Sarah closed her laptop after another marathon day of back-to-back meetings. She glanced at her phone and smiled, a notification from Granola had already transformed her scattered notes from six different calls into beautifully organized summaries. No awkward meeting bots had invaded her calls. No clunky interfaces demanded her attention. The AI had simply worked. Invisibly. Perfectly.
This moment captures something profound happening in product design right now: the rise of invisible AI that augments rather than replaces human intelligence.
In a world drowning in flashy AI demos and intrusive chatbots, one London-based startup has quietly cracked the code on AI product design. Granola, the meeting notes app that’s achieved an almost unheard-of 70% weekly user retention, isn’t winning through spectacular features or viral marketing. Instead, it’s pioneering what its founder calls “invisible AI”: technology so seamlessly integrated that users forget they’re even using artificial intelligence.
As we move deeper into 2025, Granola’s approach offers a masterclass in how to design AI products that people actually want to use every day. And the lessons extend far beyond note-taking apps.
The Human-AI Collaboration Revolution
Chris Pedregal, Granola’s co-founder and CEO, has spent the last two years discovering what makes AI products sticky. His answer challenges the prevailing wisdom about AI interfaces: “Granola augments you, rather than replaces you. You remain in control, writing what matters during the meeting, while AI assists afterwards to enhance your notes.”
This philosophy: human control with AI enhancement, represents a fundamental shift in how we think about AI product design. While most companies race to automate everything, Granola has found power in restraint.
The product lets users take their own notes during meetings, and after it ends, uses AI to make those rough notes into well-organized, comprehensive summaries. But here’s the key insight from Pedregal: “With AI it’s very easy to build a demo version of a feature…but it’s actually still a lot of work to build a great feature that works consistently and reliably and that people love.”
This reflects a broader trend we’re seeing across successful AI products in 2025: the winners aren’t necessarily the most technically impressive, but the most thoughtfully designed.
The Invisible AI Design Framework
What makes Granola’s approach so effective? It follows what I call the “Invisible AI Framework”: four principles that are becoming essential for AI product design:
1. Preserve Human Agency. Unlike fully automated note-taking apps, Granola requires users to write during meetings. This isn’t a limitation, it’s a feature. Users stay engaged, and their personal insights remain at the center of the process.
2. Enhance, Don’t Replace. The AI works in the background during meetings, transcribing locally on your device without any meeting bots. After the call, it “fleshes out your notes and makes them great,” turning bullet points into structured summaries while preserving your original intent.
3. Build Context Over Time. Granola has evolved beyond single meeting notes to become what users call their “second brain.” The app now lets users chat with all their meeting history, creating a queryable knowledge base that gets more valuable with every interaction.
4. Design for Trust Through Transparency. Every AI-generated enhancement is hyperlinked to the original transcript, allowing users to verify and understand how the AI reached its conclusions.
Real-World Lessons from Granola’s Evolution
Granola’s product development story reads like a masterclass in iterative design. The team embraced what Pedregal calls “evolutionary product development”, rapidly testing variants with real users rather than building in isolation.
For a year before launch, they worked with 150 beta users, and “it took them a long time to figure out Granola’s ‘core interaction.’” Once they found the right approach, “they cut out half the features they’d built because they didn’t work as well as they needed to.”
This willingness to eliminate features that don’t serve the core purpose is becoming critical for AI product success. In a field where it’s easy to add impressive-sounding capabilities, the discipline to say “no” creates focused, useful products.
The Collaborative Intelligence Shift
Granola’s recent evolution into a team platform reveals another crucial trend: AI products are moving from individual tools to collaborative intelligence systems. With folders for “Sales Calls, Customer Feedback, Hiring Loops,” teams can now ask questions like “Why are we losing deals this quarter?” and get answers “with source-linked citations.”
This shift from personal AI assistants to organizational intelligence platforms represents the next frontier in AI product design. It’s not enough to make individuals more productive; the most valuable AI products will help entire teams think better together.
Design Principles for the Invisible AI Era
Based on Granola’s success and broader trends in AI product design, here are the key principles every product designer should consider:
Start with Human Workflows. Don’t begin with what AI can do; start with what humans need to accomplish. Granola succeeded because it understood that taking notes isn’t just about capturing information; it’s about staying engaged and processing thoughts in real-time.
Design for Gradual Trust. Building trust with AI products requires what Pedregal calls “predicting the future”: understanding where AI capabilities are heading and designing products that will remain relevant as models improve. Start with clear, verifiable outputs and gradually introduce more sophisticated features.
Embrace the Hybrid Interface. The future isn’t conversational AI replacing traditional interfaces; it’s AI seamlessly woven into familiar interaction patterns. Granola proves that the best AI products often feel like enhanced versions of tools we already understand.
Build for Context Accumulation. Design products that get more valuable over time. Granola’s transformation from note-taker to “second brain” shows how AI products can create defensible moats through accumulated user data and context.
The Future is Already Here
It’s the stickiest product I’ve ever used. I now rely on it heavily for note-taking during meetings. This sentiment, echoed by thousands of users, points to something profound: when AI disappears into the background and simply makes your existing work better, it becomes indispensable.
The companies building the next generation of AI products would do well to study Granola’s approach. Success won’t come from building the most impressive AI demos or the most sophisticated language models. It will come from understanding that the best AI products are the ones users forget are powered by AI at all.
As we stand at the inflection point of AI adoption, the question isn’t whether your product uses AI; it’s whether your users even notice. The future belongs to the products that make AI invisible.
What AI products have you encountered that truly disappear into your workflow? How do you think about balancing AI capabilities with human agency in your own product work?
References
- Granola official websiteHow to Build a Truly Useful AI Product — EveryGranola raises $20M Series AThe Secret to Building Sticky AI Products — EveryHow to evolve a product — Granola BlogGranola 2.0: A second brain for your teamAI notetaking app Granola raises $43M — TechCrunch
The Art of Invisible AI: What Granola’s 70% Retention Teaches Us About Product Design was originally published in UX Planet on Medium, where people are continuing the conversation by highlighting and responding to this story.
