Jonas Alves’ Post

How systematic experimentation is reshaping product decision-making and organisational learning. Our latest blog explores what it really means to move beyond A/B testing toward a culture of continuous, evidence-based discovery. Check out Jonas's post below.

Have a read of our new blog which examines how experimentation maturity transforms intuition into measurable insight, turning testing into an engine of organisational intelligence. Check out Jonas's post below.

⚡️ AI made us faster - but also more distracted. Over the past few months, I’ve seen a strange trend: PMs (me included sometimes) are being asked to build, not just prototype, but build entire tools/applications using AI. And honestly, it sounds amazing on paper. “Look, the Product team is building on its own!” Cue the leadership applause. But behind the scenes? It’s usually chaos wearing an “innovation” badge. AI is brilliant for quick mockups, ideation, or validating assumptions. But when PMs start half-coding features with AI and then pass them to engineers to “just integrate”, what you end up with isn’t speed, it’s confusion with good intentions. Because let’s be honest, the idea behind this isn’t innovation. It’s optimization. Everyone wants to look efficient. “Let’s ship faster. Let’s build in parallel.” But fast and right are two very different things. While PMs are busy prompting, debugging, and pretending to be pseudo-engineers, We're losing time on what actually matters - talking to users, understanding pain points, brainstorming solutions and aligning teams. Engineers get frustrated fixing AI-generated spaghetti. PMs get burned out proving “we can build too.” And leadership gets a shiny demo… that no one touches a week later. Speed means nothing if you’re sprinting in circles. AI should be our co-pilot, not our cop-out. Use it to think sharply, prototype faster, and not to skip the hard work of discovery and collaboration. PMs don’t need to prove we can code. We need to prove we still know what’s worth building. Because when everyone’s building everything, no one’s solving anything. #ProductManagement #LifeOfAPM #Leadership #CareerGrowth #BuildInPublic #ISB #Careerreflections #MBA

𝐇𝐨𝐰 𝐬𝐡𝐨𝐮𝐥𝐝 𝐩𝐫𝐨𝐝𝐮𝐜𝐭 𝐥𝐞𝐚𝐝𝐞𝐫𝐬 𝐞𝐯𝐨𝐥𝐯𝐞 𝐟𝐨𝐫 𝐀𝐈-𝐝𝐫𝐢𝐯𝐞𝐧 𝐩𝐫𝐨𝐝𝐮𝐜𝐭𝐬? When building non-AI products, product managers were building software systems that were deterministic. These systems had predictable inputs, defined rules, reliable outputs. AI changes that. It’s probabilistic by nature. Every output is a likelihood, not a certainty. That shift forces us to rethink how we define, plan, and evaluate products. Here’s how Product Managers should evolve their thinking to build AI products: 𝟏) 𝐒𝐭𝐨𝐩 𝐰𝐫𝐢𝐭𝐢𝐧𝐠 𝐟𝐢𝐱𝐞𝐝 𝐏𝐑𝐃𝐬. 𝐒𝐭𝐚𝐫𝐭 𝐝𝐞𝐟𝐢𝐧𝐢𝐧𝐠 𝐥𝐞𝐚𝐫𝐧𝐢𝐧𝐠 𝐜𝐫𝐢𝐭𝐞𝐫𝐢𝐚. You can’t “spec” behavior in advance. You need to describe what the system should learn. Replace acceptance criteria with evaluation criteria (precision, recall, user trust, drift tolerance, etc.). 𝟐) 𝐌𝐨𝐯𝐞 𝐟𝐫𝐨𝐦 𝐟𝐞𝐚𝐭𝐮𝐫𝐞 𝐫𝐨𝐚𝐝𝐦𝐚𝐩𝐬 𝐭𝐨 𝐦𝐨𝐝𝐞𝐥 𝐟𝐞𝐞𝐝𝐛𝐚𝐜𝐤 𝐥𝐨𝐨𝐩𝐬. Roadmaps don’t capture the iterative nature of model tuning. Build mechanisms for continuous labeling, A/B testing, and post-launch retraining. 𝟑) 𝐂𝐨𝐥𝐥𝐚𝐛𝐨𝐫𝐚𝐭𝐞 𝐝𝐞𝐞𝐩𝐥𝐲 𝐰𝐢𝐭𝐡 𝐝𝐚𝐭𝐚 𝐚𝐧𝐝 𝐫𝐞𝐬𝐞𝐚𝐫𝐜𝐡. The PM’s job now extends upstream- curating data quality, understanding biases, and helping define success metrics that balance user experience and model performance. 𝟒) 𝐑𝐞𝐝𝐞𝐟𝐢𝐧𝐞 “𝐝𝐨𝐧𝐞.” An AI feature is never done. Shift the mindset from “ship and forget” to “deploy, observe, refine.” 𝟓) 𝐄𝐯𝐚𝐥𝐮𝐚𝐭𝐞 𝐰𝐢𝐭𝐡 𝐛𝐨𝐭𝐡 𝐦𝐞𝐭𝐫𝐢𝐜𝐬 𝐚𝐧𝐝 𝐯𝐚𝐥𝐮𝐞. Track precision and recall, but also measure user confidence, trust, and perceived control. AI products don’t just need different technology, they need different leadership muscle. The best PMs I know aren’t writing thicker specs. They’re building smarter systems to learn faster. What’s one way your product process has changed (or needs to) for AI-driven work? #ProductLeadership #AIProductManagement #MachineLearning #AIMindset #ProductStrategy #DigitalHealth

Yeah, this Sam Gaddis paper IS THAT GOOD. Additionally, check out The Lightcone Podcast and their latest episode on this topic as well. Their focus is on startups, not established organizations, but EVERYONE should be paying attention to how startups are approaching AI. In short, if YOUR organization doesn't figure it out (and no, rolling out Microsoft CoPilot is not "figuring it out"), there's a hungry kid in a hoodie with an aunt in your niche building an AI-native version of you, and quickly. (https://lnkd.in/gj_wN-cH)

I recently dove into Sam Gaddis’s excellent Field Notes piece, “The AI Adoption Paradox: Why 95% of Companies Fail (And How to Be in the 5% That Succeed).” The core insight is this: the technology is rarely the barrier. The barrier is in mindset, structure, and leadership. Listening to Sam Gaddis speak at Austin Tech week about engaging your software‑development team, unleashing their full talent, and directing the ship while granting crew creative agency was deeply inspiring. It reminded me of many long conversations with Paul Phelps where we explore what it means to shift project‑management culture within the enterprise (More to come in our article). The most effective teams I’ve seen are clear about direction and purpose but enable the people doing the work to drive the execution with autonomy. Sam’s approach captures that sweet spot between leadership and freedom. Really enjoy his analogy about getting as many at bats as possible within a circuit. If you’re building software, AI systems, or leading a technical team, I highly recommend you give Sam’s essay a read. Then ask: how much of our architecture, governance, and team structure is designed to enable performance, not just measure it? Great work, Sam. Looking forward to where you and your team lead next. Article here: https://lnkd.in/gjaq5N5W #AIAdoption #ProductLeadership #EnterpriseInnovation #TechLeadership #CreativeTeams #AgileMindset #ProjectManagement #FutureOfWork

🔬 The Science behind Product Management 🔬 ⚡ The end of roadmaps? The end of Agile as we know it? ⚡ According to research, “elite” engineering teams can already deliver changes with lead times under 4 hours. High performers are under 24 hours. Medium performers? 1 day to 1 week. 👉 Source: getdx.com - Lead time for changes We’re heading into a future where any feature can be shipped instantly. Yes, data silos exist. Yes, legacy code slows us down. But those are solvable problems—and companies will solve them, because the payoff is irresistible: instant delivery. And in that future, some sacred cows of product and engineering may no longer make sense: 👉 Roadmaps → Do we still need them if every feature can be delivered on demand? Imagine each customer running their own personalized version of your product. 👉 Feature teams → Why build features at all when the real game is maintaining a platform? Non-technical people could define and change user journeys or business processes directly, without waiting for engineering. 👉 Agile → If AI enables us to ship in one week what took all of 2024, what’s the point of sprints? Will they shrink into infrastructure cycles, platform upgrades, and guardrails for instant creation? We’re on the verge of a complete redefinition of what product and engineering do. 🔥 My provocation: In the AI era, “delivery” disappears as a constraint. The real challenge won’t be shipping—it’ll be deciding what should exist at all. What do you think? Are roadmaps obsolete? Does Agile collapse—or evolve into something else? And what’s the role of product managers and engineers when software becomes infinitely malleable? Let’s debate this. The future of product is being written right now.

Great roundup from Agile Analytics this week 👇 The line between AI engineers and software engineers is getting thinner by the day — and it’s wild to see how fast that’s changing how teams build and ship. From AI helping rescue struggling products to new research on what really makes Agile work (and what’s just ritual), this one’s worth a read. Personally, I love the angle on AI-powered delivery — turning chaos into clarity feels like the new superpower. ⚡

How to build a product? (That actually works) Most people start with an idea. Great builders start with a pain. When you build from pain, you see what others miss. You stop guessing. You start feeling. If you ask me to build a product, here is how I will build every product now: ➜ Start with a scream, not a survey. Find the moment where users are truly frustrated — the 3 a.m. problem. ➜ Serve the dissatisfied minority first. The average user is comfortable. The minority is desperate — and that’s where innovation lives. ➜ Build a small thing that changes one behavior. Not a big thing that does everything. Behavioral change is the only real validation. ➜ Ship fast, learn faster. Every version is a question disguised as a product. ➜ Fall in love with feedback, not features. The best PMs are students of human truth. If you ever want to know what to build next — go back to the pain. It never lies. #AI #Product #PM #ProductManagement #BuildWithNeel

From Frameworks to Foresight: Wrapping up Module 4 of my AI PM Journey Ever realised that great products aren’t built feature by feature, but decision by decision? That’s what Module 4 was all about. 💡 1. Vision is shared, not handed down. In startups and enterprises alike, product managers co-own the product vision. I’ve seen how aligning strategy with leadership intent and user insight prevents “build vs. business” conflicts later in the cycle. 🔐 2. Security and compliance aren’t checkboxes, they’re "differentiators". In regulated spaces like fintech or healthcare, products win trust when security is part of the roadmap, not a post-launch patch. For example, adding audit logs early in an MVP avoids costly redesigns once enterprise clients come onboard. 🎯 3. Value propositions must solve before they sell. We practiced crafting one-line value statements that answer: Why this? Why now? Why us? I’ve since applied this to reframing feature requests, shifting the focus from what we build to what problem it eliminates for users. 📊 4. Frameworks are only useful when they drive action. SWOT, Porter’s Five Forces, and PESTEL aren’t slides; they’re lenses. Using them together helped me map competitive gaps and spot risks like supplier dependency or pricing sensitivity. Next up → Module 5, where I’ll explore how AI product managers measure strategy through outcomes and impact. #ProductManagement #AIPM #EnterpriseSoftware #Strategy #Microsoft #Leadership