We can't find the internet
Attempting to reconnect
Something went wrong!
Hang in there while we get back on track
Big Ideas
AI is moving PM work from document-first to evidence-first. Sachin Rekhi describes five shifts: teams replace weeks of detailed specs with prototype refinement; put prototypes in front of customers before internal review; let PMs and designers make small AI-assisted code or copy fixes; shorten roadmaps to six months; and have specialists build agents for analysis, research, and design. The practical change is to make the prototype the first learning artifact: use it to test behavior, discard weak directions cheaply, and write heavier documentation around decisions that survive customer contact.
Throughput is becoming cheap; judgment and exception handling are not. Anish Acharya says one Google team compressed two years of roadmap into three months and found the harder question was what to add. He describes loops that take input through reproduction, fix, review, risk-based approval, and deployment; humans remain needed for strategy, sales, support, exceptions, and out-of-distribution thinking. A current operating example from Aakash Gupta: OLX Uzbekistan’s CPO has run a shared agent for five months that answers feature, result, and roadmap questions; a personal assistant handles 70% of recruiting, and he estimates AI now takes on work that used to consume 50% of his time. The proposed architecture separates identity, memory, recurring skills, and tools.
Tactical Playbook
Run enterprise vision as decision extraction, not executive brainstorming.
- Start with people closest to workflows and users; map current-state gaps with the other PMs before asking executives for strategy.
- In the session, align on the business outcome, affected user, current cost, materially better state, and acceptable trade-offs. Return with a proposed vision, not a blank-page exercise.
- Keep the artifact small—outcomes, target experience, principles, capability gaps, and non-goals—and delay individual feature requests. Translate vague goals such as “improve customer experience” into a measurable outcome such as an NPS or PSAT change.
Make the source of truth earn its keep. Product knowledge scattered across PRDs, Slack, memos, and unwritten rules can create misunderstanding, onboarding friction, and unclear dependencies. A lean pattern is one frequently updated product document with a table of contents and timestamps, plus searchable decision records; practitioners say AI can help maintain and navigate the canonical record. But others report that a Confluence initiative became unused overhead and that junior-PM time disappeared into document organization. Apply the smallest artifact that answers repeated questions, then audit whether people actually use it.
Case Studies & Lessons
Sell the economic outcome, not “AI automation.” A legal engineer had built time-saving tools, but months of outreach, ads, and offer experiments produced no business; they were considering a free first automation. Community feedback suggests the blocker may be buyer economics: billable-hour firms may not retain the value of time saved, while in-house and flat-fee practices do. The recommended test is a paid, fixed-scope pilot with an NDA and a measurable before/after—hours saved, error rate, or billable time recovered—not free work. Treat this as a segmentation experiment, not proof that the new segment will convert.
Test the delayed value loop. An ecommerce founder tested checkout, emails, and the thank-you page, but a “subscription” was actually a one-time purchase; renewals never fired and the number of lost customers remained unknown. For recurring billing, add a daily reconciliation between the app’s active-subscriber list and Stripe’s subscription objects so a mismatch is caught before the next renewal.
Career Corner
Price ownership against lifestyle, not just title or cash. An eight-year PM is weighing a remote consulting role with brand, bonuses, and limited product ownership against a zero-to-one agentic-AI role with a roughly 60% raise and CEO visibility—but poor prior culture, three office days, and a commute of two-plus hours each way. The useful decision rule from the discussion is phase-of-life dependent: score decision rights, learning, manager quality, health and commute, and compensation separately; a demanding role can still be the better move when the work is more aligned.