From AI Hype to Shipping - A Product perspective

Your CEO has asked you for "what are we doing with AI?"
Your Customers have "AI budgets?"
You can't figure out how to go about and what to build - it's too big, too daunting and if there's one recipe that is successful - thoughtful planning and prioritizing doesn't work with AI ...

How do you build in a world where base models are changing the ground you're standing on week by week ?

So what do you do now?

Transcript

1. AI: From Hype to "Oh Sh*t, We Actually Have to Ship This" A pragmatic framework for rapid experimentation and real-world AI delivery Chandi Kodthiwada Komodo Health
2. The Journey: From Paralysis to Shipping The AI landscape isn't just about theoretical potential anymore; it's a practical imperative. We've transformed from being stuck in evaluation paralysis to consistently shipping impactful AI solutions. Our journey takes us through strategic assessment, bold initial pilots, rapid experimentation cycles, and ultimately, a powerful rhythm of delivering AI products that customers genuinely use and truly value. Forget chasing hype or building demos that never see the light of day. This is about establishing a repeatable framework. It's about delivering AI capabilities that solve real problems, supercharge existing workflows, and create measurable value for your customers and your business. 01 Strategic Evaluation Assess the landscape and pinpoint opportunities 02 First Steps Launch initial experiments and build foundational capabilities 03 Rapid Experimentation Execute fast cycles of learning and validation 04 Continuous Shipping Sustain the delivery of tangible customer value
3. Where to Play: Four Strategic Bets Success in AI requires focus. You can't pursue every opportunity simultaneously. We've identified four distinct modalities that represent different strategic bets4each with unique value propositions, technical requirements, and go-to-market implications. Modality 1: Ubiquitous AI What it is: Embedding AI into existing workflows and last-mile experiences where users already spend their time Strategic goal: Drive delight and productivity through AI-powered micro- experiences that feel native and intuitive Think: AI where your users already are4not forcing them to learn new interfaces Modality 2: Agent-to-Agent Interfaces What it is: Positioning your platform as the definitive source that other AI agents integrate with and query Strategic goal: Become the Insight Agent for the world4the system of record that both internal and external teams rely on Think: Building the essential API infrastructure for the emerging agentic economy Modality 3: Agentic AI Experiences What it is: AI that completes complex work autonomously within your SaaS environment4not just suggesting, but executing Strategic goal: Get customers' work done faster4saving time, reducing costs, and delivering impressive results Think: Hire an AI employee who executes end-to-end workflows, not just a copilot who offers suggestions Modality 4: Deep Research Agents What it is: Specialized AI for intensive knowledge work across domains like science, finance, policy, and engineering Strategic goal: Deliver highly specialized, deeply actionable insights on- demand for expert users Think: Your expert analyst who never sleeps, constantly learning and improving domain expertise
4. AI Acceleration Principles Strong Opinions, Loosely Held Make bold bets, but remain flexible and adapt quickly. Prototype First, Perfect Later Validate with customers using prototypes before extensive investment. Re-Instrument for Speed Re-engineer processes for rapid experimentation and continuous deployment. Tightly aligned, Loosely Coupled Minimize bureaucracy and maximize cross-functional collaboration. Embrace the Duality Run parallel tech stacks for transition, experimenting aggressively while maintaining reliability.
5. How We Move Fast: The 1-2 Week Experiment Framework Speed without learning is just chaos. Our framework balances velocity with validated customer insights, allowing us to discover product-market fit through rapid iteration rather than prolonged speculation. Themes-Based Roadmap, Not Feature Lists We organized our work around broad problem spaces rather than rigid feature specifications. This gave teams room to discover the right solutions through customer interaction. Instead of "Build AI summarization," we framed themes like "Help users synthesize information faster." The what remained flexible; the why stayed constant. Experimentation as Strategy We used rapid experiments as our primary tool for discovering market needs and customer preferences. This meant embracing intentional wandering4trying multiple approaches, killing failures fast, and doubling down on signals of traction. We measured experiments not by perfection but by learning velocity. The 1-Week Prototype Sprint Created collaboratively by UX Research and Design teams, the 1-Week Prototype Sprint framework became our core operating rhythm for rapid iteration. Let the Market Lead Customers tell us where they need us most. We follow their signal, not our assumptions. The market becomes our product manager4pulling us toward real value rather than pushing us toward imagined features.
6. The Vertical Slice A vertical slice can be defined as "the sum of the work that has to be done in every layer that is involved in getting a specific feature working." Days 1-2: Concept & Visioning Define the problem space, explore solution directions, and align on what we're testing. Key deliverables include a clear problem statement and a hypothesis for the prototype. Days 3-4: Interactive Prototypes Build functional prototypes that customers can actually interact with4not just clickable mockups. Focus on core user flows and key AI interactions. Days 5-7: Vertical Slice & Beta Test Get the prototype into customers' hands and observe real usage patterns and feedback. This phase includes structured user interviews and telemetry analysis to validate hypotheses.
7. Key Takeaways: What We'd Tell Our Past Selves If we could go back to the beginning of this journey, these are the insights we'd share. They're earned through mistakes, near-misses, and hard-won successes. They're not theoretical4they're practical wisdom from the trenches of shipping AI products. Pick Your Modalities You can't be everywhere at once. Choose your strategic bets deliberately and commit resources accordingly. Spreading thin across all four modalities dilutes impact and slows learning. Ship Ugly First, Beautiful Later Perfect is the enemy of shipped. Get functional value into customers' hands fast, then iterate based on real usage. Polish matters, but it comes after validation. Build for Experimentation Velocity Optimize your systems, processes, and culture for learning speed over perfection. The faster you can test hypotheses with real customers, the faster you find product- market fit. Let Customers Pull You Stop pushing features based on internal assumptions. Let customer needs pull you toward value. Follow the strongest signals, even when they surprise you. Embrace the "Oh Sh*t" Moment Real innovation starts when you move from evaluating possibilities to shipping actual products. That moment of commitment4when the hype becomes reality4is where transformation begins.