Most PMs struggle with AI. And that's not because the tech is hard, but because they bring the wrong mindset. They try to manage AI work like traditional software: fixed timelines, predictable delivery, and clear answers early. But AI doesn’t work like that. Variance is the norm. Exploration beats estimation. Confidence intervals matter more than deadlines. The PMs who thrive in AI think differently, plan for uncertainty, and align teams on outcomes. I know, I know. Some of this might sound strange if you have only worked in traditional product teams. AI flips a lot of familiar PM instincts. And unless someone explains why, it all feels “off” or inconsistent. This is the shift we will break down in this week’s AI PM Masterclass (yep, on 22nd this week), the mental model every PM needs for the next decade. We are sharing pre-reads, examples, and frameworks in the WhatsApp group before the session. 👉 Join the WhatsApp group (link in registration email) 👉 Register for the event here See you inside.

Traditional software behaves predictably. AI doesn’t, and that changes everything. In non-AI software: Requirements = clarity + control. Same input → same output. In AI systems: Requirements = guardrails + boundaries. Same input → multiple valid outputs. You define what the system can’t do, and what “good enough” means. Most PMs miss this mindset shift because deterministic thinking breaks in probabilistic systems. For example: Wrong password → show error. (Predictable.) “Recommend the best song right now” → no single right answer. Only probabilities. That’s why AI PMs ask different questions: - What accuracy range is acceptable? - What failures must never happen? - How should the product behave when it’s uncertain? - Where are fallbacks needed? - What does “safe” actually mean here? Requirements shift from fixed outputs → flexible constraints. And this impacts everything: - Roadmaps: milestones = model improvement - UX: UIs must support uncertainty, editing, and explanations - Testing: evals > test cases; quality becomes continuous If you are building AI products, this mindset isn’t optional anymore. I’m breaking this down further in the upcoming AI PM masterclass. Register for the FREE masterclass here. And oh! Do not forget to join the WhatsApp group for updates and engaging conversations.

Hey everyone, Quick update, we are moving the masterclass from Saturday, November 15th, to Saturday, November 22nd, due to some unexpected circumstances. Sorry for the late change, and thank you for being so understanding. We really appreciate it. We will make sure the session on the 22nd is absolutely worth your time. In the meantime, we are active inside the WhatsApp group where we will keep sharing resources and continue the conversation. If you want to join the group, just register for the event using this link. The WhatsApp link will be sent to you in the confirmation email. 👉 Register HERE Thanks for being part of this journey.

Last month, a PM friend told me he spent two sprints adding an “AI-powered” feature to his app. When I asked what it did, he said, “It summarizes user feedback.” I asked if users were asking for that. He paused and said, “Well... not really.” Sound familiar? Most PMs are using AI even when it doesn’t make sense. They want to stay relevant but end up chasing the buzz. The biggest mistake they make is starting with “How do we use AI?” instead of “Should we use AI?” Here’s a simple checklist to answer that second question 👇 ✅ Use AI when... ✚ The problem is ambiguous... there’s no single right answer. Example: writing ad copy ✚ You have lots of data to learn from. Example: recommending products, ranking content ✚ The task needs human-like judgment or creativity. Example: suggesting designs, reviewing resumes ✚ The system can tolerate small errors. Example: auto-suggesting captions, drafting emails ✚ AI can make it 10× faster, smarter, or more personal. Example: chat-based help desks, content assistants 🚫 Don’t use AI when... ➖ The task is deterministic... there’s one correct answer. Example: calculating tax, generating invoices ➖ You have little or no data. Example: brand-new apps, small internal tools ➖ The problem can be solved with clear rules or formulas. Example: filtering spam with keywords ➖ Mistakes are high-risk. Example: approving loans, medical diagnosis ➖ AI adds complexity without clear value. Example: replacing simple forms with generative chat I understand this can be overwhelming. Everything around you seems to need AI, and it’s easy to feel like you are missing out if you don’t add it. But the truth is not every problem needs AI. The real skill is knowing when it actually makes sense. Want to see this in action? We are running a FREE masterclass next week: 🗓️ 15th Nov | ⏰ 4:30 PM IST / 11 AM GMT 💰 Free entry | 🎟️ Limited seats Learn how to decide if your product idea really needs AI and get your AI Fit Checklist to test it yourself. Sign up HERE