Inari (YC S23)

Tips on making annual planning actually work for your product team. - with ❤️ from recent planning struggles from Inari (YC S23)

We’re stoked to ship a big revamp on how product teams explore and discover new customer insights using Inari (YC S23). We reworked the flow so that when generating insights, feedback gets clustered immediately and new insights are streamed live. You can set a prompt, ask specific questions, set custom filters for the feedback you’re looking to analyze, then identify product insights in real-time. If you like the insight generated, you can save it and we’ll create a robust deep dive and monitor the topic moving forward. If you dislike the trending insight, you can ignore it then run a new query to replace it with different results. This makes asking questions and surfacing insights 100x faster and easier than before and is extremely powerful since Inari unifies all of your customer data sources like Gong, Intercom, Zendesk, Slack, HubSpot, and Salesforce. If you’re currently stuck annual planning and want to summarize your customer’s top problems and requests with zero work to inform your roadmap, hit me up and I’d love to run this for you to get feedback!

Christmas feels like it came early today for folks excited about the AI code-gen and AI product building space: 1. Devin, the first publicly announced autonomous AI software engineer from Cognition, moved into general release so you can now just delegate issues to the bot and it'll draft PRs for fixes. 2. Replit's Agent product also jumps out of early access today and they've made big improvements in UI-generation through writing React + releasing "Assistant" for iterating on existing projects. 3. ChatGPT's Canvas now lets you write and edit code directly within the app, more directly competing with products like Claude's Artifacts and Vercel's v0. 4. 2-3 weeks back, GitHub also demoed an updated Copilot Workspace for running agentic workflows in your repo and generating 1st draft solutions. We'll be spending our week @ Inari (YC S23) playing around with a bunch of these tools, seeing how far these agents have improved, and figuring out how we can revamp our product development workflow. We have 100's of little requests, improvements, and bugs collected in Inari from customer feedback that we've been itching to throw at these agents. If it works, we're one big step closer to the dream of a self-organizing and developed backlog 💪🏼 If there's any apps or improvements that you're curious if the AI coding agents can handle, I'd love to hear about them so we can try them out!

Fun story from Elle Luna on walking up to Travis Kalanick 3 whiskeys deep at a bar, telling him she could redo Uber's entire design, and TK pulling together a team by Monday morning to redesign the whole thing in 3 weeks. Real builders move extraordinarily fast and get shit done.

I'm calling it now: 2025 will be the first official year of the full-stack “AI product builder”. 1️⃣ Product, design, and engineering roles will merge or hyper-specialize 🔀 With AI handling initial technical tasks, operators either become hyper-generalists that navigate product, design, and engineering to ship products quickly or hyper-specialists that push the boundaries in their domains. Those in between, like middle managers, become less relevant. 2️⃣ Everyone gets an AI coding agent 🤖 AI coding tools like Cursor, Bolt, and v0 will generate 90% of new features with few-shot prompts once they’re integrate with existing codebases. Paradoxically, total working hours won't decrease. Instead, time shifts to fixing AI-driven tech debt, addressing last-mile edge cases, and building new + creative products instead. 3️⃣ Product transitions from prioritizing to polishing 💎 Why debate amongst stakeholders for weeks about which initiatives to take on when you can reallocate that time to build prototypes with AI then get feedback from users instead? PMs will be valued less for prioritizing roadmaps and more for having the taste to prompt and shape experiences that customers love and polishing the last mile with their teams. 4️⃣ Product ideas grounded in customer needs becomes the limiting factor 💡 The bottleneck for growth shifts to finding better product ideas grounded in genuine customer needs. Workflows that capture customer insights and help teams discover differentiated solutions will emerge. AI coding agents perform 100x better with context about customer journeys so they’ll be better integrated with tools that analyze customer data and help teams plan and ideate (like Inari (YC S23) 😜). 5️⃣ AI floods inboxes and tanks SEO so builders will go “direct to humans” 📥 AI SDRs will saturate inboxes while ChatGPT and Perplexity erode traditional SEO-led growth. As a result, product builders unfortunately have to become content creators to stand out and community managers to gather feedback and build deeper customer relationships. I'd love to hear other ideas + hot takes on how you think product building and startups change next year! Roast these takes here or DM me 😅

This Thanksgiving, I’m most thankful for: 1. The 3,254 cold outbound emails I received from AI SDR’s this year 🤖 2. My crypto portfolio recovering so my family stopped disowning me 📈 3. Replacing my ability to think for myself by prompting Claude 🧠 4. Replacing my wife with an AI girlfriend ❤️ 5. OpenAI board not firing Sam Altman this Thanksgiving 🙏🏼 6. OpenAI and Perplexity incinerating VC money so that I don’t have to see sponsored Google ads (for now) 🔥 7. America not implementing an unrealized capital gains tax 🇺🇸 8. Having the opportunity of building dashboards in Inari (YC S23) that give executives deep insight into critical business functions 🎤 9. Most importantly, delivering stakeholder value 💯 Have a Happy Thanksgiving everyone 😅

These are the harsh truths we’ve seen after talking to 100’s of teams running continuous product discovery processes at Inari (YC S23): 1️⃣ Your favorite ideas will be the first to go Many teams start product discovery wanting to validate a pre-conceived idea. More often than not, those ideas don’t survive real-world feedback. Stay open-minded, let data instead of ego drive decisions, and be willing to give away your Legos. 2️⃣ Users don't always tell the truth — even when they think they do It’s human nature for interviewees to tell you what they believe you want to hear, so you’ll get dramatically different answers depending on how you phrase your interview questions. Ask about their actual observed behaviors, workflows, and attempts instead of asking whether they have a problem or request in the first place. 3️⃣ Customers don't know what they want until they feel it. It’s tempting to ask questions like “what’s the ideal solution for X…?” or “if you could wave a magic wand for Y…?”. 99% of those answers go nowhere - most customers haven’t thought about deeply enough about a specific problem to give you the ideal solution. Anticipate their needs, share prototypes, and you’ll get 10x better feedback. 4️⃣ You can't please every stakeholder and many truths are unwelcome Stakeholders resist uncomfortable findings, and trying to appease everyone slows customer learning and dilutes feature quality. Focus your discovery on customer problems and use collect a strong mix of qualitative and quantitative evidence to align stakeholders. 5️⃣ It's never truly "done." Bad discovery is a one-time sprint to validate a pre-conceived notion. Great discovery is an ongoing and endless habit. To quote Jeff B - “customers are beautifully and wonderfully dissatisfied” and “always crave something better”. Teams that continuously talk and act on customer needs win in the long-run. 6️⃣ Teams waste a HUGE amount of time on work about work. 99% of product discovery processes fail since they’re manual and time consuming. We’ve seen way too many messy Figjam boards and complaints that linking feedback, synthesizing insights, and connecting requests to revenue impact in your CRM is mind numbing. Instead, you can automate all of it with Inari (YC S23) 😁

The startup playbook of shipping MVPs sounded sacred to me: ship fast, get feedback, iterate based on customer needs quickly. But when we applied it at Inari (YC S23) and Main Street, it backfired. The common startup failure case we tried to avoid was building the perfect product in isolation for years, not getting feedback or validation, then wasting years after shipping the wrong thing. So in early stages working on products, we often optimized for velocity at the expense of quality. Instead of unlocking delight, we ran into 2 failure cases: 1. Instead of insights about the product’s core thesis, most complaints centered around bugs and confusing designs. Testers wasted time and goodwill stuck on broken buttons or flows and rarely shared feedback on the idea itself. 2. The faster we shipped, the wider the types of feedback received. Everyone pulled us in different directions. When trying to follow it all, we ended up with an unfocused product that didn’t excel at solving any one use case. A few months ago I stumbled on Jason Cohen’s take on SLC products (Simple, Lovable, and Complete) and tried it out. Once we stopped only “shipping fast” but instead “shipping right”, our results improved. - Simple: Ruthlessly narrow down the product that solves a core problem, but build quickly since resources and time are limited. - Lovable: Create an experience that customers love to use for the narrow use case. Build something well-designed, delightful, and emotionally resonant. - Complete: No one wants multiple half-baked tools. Build a v1 product that solves at least one problem well instead of doing multiple things poorly. Talk to users. Find a really painful problem. Build a simple, lovable, and complete solution to delight them. Then expand from after the customer delight is there. PS: please don't roast me too hard for my 5 minute Canva creation 😂

On how to build delightful and fun products. Make products playful 🚂 Pokemon Go turned walking into an addictive game by adding collection, progression, and scarcity. Players discover new Pokemon, collect them to increase their levels, with weekly rotations and varying rarity levels that make collecting satisfying, delightful, and time-bound. Fun fact: Pokemon Go has made $8B+ revenue so far and the new Pokemon Sleep, which gamifies sleep, earned $120M+ in its 1st year. Design products for emotional resonance ❤️ World of Warcraft drives emotional resonance by letting players tackle epic raids and events together in guilds and allows infinite ways to personalize through playable races and gear. There’s great reasons why most WoW players played 20-40+ hours per week - it was a masterclass on progression (new levels, rewards, and quests), personalization (different races, gears, and skills), AND community (raids, guilds, shared world) Craft rewarding journeys 💎 Right after sign-up, Marvel Snap players jump directly into a simplified version of their battle mechanic where Nick Fury walks through how the game works. You play against easier bots or similarly-leveled humans with a basic set of cards to ensure you feel confident and not overwhelmed. You also can’t access the complex cards and battle mechanics without progressing naturally. Build shared experiences 👥 Let’s be real: the best part about Strava is stalking your friends on the leaderboards or feed then bragging to each other about your mileage and speed. We learned on NBA Top Shot that people love getting to the top of leaderboards to feel accomplished and a huge side benefit is that other less accomplished players can see the playing and collecting behavior of those at the top - providing examples of how best to “hack” the game. Simplify and iterate 🔁 Superhuman makes email faster, manageable, and “fun” by giving you mastery, autonomy, and purpose. You can unlock keyboard shortcuts to navigate quickly, customizable and split inboxes to organize your workflows, and gives subtle rewards like visual celebrations when clearing your inbox or undoing sends. Every feature is fast and useful, solves a real problem, but fun without being distracting or gimmicky. cc Inari (YC S23)