Skip to content
Aakash GuptaAakash Gupta

What AI PMs REALLY Need to KNOW in 2026 (Agents, Discovery, EVERYTHING)

Todd Olson spent 28 years in product management and built Pendo to $2.5B. He reveals why AI PM jobs doubled to 20% of all postings (and pay 30-40% more), the exact 5-layer technical pyramid to upskill from core PM to AI PM, and how to ship AI features at scale with proper evals, cost optimization, and the right product strategy. Full Writeup: https://www.news.aakashg.com/p/todd-olson-podcast Transcript: https://www.aakashg.com/the-complete-ai-pm-roadmap-how-to-upskill-from-core-pm-to-ai-pm-with-pendo-ceo-todd-olson/ ---- Timestamps: 0:00 - Intro 1:29 - Episode Begins 3:24 - Why AI PMs Get Paid 30-40% More 6:07 - How to Upskill into AI PM 11:50 - Ad 12:54 - The 5-Layer Technical Pyramid 16:30 - AI/ML Fundamentals 23:00 - Data Pipelines & RAG 33:02 - Trace Analysis & PM-Eng Tension 40:44 - Cost & Performance Optimization 48:56 - Evals Are Your Domain 56:03 - AI Product Roadmap 1:04:16 - Live Demo: Pendo's AI Features 1:13:07 - Ad 1:14:12 - Stakeholder & Board Management 1:22:03 - Outro ---- 🏆 Thanks to our sponsor: Reforge: Get 1 month free of Reforge Build with code BUILD - https://reforge.com/aakash ---- Key Takeaways: 1. AI PM market exploded - Last year 10% of PM jobs were AI PM jobs. This year it's 20%. They pay 30-40% more because of scarcity and skill level. But Todd warns: "You better damn well be good and know what you're talking about if you're gonna call yourself an AI PM because we are going to interrogate the hell out of it." 2. Real requirement is production at scale - Not "I built prototype at 1-person startup." Hiring managers want 20,000 paying B2B customers experiencing your AI feature successfully. To get there: upskill internally at current company by shipping AI features on your roadmap. 3. The 5-layer technical pyramid - Foundation: AI/ML fundamentals, data pipelines, prompt engineering. Middle: Observability (trace analysis), cost optimization, evals. Top: Product strategy, stakeholder management, leadership. You need to climb all 5 layers. Most PMs stop at layer 1. 4. RAG is table stakes - "RAG is the de facto way to build." You ingest data, create embeddings, feed into vector database, look up relevant context, pass to LLM. Todd: "If you put too much in context window, just like a human, you get confused. You want to give the right context." 5. PM-engineering tension is real - At startups, PMs do trace analysis. At large companies, engineering managers push back: "This is my world. I don't want some PM shadowing me." Similar to Data Dog—most PMs don't have login. Know the line. Be fluent but respect boundaries. 6. But evals are YOUR domain - Unlike trace analysis, evals are where PMs are the expert. "The PM is probably the best-suited human being to author and manage eval sets." You understand user and business needs. Engineers don't have that context. This is must-have competency now. 7. Cost optimization will matter - Some AI companies have sub-15% gross margins. Traditional software is 70-80%. Todd: "It's not a business at sub-15%." Eventually you'll rearchitect systems because infrastructure is too costly. Rule: when something's faster, it's cheaper (both buying compute). 8. Solve hard problems, not shiny objects - Todd's test: "Are we gonna do much better job than ChatGPT out of box? Why would we just wrap that and slap Pendo logo on it?" His discovery agent example: hard part isn't interviewing customers—it's finding which to interview, prioritizing, scheduling. Automate that workflow. 9. Kill bad features ruthlessly - Todd shipped features couple years ago that weren't great and turned them off. "Too often we hold on to something. Turn them off. Be unafraid. The more stuff in your product, the worse the experience is by default." 10. Control the narrative with boards - Don't show up with no story and get crushed with random requests. Todd: "Show them how you actually run your business. I want to see what you're looking at, not something just made for me." Think deeply about how each bet drives shareholder value. ---- 👨‍💻 Where to find Todd Olson: LinkedIn: https://www.linkedin.com/in/toddaolson/ Twitter/X: https://x.com/tolson Company: https://www.pendo.io 👨‍💻 Where to find Aakash: Twitter: https://www.x.com/aakashg0 LinkedIn: https://www.linkedin.com/in/aagupta/ Newsletter: https://www.news.aakashg.com #aipm #productmanagement #pendo ---- 🧠 About Product Growth: The world's largest podcast focused solely on product + growth, with over 200K+ listeners. 🔔 Subscribe and turn on notifications to get more videos like this.

Aakash GuptahostTodd Olsonguest
Dec 2, 20251h 21mWatch on YouTube ↗

At a glance

WHAT IT’S REALLY ABOUT

2026 AI PM playbook: technical fluency, evals, roadmaps, governance, execution

  1. AI PM demand is rising fast, but “AI PM” labeling is a marketing and scarcity game that will be heavily scrutinized in hiring due to rampant AI-washing.
  2. To upskill, PMs should build firsthand model fluency (tradeoffs, privacy, residency), understand data pipelines/RAG, and practice prompt engineering as a core communication layer with LLMs.
  3. As products become agentic, PMs need working knowledge of observability (trace analysis), production realities (SRE boundaries, access controls), and cost/performance optimization tied to gross margins.
  4. Evals become a PM-owned domain because PMs best define quality and user outcomes, while engineers supply the harnesses and infrastructure to run evaluation at scale.
  5. Strong AI roadmaps avoid shiny-object chasing by focusing on hard workflows, leveraging unique data/context, killing weak features quickly, and communicating a clear narrative to stakeholders and boards.

IDEAS WORTH REMEMBERING

5 ideas

Don’t claim “AI PM” unless you can back it up in depth.

Olson warns that premium pay invites deeper interrogation, and AI-washing is increasingly obvious in resumes and company positioning; credible experience means shipping successful AI features in production at scale.

AI is both a work accelerant and a product capability—PMs must do both.

PMs should use AI for prototyping, competitive research, and speed (e.g., Deep Research, rapid prototyping tools), while also identifying where LLM APIs can quietly improve product features even without “AI” marketing.

Model fluency is table stakes: tradeoffs beat brand loyalty.

Teams should continuously test models (OpenAI, Anthropic, Gemini, open source) across use cases, weighing quality, latency, cost, data residency, and vendor/legal friction like DPAs and country availability.

RAG and data pipelines are foundational because context quality determines output quality.

Understanding embeddings, vector stores, ingestion, retrieval, and context window limits helps PMs reason about relevance, confusion from too much context, latency, and scaling constraints in real products.

Trace analysis matters for agentic systems, but PMs must navigate ownership boundaries.

As agents call tools/other agents, tracing helps pinpoint failures and inefficiencies, yet larger orgs often reserve deep ops/debug work for engineering/SRE due to division of labor and access controls.

WORDS WORTH SAVING

5 quotes

You better damn well be good and know what you're talking about if you're gonna call yourself an AI PM.

— Todd Olson

RAG is kind of a de facto way to build… you wanna give the right context.

— Todd Olson

This is a real issue… how you build and design systems affects your cost of goods sold… and gross margin.

— Todd Olson

The PM is probably the best suited human being to author and manage these [eval] sets.

— Todd Olson

Throw it away. Do not hold onto it.

— Todd Olson

AI PM hiring trends and AI-washingUpskilling via hands-on model experimentation5-layer AI PM technical pyramidModel selection tradeoffs (quality, speed, cost, privacy)RAG, embeddings, vector databases, context windowsPrompt engineering as instruction and context designTrace analysis for agent/tool orchestrationPM–engineering boundaries and ownership tensionCost/performance optimization and gross margin realitiesEvals, experimentation, and outcome-based metricsAI roadmap strategy: workflows, unique assets, feature killingBoard/stakeholder narrative controlAI governance: privacy, bias, safety, regional regulation togglesPendo demos: agent analytics, dashboards, MCP, discovery automation

High quality AI-generated summary created from speaker-labeled transcript.

Get more out of YouTube videos.

High quality summaries for YouTube videos. Accurate transcripts to search & find moments. Powered by ChatGPT & Claude AI.

Add to Chrome