes away and there will be 100 million developers. His advice to PMs is to be the vanguard — automate for yourself first, then you can build software that reaches a million people.

  • High-agency people will write software to automate their own lives
  • The 'one-man SaaS' business erodes as people build bespoke personal tools
  • There will be ~100 million developers as anyone who cares can build
  • PMs should be the vanguard: automate for yourself, then scale to a million users
  • It's still day one — nobody who seems ahead has actually won yet

“a lot of high agency people not just PMs will essentially be writing the software that automates their lives right and so the one man…”

Oji Udezue · 35:30

“there'll be 100 million developers because we will people who really care will automate their lives”

Oji Udezue · 36:00
#ai#predictions#personal software#product management
Hot Take45:00

Chat Is Not the Final Boss of AI User Experience

Aji pushes back on chat as the endpoint for AI interfaces, using the GUI-versus-command-line analogy: if command lines were enough, DOS would be the biggest OS. Winners are taking chances on new, dynamic, personalized user experiences instead of defaulting to a chat box.

  • The chat interface is not the final form of AI UX
  • Analogy: GUIs won for a reason; if command lines sufficed, DOS would dominate
  • Companies succeeding are taking chances on novel user experiences
  • Dynamic UX that personalizes to the customer is a big opportunity

“We don't believe the chat interface is the final boss for AI user experience. We think that there's a reason viewers exist and the command…”

Oji Udezue · 45:00
#ai#ux design#product
Hot Take45:30

AI Is 'Ordinance-Level' — Build It With Ethics

The guests warn that AI's capabilities are 'ordinance-level' — eventually more powerful than a fusion bomb — yet many software people build with it with no consideration of ethics. Aji recalls the inventor of deepfakes shrugging off the implications of her work, and calls on PMs to lead on responsibility given the power they wield over developers.

  • AI's core capabilities are 'ordinance-level,' eventually more powerful than a fusion bomb
  • Many builders play with it with no consideration of ethics
  • Anecdote: the inventor of deepfakes said 'I don't care' about the implications
  • PMs direct many developers, so that power comes with responsibility
  • Best companies should lead on ethics rather than ignore it

“we think of AI and its core capabilities as ordinance level meaning like at at some point we'll think of this as”

Oji Udezue · 45:30

“the person who invented deep fakes was interviewed one time and she said and they're like what are the implications of your stuff? And she's…”

Oji Udezue · 1:00:30
#ethics#ai#responsibility#product management

Explainer· 1

Explainer05:00

What's Changing (and What Isn't) in the PM Role

Ezinne argues the PM's core value hasn't changed — derisking delivery and maximizing business value — but AI frees PMs to invest more in true customer insight. The job shifts toward orchestrating software, feedback loops and LLM touchpoints rather than orchestrating people, treating the product as living, breathing software.

  • Core PM value is unchanged: derisk delivery, maximize value from investments
  • AI frees PMs to develop deeper, confirmed insights about the customer
  • Orchestration moves from people-to-people to managing software, feedback loops and LLM touchpoints
  • Static software is dead; treat products as living systems that need to keep learning
  • New emphasis on data literacy and building the right guardrails

“it really is uh derisking the product delivery process while also really trying to maximize the value the business gets from these investments.”

Ezinne Udezue · 05:00

“So that idea of working static software isn't true anymore. It's almost like living breathing software”

Ezinne Udezue · 06:00
#product management#ai#customer insight#data literacy

Story· 1

Story31:00

Aji's AI Smart House and the DIY 'Super Sensor'

As a learning project, Aji is automating his house so it has eyes and ears. He's building the hardware himself around a custom 'super sensor' that sees, hears, and senses humidity and temperature. The house adapts to occupants — precooling kids' rooms before they return, cutting HVAC to empty rooms — running on an on-device AI chip while looking deliberately dumb.

  • Give the house eyes and ears so it senses and adapts to occupants
  • DIY 'super sensor' sees people, feels heat, hears, senses humidity and temperature
  • House manages energy: precools rooms an hour before people return, cuts HVAC to empty rooms
  • Runs on a processor with an AI chip and local inference, kept intentionally invisible
  • A concrete personal project is a powerful way to force yourself to learn the tech

“The coolest thing is I've specked out what I call like a super sensor. It will see people. It will feel the heat. It will…”

Oji Udezue · 31:30
#ai#hardware#home automation#learning

Q&A· 1

Q&A17:30

What Ezinne Hires For: Curiosity, Humility, Agency, Evals

Asked what she looks for in PMs now, Ezinne names a blend of attitudes and skills: curiosity and humility (willing to be a learner and taught by juniors), high agency and ownership (a thermostat, not a thermometer), plus hard skills like data literacy, writing evals, constraining hallucination, and combining multiple models.

  • Curiosity plus humility: willing to admit you don't know and start as a learner
  • High agency and ownership: act like an owner, be a thermostat not a thermometer
  • Data literacy: understand how data is organized and leveraged
  • Write evals to verify LLM output and catch hallucination, beyond just prompting
  • Use multiple models and combine their strengths into something greater

“You're more of a thermostat than a thermometer.”

Ezinne Udezue · 19:00

“there's this skill of being able to write evals”

Ezinne Udezue · 20:00
#hiring#product management#evals#agency

Tool· 2

Tool33:00

Home Assistant, Local Models, LM Studio and Ollama

Aji details the stack behind his smart house and hobby learning: the open-source Home Assistant platform (extensible enough to plug in fine-tuned or large LLMs plus small Whisper voice models), and local open-source models run via LM Studio and Ollama, which he loves for being deployable in different places.

  • Home Assistant is an open-source, extensible home automation platform
  • You can plug in fine-tuned LLMs, big LLMs, and small Whisper voice models
  • Voice wake-word ('hey Jarvis') lets the house talk back
  • LM Studio and Ollama for running local open-source models
  • Local models can be placed in different locations, useful for hobby and privacy

“There's this open- source software called Home Assistant that I use”

Oji Udezue · 33:00

“I'm super into local models the these open source models running them because I feel like I can put them in different places.”

Oji Udezue · 1:11:30
#tools#local models#home assistant#ai
Tool1:09:30

The AI Hack: Be the Generator, Let It Refine

Aji names Claude his everyday sidekick, but shares the core trick for using generative AI well: despite the 'generative' label, you should be the generator and let the model refine your work rather than having it generate from scratch. Generating first yourself, then letting AI refine, is the hack he thinks everyone should learn.

  • Claude is used as an everyday sidekick for anything and everything
  • The trick: be the generator yourself, let the AI refine
  • Don't let generative AI do the initial generation — you do it, it improves it
  • This inverts the default 'ask it to write everything' pattern

“be the generator and let it refine. That's the major trick.”

Oji Udezue · 1:10:00
#tools#ai#claude#workflow

Takeaway· 5

Takeaway27:30

Humility Is Teachability, and Teachability Is Survivability

Aji reframes humility as teachability — essential precisely because there's no blueprint yet. Even seemingly successful AI companies may not survive a few years, so the ability to keep learning is what makes a career (and a company) durable.

  • Humility is really teachability — the ability to keep being taught
  • You need teachability most when everything has changed and there's no blueprint
  • Today's AI winners may not exist in two to three years; the playbook is unwritten
  • Teachability is what makes careers and companies survivable

“So humility is teachability and teachability is survivability.”

Oji Udezue · 28:00
#mindset#careers#learning#ai
Takeaway27:00

Get Hands-On: Code Is Now Architecture and English

Aji, a career PM who rarely wrote code, says he's written more code in the last year than the previous ten because code is now essentially architecture plus English. He's converted PRD writing into prototype writing, calls APIs himself with Postman, subscribes to every AI tool, and learns MCP by connecting things — staying sharp by doing, not pontificating.

  • Code is now essentially architecture and English (or any language you speak)
  • Convert PRD writing into prototype writing
  • Write and call API interfaces yourself with tools like Postman
  • Subscribe to all the AI tools as a cost of learning what each model is good for
  • Learn MCP by connecting different systems together

“I've written more code in the last one year than I have in the last 10 years because code is now essentially architecture and English…”

Oji Udezue · 29:00
#ai#vibe coding#product management#learning
Takeaway34:00

Learning AI? Build Around a Passion, Not in the Abstract

For people who don't know where to start with AI, the advice is to build a project around something you already love rather than aimlessly playing with tools. Ezinne's example: an older woman passionate about her wardrobe photographed all her tops, bottoms, dresses and shoes to build an 'outfit of the day' recommender based on the temperature.

  • Motivation from a specific personal problem beats generic tinkering
  • Anchor the project in an existing passion (fashion, reading, your home)
  • Example: cataloging a full wardrobe to get temperature-based outfit recommendations
  • The passion supplies the drive to keep learning as the project deepens

“she took the time to take all her tops, all her bottoms, all her dresses, all her shoes, and now she's working on based on…”

Ezinne Udezue · 35:00
#ai#learning#personal software
Takeaway49:30

Have an Opinion and Ship It — Simplicity Is Courage

Ezinne argues complicated products come from fear — not being opinionated enough because you lack conviction from time with the customer. The fix: form a point of view, ship the simplest opinionated experience, hide configurations behind the scenes, and accept being wrong. You learn more from a wrong opinion than from offering ten murky options.

  • Complicated solutions come from being afraid to take a stab and be opinionated
  • Low conviction usually traces back to not spending time with the customer
  • Ship a simple, opinionated experience; leave configurations behind the scenes
  • You're better off picking an opinion and being wrong than offering too many options
  • Too many options means you never learn which experience was better

“One of the reasons people create complicated solutions is because they're afraid to make take a stab to to to put their point of view…”

Ezinne Udezue · 49:30

“you're better off picking an opinion, shipping it shipping it out there, being wrong and then adjusting that leaving too many options because you then…”

Ezinne Udezue · 51:00
#product management#simplicity#design#decision-making
Takeaway51:30

You Can Never Over-Communicate the 'Why'

Ezinne's biggest strategy lesson: activating an organization from strategy to execution is about communication — you can never spend too much time explaining the why. She applies a crossing-the-chasm lens to strategy adoption: ~5% are early adopters who get it and move fast, while laggards are still asking why, which explains why strategy feels muddy across a team.

  • Communication is the bridge from strategy to execution; repeat the why relentlessly
  • The NASA janitor knew the goal was to send a man to the moon — everyone should know the why
  • Treat strategy communication as change management, like crossing the chasm
  • ~5% are early adopters who already want the next version while others are just getting it
  • Lagging adoption, not a bad strategy, is often why strategy feels muddy

“you you will never never spend too much time communicating the why to your organization over and over and over again.”

Ezinne Udezue · 51:30
#strategy#leadership#change management#communication