50 AI Leaders Redefining Product with Copilot Technology

Slapping AI onto your existing product isn't a strategy. It's a shortcut to irrelevance. Michael Smith has been building products long enough to know the difference between genuine innovation and shiny object syndrome. His take on AI product management cuts through the noise that most of our industry is drowning in right now. Three insights that deserve your attention: • AI product discovery operates on fundamentally different rules than traditional SaaS. You can't just run the same discovery playbook with an AI layer. The problem space, the solution space, and the validation methods all shift. Prompt engineering during discovery is now a core PM skill. • The MVP for an AI product should target core capabilities, not polish. Most teams over-engineer the AI component and under-invest in understanding whether the problem even warrants an AI solution. Start with the question: "Would this be better solved without AI?" • Vanity metrics in AI products are particularly dangerous. High usage doesn't mean high value. Teams need metrics that capture whether AI outputs actually changed decisions or outcomes for the user. 🎧 Full episode: https://lnkd.in/gBzpRvJ4 #AIProducts #ProductManagement #ArtificialIntelligence #ProductStrategy #AIProductManagement

Managing Multiple MRs During AI Adoption (Without Breaking Your Sprint) AI adoption creates MR chaos: experiments, prompt changes, infra tweaks, UI updates - all at once. Sprint velocity suffers when everything moves in parallel. Here’s the lean approach: 1️⃣ Separate Discovery from Delivery Experiments live outside the sprint. Only validated approaches enter the backlog. 2️⃣ Keep MRs Small Split infra, AI logic, and UI into separate MRs. Small merges = faster reviews, lower risk. 3️⃣ Add an AI Merge Gate Before approval: ✔ Observability ✔ Fallback logic ✔ Cost visibility ✔ Basic edge-case testing 4️⃣ Enforce WIP Limits Limit active AI MRs per squad. Flow beats multitasking. AI doesn’t slow teams down. Uncontrolled experimentation does. Control the flow - and AI becomes a velocity multiplier. #AIAdoption #MRchaos

Key question for our weekly AI Ship Day today: how can we compound our advantage in our AI operating system, workflows and AI teammates? As AI adoption accelerates, what keeps me up at night is how fast we can build and adapt. The team's demos today highlighted this accelerating pace at which our platform for AI workflows and agents is evolving. My challenge to the team this morning was to embrace a deeply AI Native perspective in generative AI products. What was best in class last year is no longer best in class today and certainly not future proof for tomorrow. For some legacy apps, we have decided to pause and sunset. For many apps, we have decided to upgrade with frontier augmentation capabilities. For our fresh innovations, we have streamlined how AI enhances how we build, with scoring parameters across automation, augmentation and target impact (relative to token costs). Systems level visibility, aka our "AI HR Department," guides the performance of AI apps and teammates in real time, benchmarking where we should set for semi-autonomous or autonomous workflows. A new MCP Hub now connects more than 30 live AI applications into a single operating environment where agents can discover context, tools, and capabilities and self coordinate as part of a larger system. Underneath it all, our memory layer allows context to persist across the workflows of AI and Humans. Instead of interactions resetting, the system logs actions and decisions, compounding how his recursive learning strengthens the collective intelligence and builds durable operational knowledge over time. Always energizing to have every AI Ship Day push the frontier forward of what we do, learn and improve. Learn by Doing.

"I used to spend hours on repetitive manual tasks. Now, I have a 'Gemini Gem' doing the heavy lifting for me." Automation is great, but personalization is the game-changer. I’ve been exploring how to move beyond basic AI prompts to create custom workflows that actually understand my specific business needs. Here are my three biggest takeaways from my recent deep dive into Generative AI and workflow optimization: -Customization is King: By building specialized "Gems," I’ve been able to automate the nuances of customer service management and billing coordination that a generic AI would miss. -The 80/20 Rule of AI: AI can handle 80% of the drafting and data sorting, leaving me the final 20% to focus on high-level strategy and human-centric problem solving. -Future-Proofing: The labor market is shifting. Staying ahead means not just knowing that AI exists, but mastering how to integrate it into a professional branding and productivity roadmap. The goal isn't to replace the human element—it's to amplify it. 🚀 How are you leveraging Generative AI to reclaim your time this week? Let’s swap notes in the comments! #GenerativeAI #ProductivityHacks #DigitalTransformation #FutureOfWork #AIWorkflows #ProfessionalGrowth

After going deep into AI over the past several months, one thing has become clear: We don’t have an AI problem. We have an infrastructure problem. The models are good enough. The issue is everything around them. When you try to move from demo → real system, things start breaking: • workflows need orchestration   • agents lose context between steps   • execution has no memory or structure   • reliability becomes unpredictable  So teams end up doing what we’ve all been doing: Building glue code around AI. But that doesn’t scale. What’s missing isn’t another model. It’s a layer that can actually manage how AI systems operate. Something that can: • coordinate multi-step workflows   • manage state across execution   • balance and bound quality vs cost   • structure how agents work together  Right now, most approaches still hardcode too much of this into workflows instead of letting systems evolve dynamically. That’s why we’re still "babysitting" agents. And why moving from prototype → production feels harder than it should. I’ve been spending a lot of time thinking about what this layer should look like. Starting to see some patterns emerge. More on that very soon.

Question: Why aren't you using AI if it's faster? Because fast automation of a bad process = fast failure. Here's what I'm doing instead: ✓ Run my manual template for 2 weeks ✓ Document every friction point ✓ Identify what actually matters ✓ Understand what "good" looks like THEN build the AI layer. The difference: Most people: Manual → AI → Hope Systems builders: Manual → Clarity → Structure → Automation One creates automated chaos. One creates scalable systems. Day 4. Choosing clarity. @QarmiQarma Documenting AI systems from real frontline work. #QarmiQarma #SystemsThinking #AIFromRealWork #FrontlineSystems #CareerTransition

Dev Tip: Split your AI workflow by cost vs capability. Use a premium model (like Anthropic) to design, refine, and upgrade your agents + skills — this is where quality matters. Then shift testing and iteration to a cheaper model (z.ai, OpenRouter, etc.). Run it repeatedly, break things, tighten prompts, and get it passing consistently. You end up with: • Higher-quality agents • Way lower cost during iteration • Better real-world robustness (because cheaper models are less forgiving) Think of it like this: Strong model = architect Cheap model = stress tester If it works well on the cheap model, it’ll usually perform even better on the premium one.

AI is easy to get excited about because the first version always looks good. You wire up a model, throw a prompt at it, get something clever back and suddenly everyone starts talking about transformation. Then comes the question that actually matters. Can this live inside the product, hold up under real usage and not create fresh problems somewhere else? That’s where most teams find out the real work was never the demo. TSH builds AI that holds up, not just demos that look good in a room!

If you scroll LinkedIn long enough, most AI posts start to blur together GenAI. Agents. Vibe coding. New tools every week. The real signal, though, is the rise of AI-native organizations. Everyone has an AI opinion. But the shift is simpler: Not pilots. Not experiments. Not another chatbot. PwC helps cut through the AI noise so you can solve the business problems that matter. Learn more at https://pwc.to/3N8p1dR #AgenticAI #ResponsibleAI #Copilot #FinancialServices #PwC