At a glance
WHAT IT’S REALLY ABOUT
Building an AI-native product org using agents, knowledge graphs, rituals
- Mikhail’s core premise is to prevent “knowledge leakage” by continuously digitizing company context into an AI-accessible knowledge graph, enabling delegation of increasingly high-level work to agents.
- He introduces a measurable “product context coverage” KPI (e.g., 54% today) and argues that higher coverage unlocks more autonomy—from backlog decisions now to potential strategy support at 70–90%.
- He reframes the modern CPO role from process scaffolding to owning the agentic operating system: architecture, memory, tools, prompts/imperatives, rituals, training, and evaluation.
- The system automates large portions of PM and leadership overhead (status reporting, stakeholder intake, email/calendar, backlog updates, recruiting), aiming to shift PM time toward customer discovery and value creation.
- He predicts PM roles will persist but teams will get smaller and role boundaries will blur as PMs, engineers, and designers increasingly orchestrate quality and token budgets with AI collaborators.
IDEAS WORTH REMEMBERING
5 ideasTreat organizational context as an asset to be continuously captured, not tribal knowledge.
Mikhail’s system aims to reduce dependency on individual “knowledge bottlenecks” by storing product, customer, business, and technical context centrally so it persists even when employees leave.
Measure AI usefulness with a context coverage KPI tied to autonomy.
He uses a prompt-based “product context coverage” score (industry, business model/PNL, customers/segments) as a CPO KPI; higher scores justify delegating larger decisions to the agent.
Use the knowledge graph to diagnose silos and discovery quality, not just to answer questions.
The graph visualizes how teams/PMs connect to customers, projects, and stakeholders; sparse connectivity can signal poor discovery, weak stakeholder management, or organizational silos.
Store raw transcripts for retrieval; summarization can reduce recall.
After testing, they found summarization harmed retrieval accuracy (reported ~20–25% worse recall) due to lost nuance and imposed templates, so they keep raw meeting and conversation transcripts.
The “agentic CPO” owns scaffolding, rituals, and evals—don’t outsource it lightly.
Mikhail argues CPO ownership maximizes iteration speed and ensures accountability because the agent changes decision-making and business outcomes; delegating to AI ops/engineering risks slower feedback loops and misaligned incentives.
WORDS WORTH SAVING
5 quotesIf you could have a single storage of this entire business, customer, product, technical knowledge in one place, then, uh, it would increase, like, the entire value of your organization.
— Mikhail
Summarization actually hurts a retrieval.
— Mikhail
I don't read my email anymore, like an agent does it for me and pings me in case if it's, uh, if it's urgent. I don't manage my calendar anymore.
— Mikhail
This automates probably like 70, 75% of the, of the, uh, recruiting workflow.
— Mikhail
Not only product management is going to exist, I think it's gonna thrive.
— Mikhail
High quality AI-generated summary created from speaker-labeled transcript.
