This AI note-taking startup thinks it’s building the ‘steering wheel’ for chatbots

Early drivers steered cars by pushing a lever left and right. That was fine at slow speeds, but disastrous when you accelerated. It took years before the steering wheel arrived. Granola CEO Chris Pedregal says AI interfaces are still in the lever era. 

Pedregal, who in 2019 sold the edtech startup Socratic to Google, says we’re just beginning to figure out how humans should interact with AI. Three years after the launch of ChatGPT, people still associate AI with typing into a chat box. 

Granola is betting on a new approach to AI-enhanced note-taking. The London-based startup doesn’t record audio or video or send bots into your meetings. Instead, its tool sits on your computer or phone, transcribing in real time while you maintain control.

You can jot notes alongside its transcription, building a personal knowledge base instead of a raw archive of recordings. The viral spread of its tool helped the company raise $43 million last year, bringing its total funding to $67 million at a valuation of $250 million. It’s also grown from a team of 4 to 35. 

Fast Company spoke with Pedregal about the “steering wheel moment” still ahead for AI interfaces and the surprising ways people are using Granola to take notes on everything from therapy to vet visits. The conversation has been edited for length and clarity.

You’ve described Granola as a steering wheel for large language models. What do you mean by that?

I think it’s very, very early days in this new wave of AI, particularly on the user interface interaction side of things. The technology developed very quickly, but it takes human time to figure out the right interaction patterns. 

I looked it up—it was over three years from when the iPhone came out to when Instagram launched. I think we’re in that time period right now. People might be like, ChatGPT has been out for three years, and we’re still just dealing with chatbots. Is that the end of it? I think it’s just early days.

Early cars were driven with a lever, literally a stick that you’d move left and right to steer. It was fine if you’re going slowly, but the moment you started going quickly, it was easy to go off the road. It took quite a while for them to develop the steering wheel. Once they figured out the steering wheel, it became very natural and it stuck. AI interfaces are still in their lever era.

Granola deliberately doesn’t record audio. Why make that choice when competitors do?

Granola doesn’t record audio by design, which is probably annoying if you ever try to use Granola for interviews. But it makes it less invasive for work conversations, because really what you want are the notes. The goal is not to have an audio recording.

The way I think about it is: What’s the minimum amount of invasiveness for the most value? That’s how you have to thread the needle. AI is here, we’re all going to be using tools like this in the future because they’re so useful. But what are the norms? What’s the thoughtful, ethical design of these tools so that we maximize the gains for the cost?

How is Granola different from Otter, Fathom, and other meeting notetakers?

It all really comes down to this: Granola feels like a tool that lets you be your best self in meetings. The operative word there is tool, and that means you control it. You can write your own notes. When the AI generates notes, you can edit them. The AI is subservient, augmenting your abilities. It’s your personal place where you have all this information.

I think a lot of the other tools—Otter is like 9 years old at this point—are really about meeting capture, meeting recording. You log in and here’s all your meeting recordings. That’s useful, but it feels very different than when you open Granola. It’s like, here’s my personal context where I can ask questions. It’s not really about the meetings. It’s about the notes, the knowledge inside of it.

As we look towards the future, Granola and those other tools are going to look more and more different. I see Granola as being much more of a contextual workspace where Granola has all this helpful context about you. Now if I need to go write an article or a blog post, or institute some process changes inside the company, I will go into Granola and write that first draft because it has all that context. I can’t imagine doing that in Otter or Fireflies—it just doesn’t feel like the right place for it.

You’ve found that mixing work and personal contexts in Granola is actually more useful. Why?

Right now, I use Granola for all my work meetings, therapy sessions, and logistics conversations about my life. If you had sat me down two years ago and asked if that’s really useful, I would have said no—I want those things separate. It turns out when you’re asking Granola questions, it having a 360-degree view of different things that are going on in your life is very useful. When you’re making decisions, you’re weighing all those constraints and priorities—not just the ones tied to this specific project at work. 

I was just at the vet this morning, and I used Granola because it’s my mom’s cat and I’m not going to remember exactly what the vet says. Moments like going to the doctor, parent-teacher conferences, talking to a construction worker or plumber—any situation where there’s sometimes technical language that’s really important to get right, that you’re not familiar with—are incredibly valuable to capture accurately.

There’s a different question around data ownership. I don’t necessarily want my company to have my therapy notes. But as models get better, the AI having access to the right context makes all the difference in terms of the quality of the response.

What’s appropriate etiquette around recording conversations with Granola?

I think right now, the etiquette is simple: Ask.

I imagine that the norms around this will change quickly, but it will remain very situation-dependent. Inside our company, it’s expected that meetings are Granola’d unless someone asks not to be. But in social environments the norms will be very different. I’ve tried some of these pendants that record everything, and the idea of wearing those at a party just makes me feel a bit icky.

I also think the video conference providers will adapt and make it easy to show meeting participants that you are using something like Granola, so you won’t have to think about it. I think it really comes down to the social nuances of the situation.

I usually frame it simply: Talk about it in terms of notes and transcription. “Is it okay if I take notes? This thing will transcribe so I don’t forget the important stuff you say.” That’s basically what I say.

What’s been your biggest mistake as you’ve grown Granola?

The biggest mistake I’ve made so far was that we didn’t grow the team fast enough. We had product-market fit in a fast-moving space, and I didn’t recognize that early enough. By the time I did, we were drowning in user tickets, requests for billing—all the kinds of stuff that happens when you grow. There were only four of us on the team when we launched the product.

I was trying to use my playbook from my last startup—keep the team super small, grow slow and steady. [I realized] that’s great, Chris, but actually the world wants this and you have to respond. I thought growing quickly meant sacrificing how thoughtful we could be about product, and I wasn’t reactive enough. We’re 35 people now, and most of that has happened in the last couple of months.

What are some surprising ways people are using Granola?

All the personal stuff was surprising at first—therapy, vet visits, parent-teacher conferences. Then there were these founders early on who used Granola as their collective brain. They logged in with the same account and would record every conversation they had because they were early in their startup—every brainstorm, every argument. It became a single shared memory between the two of them.

One user followed this famous sales methodology where every conversation falls into one of 14 buckets. He created very specific templates for each bucket, and at the end of the meeting he’d select the right one. Granola would basically spit out all the next steps to win that deal based on that framework. He encoded his entire sales process into it—super intricate. I didn’t see that coming.