Close sidebar

Clear
Thank you! Your submission has been received!
Oops! Something went wrong while submitting the form.
Ep 01
Dylan Fox of AssemblyAI
Coming Soon
NOW LIVE
Ep 02
Daniel Sturman of Roblox
Coming Soon
Coming 10/24
Ep 03
Mike Murchison of Ada
Coming Soon
NOW LIVE
Ep 04
Shensi Ding of Merge
Coming Soon
NOW LIVE
Ep 05
Alexandr Wang of Scale AI
Coming Soon
NOW LIVE
Ep 06
Jack Krawczyk of Bard
Coming Soon
NOW LIVE
Ep 07
Victor Riparbelli of Synthesia
Coming Soon
NOW LIVE
Ep 08
Cai GoGwilt of Ironclad
Coming Soon
NOW LIVE
Ep 09
Daniel Yanisse of Checkr
Coming Soon
NOW LIVE
Ep 010
Glen Wise of Cinder
Coming Soon
NOW LIVE
Ep 011
Kate Parker of Transcend
Coming Soon
NOW LIVE
Ep 012
Rene Haas of Arm
Coming Soon
NOW LIVE
Ep 013
2024 AI Predictions
Coming Soon
NOW LIVE
Learn more about Season 01
Ep 01
Andrew Bialecki of Klaviyo
Coming Soon
Now Live
Ep 02
Vlad Magdalin of Webflow
Coming Soon
NOW LIVE
Ep 03
James Theuerkauf of Syrup
Coming Soon
NOW LIVE
Ep 04
Suresh Vasudevan of Sysdig
Coming Soon
NOW LIVE
Ep 05
George Kurtz of CrowdStrike
Coming Soon
Ep 06
Ivan Zhou and Amit Kumar of Accel
Coming Soon
COMING 5/2
Ep 07
Marcelo Lebre of Remote
Coming Soon
COMING 5/9
Ep 08
Jon Noronha of Gamma
Coming Soon
COMING 5/16
Ep 09
Barr Moses of Monte Carlo
Coming Soon
COMING 5/23
Ep 010
Alex Bovee of ConductorOne
Coming Soon
Ep 011
Sanjay Beri of Netskope
Coming Soon
now live
Ep 012
Jackie Burns Koven of Chainalysis
Coming Soon
Ep 013
Marc Lore of Wonder
Coming Soon
Ep 014
Jeff Shiner of 1Password
Coming Soon
COMING 6/27
Learn more about Season 02
Season 01 • Spotlight on AI
Season 02
Episode 08

Gamma’s Jon Noronha on how early-stage startups can challenge industry incumbents

A conversation with the Co-Founder of Gamma

In a landscape dominated by incumbent players, Gamma takes a bold approach: instead of trying to be better than the existing players, they aim to be an entirely different tech company. They've reimagined the very essence of presentation tools and the creative possibilities they offer. In this episode of Spotlight On, Gamma’s Co-Founder, Jon Noronha, delves into the team’s journey, from an early acquisition at a former startup to somewhat coincidentally building a product that dovetails perfectly with AI. He also unpacks the challenges and opportunities early-stage companies face navigating today’s landscape.

While still in the early innings, Gamma’s journey underscores the importance of staying nimble and ahead of the curve. They’ve learned to anticipate and adapt to advancements in AI models and how to prevent the common (yet often surprising!) misuse of artificial intelligence products. On the heels of their Series A round, the discussion offers advice for early-stage startup founders on overcoming hurdles in aligning a product with market needs, fundraising in different economic climates, and more.

Conversation Highlights:

00:00 - Introduction to Gamma

03:10 - The foundations Gamma had in place to take advantage of the AI wave

08:42 - Structuring early-stage teams to leverage AI UX designers

14:31 - Outpacing powerful incumbents by embracing uniqueness to earn market share

23:02 - Building an AI company with customer data sensitivity in mind

30:02 - Overcoming challenges of fundraising during boom times

32:08 - Reflecting on the importance of a lean early-stage team

Featured: Jon Noronha, Co-Founder at Gamma, and Vas Natarajan, Partner at Accel

Explore more episodes from this season:

Access Spotlight On Season 1 episodes here.

Read More

Vas Natarajan (00:07):

Welcome to Spotlight on, I'm your host Vas Natarajan and I'm here with John Noronha, the co-founder of Gamma. Thanks Jon for joining us. Gamma has become such a special business for a couple reasons. One, it's a company that we led the seed in series A of which is part of the news that we're here to convey. We're so excited to be continuing this journey with you guys, but it's also a business that was born right before the AI revolution and has since become very much an AI native company. And so we want to unpack that history and use, I think the Gamma narrative to help explain what's happening in AI and also help coach future founders on how to build a great successful AI native business. So we're so excited to have you here. Why don't you kick off and just tell us really simply, what does Gamma do?

Jon Noronha (00:55):

Well, it might help to start with actually the origin story then to tee into it. So where we started and the beginning point for all of this was reinvent PowerPoint. We had this idea to take this ubiquitous process, which is making slides. It's something that hundreds of millions of people every month do, but nobody likes doing. It's frustrating. It's a pain to format things. It's a pain to think of what you're going to say, structure your ideas, find the pictures even to present it. And so that was our goal all along. Let's reinvent PowerPoint and not just build a new editor for PowerPoint, like a new way to draw the same kind of boxes and rectangles, but actually rethink the medium itself and make a new format. So we were actually born in 2020 out of Covid and the pandemic and the lockdown we were all in and we thought PowerPoint was made for sitting around a room, pointing at a screen, and there'd be one person who would stand at the front and talk at everybody and everybody would pretend to pay attention, but mostly be in their own laptops or in their own world or whatever, dozing off.

(01:48): And so we thought everything about this doesn't make sense. PowerPoint was invented in, I think it was 1989, which by the way is the year I was born. So it's been around for a bit and a lot of the assumptions behind it have changed. And in Covid it was most obvious that this assumption of we are all in the same room pointing at a screen didn't make any sense anymore. And so we started as let's reinvent PowerPoint, particularly for remote work, but along the way, let's rethink a lot of the deeper assumptions of it. It has to be live at all. It has to be in person. It has to be a perfect 16 by nine rectangle. All of these things, it had to be a linear story. None of those necessarily make sense anymore. And so we had this idea of let's rethink it all and build a new way to present ideas.

(02:26): So 2020, we've started with this balance of let's reinvent this medium, let's rethink how the audience is receiving content, but more importantly, let's also rethink how creators are choosing how to present content. And we were born in a pre AI pre LLM world. That's right. 

Vas Natarajan (03:20):And so what were some of the technical choices we were making then? And then if I could have you fast forward to 2022, AI shows up on our doorstep, what were the technology choices that we pivoted to? 

Jon Noronha (03:10): Well, one funny thing is we were pre LLM boom, but we were not pre LLM. And so I distinctly remember using the very first version of GPT-3 in summer 2020 and evaluating could this be useful for what we do? And the answer was no. I tried, why not? Why not? The quality just wasn't there. So I was testing very simple things like just take a long written document and condense it to the key bullet points, which is a core thing that you do in making a presentation.

(03:25): It's take some long-winded idea and distill it down to the essence, and it did okay and sometimes it even wowed me, but it would wow me 25% of the time and mess up 75% of the time, and that's obviously not enough to build a consumer grade product out of. So it was a neat idea, but I just thought, okay,  let's focus on a whole different side of things, which is funny because there was this moment then in 2020, I guess it was 2022, 2023, where they really took off and the technology actually hadn't changed that much. It turns out that just going from that three to 3.5 upgrade was enough to totally change the equation. But going back to technology choices early on, it's ironic how some of the early technology choices that we made thinking that LLMs were irrelevant ended up actually paying off really well.

(04:11): And I would say one of the biggest tech choices that we made early on, well it was really a product choice actually, was we decided we wanted to merge the idea of document and presentation. So if you think typically a document, it's one kind of run on sequence of thousands and thousands of words all in a row, whereas a slide deck is a bunch of independent chunks and each of those chunks, each slide is this very sort of freeform canvas where you're dragging around boxes and putting little arrows between things and all of that. So presentation and document are these very different worlds that each play by their own rules, which is why it's a huge pain to create both, you've thought in writing mode and now you're going to slide mode or vice versa, and we're all just kind of reproducing the same stuff in different shapes.

(04:54): Our kind of core approach early on was what if we create something that combines the best elements of both of these, and so on the document side that meant a writing based interface, you're not actually dragging and dropping things, you're just typing in your words and using commands to pull in these more dynamic blocks like an image or an embed, take a lot of inspiration from tools like Notion that had kind of gone and reinvented the document as a core sort of building block of work, but then also pulling in these elements of presentations. So a kind of big beautiful present mode that zooms in this ability to create kind of easy layouts and visuals and drag them in lots of stuff around imagery. And so we'd lean into this sort of hybrid medium that almost had this aspect of a responsive website. You could make it big for a presentation, but you could also make it small for a document.

(05:39): And all of that was based on this conviction that communication is going more async and we want this sort of hybrid approach. Then LLMs come on and they're selling this opportunity to plug them in and help with these problems. Turns out large language models are really good at writing. They are good at writing tons of, and they're even okay at structured writing, meaning things like HTML or whatever it is. One thing they're terrible at though is dragging shapes around on a rectangle, so nobody's got LLMs to really draw very much. In fact, I remember early on experimenting with GPT-4 saying, draw me a pyramid like three stages with words inside and it could not do it. It absolutely stumped this genius level LLM that could pass the bar exam and the MCAT and all this stuff, but it can't draw but it can write.

(06:28): And so it turns out we had created this medium for turning all this structured writing into a beautiful dazzling presentation, and so it put us in this amazing position to actually exploit this new technology. Right when it came out before an incumbent like PowerPoint was in position to take advantage for the entrepreneurs that are listening, we at Gamma have become very multi-model in our approach. So early days we were playing around with GPT-3 that evolved to 3.5, obviously four is on the horizon. The technology itself I think became a lot more consumable. And so just the developer experience around it, the sophistication of the API, what we could pass in, pass back and forth with these systems has gotten a lot better. But there are many models out there each that have their own in some ways domain expertise. One model might be really great at certain functions or outcomes. Another model might be great at other things.

Vas Natarajan (07:22): And so how have you guys thought about this? Because I think a lot of founders are wondering, do we need to be orchestrating across multiple different models to get to the right output or can we build a business just thinking in a singular model way?

Jon Noronha (07:33): We certainly started singular model because the time that we got into this GPT-3 0.5 was kind of the thing, and so we built everything on one. And I have to say it was an easier starting point, not having to think about any of these questions. If I were advising a brand new founder, I would say don't dabble too much in testing 10 different models. Honestly, you can get so much out of a prompt engineering just one model that you should first figure out your core business around the model. And then once you've done that, iterate on the models.

(08:03): But for ourselves who are now sort of scaling an AI native company, the choice is starting to flip there where now we're having to put a lot of focus in it. And when I think about our success and the drivers of that success, a lot of it's about our ability to be at the forefront of what new models can do. We need to be in a position that when a great new model comes out, we can take advantage of it within a week, within two weeks, not just with local testing but actually rolling it out for new use cases. We need to really understand exactly what the frontier looks like of all of the features we are ready to build once models can do them, but haven't built yet because they can't. 

Vas Natarajan (08:42):

And so what does that imply about the team that we're putting together? Do we need some subset of our EPD resources tinkering with the new models that are out in the wild and almost thinking in sort of a labs and a research concept or have you designed our product organization and the way that you guys designed it at Optimizely or at other software companies?

Jon Noronha (08:51): One interesting thing about our organization was we are very UX heavy, so we've really leaned into UX design as sort of a core differentiator. So for a long time we were a team of 12 people and of those 12 four were UX designers, which I think is pretty much unheard of in startups at that scale to lean into it. But the reason we've done that is that we really believe that actually probably the single biggest differentiating skillset in terms of taking advantage of LLMs is sort of machine learning engineering, it's actually user experience at the application layer.

(09:32): It's that ability to deploy these things. And so it's not so much that we have a subset of our team tinkering. We actually try to have a superset of our team tinkering. We try to have lots of folks trying out LLMs in their day-to-day life including so-called non-technical folks who you might not think of as the initial deployers. And so often our designers are the ones writing prompts, they're the ones who are a bit out ahead trying to take advantage of these things. We try to encourage every single person on our team to use AI both in sort of producing outward facing AI product, but also in terms of how they get their own work done. And that's because I actually don't know if it's just about frontier model performance, it's about realizing all the things you can do with the models we've already got.

Vas Natarajan (10:14): I think one of the things I appreciate most about the team at Gamma, you aren't just leveraging the AI technology, you are building a software company. And I think many founders that we're meeting right now are just building AI technology and they're forgetting to build a software company. And what that implies is at the end of this whole equation, there is still an end user who has a job that wants to get to a certain outcome. 

Jon Noronha (10:22): We were talking about our product roadmap at our last board meeting and some of the things you guys were describing didn't have anything to do with the actual core AI. It was just traditional user workflow. Things like how do you get data in, how do you build charts, how do you express different complex concepts? And AI is going to help you with that, but you still have to build software around it.

(10:58): And I think that's to your point around having the right UX and design experience on our team to leverage the underlying technology but still recognize that there is a human at the end of this equation that needs to fulfill some sort of job. It's a real challenge I think, because the draw of just building cool AI stuff is so powerful and you can easily spend a full-time job just paying attention to all the new models coming out and all the different techniques of rag and fine tuning. And so even though we do try to get everyone involved in AI, we also try to tune it all out a bit and focus on the software engineering fundamentals. It probably does help us that we didn't start by conceiving ourselves as an AI company. We started by conceiving ourselves as solving a certain problem, and we got very lucky in that AI came along and was a great solution to those problems, but we really use this phrase powered by AI, so it's not just like AI for or whatever, it's there are certain things that we could not do before, but now we can do because of AI.

(11:56): One of the things we also talked about on our board, maybe we wanted your advice on is how much do we even lean into this AI hype? How much do we just say slap the word AI all over everything? And it's a delicate balance because as a founder you always want to be riding the wave. You want to be taking advantage of whatever novelty and talk is out there among people, and AI is the hot thing, but for many of our features, AI is almost invisible and we like it that way because we want it to be geared towards solving a specific problem, not just announcing a buzzword. 

Vas Natarajan (12:55): When you think about that concept of Gamma being powered by AI, I love that concept because at its core we're leveraging all this AI technology, but we're still building a lot of software around and you guys have made a lot of the hard choices to make sure that for core business professionals, they're able to take advantage of the technology but still be able to finish the product in their own way, tweak text, tweak graphics, present those concepts in the design language of their own business. A lot of our product choices have been about helping to deliver this business content over that last mile in some ways. Where do you think we're going to be in five to 10 years? Will all of that software still matter? How much of that outcome will be tucked fully behind the AI versus how much of it will we still need to deliver via really good software around ai?

Jon Noronha (13:31) This is one that keeps me up at night. I have this sort of nightmare that no matter how good we do UX and software development one of these days, a great AI agent's going to come along that just uses the computer for you. And so no matter what we do, it kind of becomes irrelevant because the AI is actually the one pushing the buttons and using the software. But so far, nothing on our current trajectory actually really tells me that we're going that way.

(13:46): So far what I see is that people want to be in the loop. AI itself seems to go off the rails and the cost of the models to stay on the rails is becoming sort of exorbitantly exponentially expensive. And all of our success so far seems to be coming from sort of empowering that human and almost tucking the AI away in the background versus making it the forefront. But with the pace of innovation in this area, I think we're all just watching and waiting to see what happens and we view our job as just constantly being aware of what that frontier is so we can adjust our strategy over time if the core capabilities of the models or the core behaviors of the market change. 

Vas Natarajan (14:28) And is there anything that we can do to prepare ourselves for that? 

Jon Noronha (14:31): I think the biggest thing we can do is just always to be first. If there's some new way to use AI to solve a problem, we don't want to find out about it because a competitor ships it. We want to be the very first ones to at least to try it, not necessarily to ship it. There's a lot of AI stuff that we try out internally and just throw away because it actually doesn't seem like a better user experience, but we need to know and we need to constantly be aware of what's out there. Anyone starting an AI company today is logically considering the responsive incumbents. And if you're in the HR tech space as an example, you want to build an HR app that's powered by AI. Well, there's Workday, there's Greenhouse, there's Bamboo, there are any number of HR technology companies that already have an existing built-in distribution base that can logically turn on AI. How is a new founder to a space with a lot of well entrenched incumbents that have the power of an existing install base, tons and tons of distribution.

(15:23): And in our world, obviously that's PowerPoint to a lesser extent Google Slides, but there are any number of other players that have come before us in the presentation game. 

Vas Natarajan (15:28): How did you guys consider that and what are some of the chess moves that you have contemplated making to make sure that we still have a seat at the table in spite of their incumbent install base advantage? 

Jon Noronha (15:34): When we were thinking about starting this company, there was a lot that I liked about it including really about the space, the breadth of usage of tools like PowerPoint and the dissatisfaction with them. It felt like a juicy recipe for a startup, but the biggest thing that gave me pause is that presentations as a space has been just a graveyard of startups. We have not seen a lot of companies really take off in this space. I would say the last successful company in the space was PowerPoint before it got bought, which as I mentioned was a while ago now.

(16:10): And so that worried us. It seems like it's very hard to break in even before AI comes along. There's a massive distribution advantage for an incumbent like office. It's bundled as part of a suite. People standardize on their templates and themes and as we sort of built the company, and especially before we had a lot of our traction take off with AI, we really felt like we were running into this brick wall of how do you break through the incumbent? And so I think it's actually a really important question to ask yourself early on. I actually wish we had spent more time thinking and talking about this when we were first starting the company, but all that said, I don't think it's a reason to not enter this space or not take on the incumbents. I think every great startup probably was born by taking on an incumbent in some form, but at least for us, and I think this is generally a good strategy, you don't usually win by going head on and building the same thing.

(17:00): You win by building something that's a little bit different. And if you think of sort of a classic disruption theory, building something that's almost unappealing to the incumbent in the first place. For us, the way we phrased this in our very first pitch deck was we wanted to build the anti PowerPoint, not the better PowerPoint. So for us, the anti PowerPoint was we wanted to almost invert a lot of the assumptions the PowerPoint is based on. And by doing that, I think we intrinsically made it hard to follow. The thing that gives PowerPoint its strength is kind of the standardization of this 16 by nine linear slide format. Literally the Pptx format is just the standardized thing that office has to speak. We wanted to build our own different format. We wanted to sort of break all those assumptions and in doing so, we make it harder for any incumbent to compete with what we're doing and also unappealing.

(17:50): They can't really make sense of us. They don't really see why we necessarily make the choices we do because we don't really fit into their headspace. A good example is this idea of melding document and presentation. One simple feature that we built very early on is when you're building a slide in Gamma, it can be any height you want. It doesn't have to be 16 by nine. It just grows taller and taller and taller. And if you work at PowerPoint, you think that's really weird. I think you're like people are going to be scrolling in their presentations. People will never want to deal with this. That sounds awful. We were thinking about responsiveness. We were thinking about something that could actually scale directly to viewing on a phone. And again, this all started to be useful when LLMs came along because it turns out when you had these more scalable slides, it was way easier for an AI to fill them out than it is for a tool like PowerPoint.

(18:34): So taking it back kind of the more general answer, I think it's all about finding your unique lens on the situation and leaning into Differentness. When I look back on our early years, the decisions I'm grateful for are the ones where we tried to be different, not the ones where we tried to be the same. And I think the more you can do that, the more you can sort of build your own seed before any incumbent really notices.

Vas Natarajan (18:44) Let’s talk about a thorny issue that we at Gamma face, but other consumer AI companies face too, which is how our technology is used not just for good but sometimes for bad and the way in which one can run afoul of our terms of service or just what is right, what is ethical. 

Jon Noronha (18:49): As we maybe touching upon the very broad top of funnel that we have, this is technology that is leverageable by people from all walks of life and we have seen Gamma used in some nefarious ways. This is a new phenomenon for us. 

Vas Natarajan (19:00): How are you starting to think about just putting the guardrails in place to make sure that our technology is used for the right purpose? And how would you coach other maybe consumer AI companies who are also going to run into trust and safety issues broadly in the future? What are decisions that they should be front running that you wish maybe we had made earlier? 

Jon Noronha (19:52): One of the biggest challenges is even knowing how your product is being used. And so we opened the floodgates of our product with AI and just saw this true hockey stick of growth. And what was really surprising about that hockey stick was that it wasn't just people in say the US or in Silicon Valley that were using it. It was incredibly international. For a long time, probably less than 10% of our signups were coming from the US and many of them were using gamma in languages that nobody on our team spoke for.

(20:26): Things that we could barely understand aside from copy pasting stuff into Google Translate and making our best sense of it. That's actually been great for growth. It's great for top of funnel, but it's also bad because you start to lose a grip on like, well, what are people even doing with our product? And one thing we realized is that not everything they were doing with our product was good, particularly when you have great AI content generation. A simple example of that is we've seen so many Gammas created to sell CBD pills online or whatever it is. And so we honestly, as a small startup, we're not going to focus on this. We are just trying to build our core business. We have a million requests from our core users of what we need to do. Why should we worry about what a few bad people are doing with our product?

(21:11): It's just not the right priority. That turned out to be a mistake. And maybe it's even obvious now when I say why it's a mistake, it hurts reputation to let these things go out there. It's also just not the right thing to do. I think every software creator has a responsibility for how their software gets used and it's now come back to bite us in a few areas. So we've seen cases where Gamma as a whole will get blocked by an ISP because some nefarious actor used it for something like phishing or even just for spammy CBD pills or whatever it is. And so we have now had to actually mount a real defense against this, and I wish we had done that sooner. I wish we had sort of recognized that whatever you see is a small problem now is just obviously going to grow and turn into a big problem.

Vas Natarajan (21:54): Are there some really simple wins that a founder can think about on the front end as it relates to gating user signups or figuring out different security protocols to put into place? There are, I would say some easy wins. So just basic things like a capcha on your signup page. It turns out we had a problem where bots were signing up and using Gamma to send emails to people as spam. So validation of your form fields. One really stupid one that we realized was we never validated that your last name in Gamma was actually a last name. And so people were putting a 500 word spam email into their last name with a bunch of links in it and using it to send emails. So some of this is like engineering 101 that you just don't think of when you are just trying to scale your startup and get it off the ground.

(22:40): A cooler one that I'm excited about, it's not quite 101, maybe 201 level, is we're actually now using AI to review Gammas that are created and decide if they meet our terms of service or could potentially violate it. And when they do, we're building a review process to actually try and take those things down or do something about them. And so this is a case where we're actually trying to once again deploy new technology in maybe an unfair way that's hard for other companies to do.

Vas Natarajan (23:02): How do you think about all the business users out there that are communicating complex business logic? They're putting sensitive data into your system, they're talking about revenue and user numbers, and it's going to be sensitive corporate content overall. And as an AI company, one of the big benefits is you get to take all of that data and effectively train on it to make the system better and better.

(23:31): And so there's just this amazing data network effect that AI specific businesses have that maybe traditional software companies don't have. And how are you thinking about that cut line of data and user interactions that you want to learn from, that you want to leverage to make the system better for everyone else while still making sure that you're honoring the potential sensitivity of what they're using Gamma for?

Jon Noronha (24:00): We have erred on the side of being conservative, so we have never trained on anyone's data. I understand why other startups are taking the opposite choice of trying to build that data flywheel because it is a core one. But I think our hunch is that because we know we want to grow into more and more professional and serious use cases, we don't want ourselves to be sort of hooked on using data in a way that if our customers knew about it, they would be upset about.

(24:25): That's a very dangerous position to put yourselves in. And so we have avoided it, and I think that doesn't mean you can't learn from what people are doing. So I would say overall, from an AI strategy perspective, we have relied less on fine tuning and more on prompt engineering. So what we try to do is we try to look at failure cases and say, how can we improve our core prompts for everyone based on that? But what we're not leaning into as much is let's actually train on a huge number of things these to produce new models.

Vas Natarajan (25:00):  I love that distinction of focusing on prompt engineering over fine tuning. You can focus your time and energy on improving the inputs or you can focus your energy on improving the core machinery. We're choosing to improve the inputs and over time the machinery is just getting better and better as these models. improve.

Jon Noronha (25:11): That's right. Yeah, exactly right.

Vas Natarajan (25:13): So walk us through the post seed era of Gamma. We were well funded. We had a great team. We were trying to figure out product market fit in that same period of time. The energy around productivity tools and remote work from the Covid era was starting to die down a bit and we were just overall, I think searching for our identity. What was that like? 

Jon Noronha (25:55): It started out really fun and then it got kind of hard and stressful. So like I mentioned, we had roughly 12 people for most of this time. We launched our beta on product hunt over that summer, I guess it was summer 2022, and we started sort of growing our user base. We actually were lucky to be number one product of the day on product hunt, which drove all of our early usage, but we're talking about on the order of a couple thousand people signing up.

(26:07): And that boiling down to maybe on the order of hundreds of weekly active users so enough to see that there was clearly something here that there was real value people were getting and some of them were becoming super fans. They were raving about it and loving it, but there just weren't that many of them and growing, but growing linear. So more people are using us over time, but not exponentially, not the kind of thing where it's clearly a fire that is taking off. And around that time, our runway was starting to get low, so I would say by summer 2022, we had maybe a year, year and a half of runway left. So not an emergency, but you're just starting to play that forward and you're like, oh, well, you want to ideally raise funds when you have a year of runway left so you're not cutting it too close.

(26:54): That means in the next six months we need to be in the position to raise a series A. Also, we raised a seed round in the boom times, and so we can't just raise any old series A. We kind of have to raise at least as good as we had before, which for almost any investor, and I remember we even talked to you at the time and you're like, just so you guys know, this is not going to be a walk in the park. You need to be really gearing up for battle here. You need to be really, this won't come so easily as the last rounds came. And so it took this process that had been really fun of building this company, but if I'm being honest, lacked a sense of urgency and suddenly cranked up that sense of urgency to the max. We really realized that we had this narrowing window of time where we could take off.

Vas Natarajan (27:37): And can I ask you, Jon, in that moment, you're contemplating raising more funds for the company, but at the same time you had brought all these great ex-coworkers of yours into this business on the promise of this next generation future productivity SaaS company. And that hype, that energy, that curve was maybe starting to flatten a bit. And how are you guys balancing both keeping the energy and the enthusiasm alive for the investor community who you need to motivate to put more money to the business and almost as critically, if not more consequentially, keeping that energy up for the team who's probably starting to realize, okay, wow, we haven't quite taken off in the way that we thought we were going to take off. 

Jon Noronha (28:30): Yeah, it's a really delicate balance. It's like I remember this feeling of with the team having to convey both we're doing really good and also we're not doing really good.

(28:34): Basically, this is great. We finally have traction. Let's be motivated and excited that people are using our product. But also this level of good is we actually think would've cut it for a series A two years earlier but didn't cut it. It actually really helped. The macro environment had changed because we really kind of put all of our team communications in that lens. The world has changed around us, and so it actually doesn't matter that we are doing better than we were doing for the team. It was almost jarring because if they compared the progress we've made compared to we raised the seed, it was night and day. We had been just a really crappy beta at the time we raised the seed, whereas now we finally had a product we were using internally that we actually thought was good and we had real external customer love for the first time.

(29:18): And so from that perspective, from the team, it was actually really great progress and everything was good. We almost needed to crank up the urgency with them and say, it's actually not as good as you think, because even though, yes, we're getting customer love, we're not seeing these numbers.

Vas Natarajan (29:50): How did they react to that?

Jon Noronha (29:52):  I think overall it was a motivator, but I think we really struggled with how to convey that urgency without scaring people. Because of that, I think we did get the team to really take on a sense of urgency. And honestly, we got really lucky that our incredible sense of urgency coincided with this chat GPT moment where all of a sudden a new market opportunity just blew up in our face, and so we seized it with full force. The ultimate test of the medal of a company and of the culture of a company is not in the great times.

(30:00): It's in the tricky times. And how do your employees rise to the moment in those crucible eras of a business? And I was so impressed by in the 20, 22, 20, 23 times where we really had to focus in to try to get product market fit and get this business to ignite. 

Vas Natarajan (30:02): That brings me to our series A, and one of the things I've always appreciated about Gamma is how you endeavor to do a lot with the little, and that manifested in this fundraising process where there was so much more money that you could have raised that you didn't raise. And I'm curious, walk us through that decision process.

(30:43): How is this series A going to be the right amount of capital for Gamma today? How does it let you focus in on the things that you want to focus in on? And conversely, how do you make sure that you don't do too many things in parallel such that you take your eye off the ball in terms of what's important? 

Jon Norohna (30:56): There was a point a little over a year ago where we would've taken any amount of money from anybody. We were just in survival mode and looking for a lifeline. It was nice to be in a position when we actually did raise this series A where we were not so desperate, we had seen this huge acceleration of growth from AI, we started monetizing and we've actually been cashflow positive for most of the past several months. And so we were actually at a point where we didn't need the money anymore.

(31:27): And the advice I have heard a hundred times from founders is always raise money when you don't need it because you're in a position of strength. And so we were lucky to do that. We did have people pounding down our doors and because we'd been so recently in a time when we didn't have that, it was just so obvious to me what a different feeling it is. But it made me realize the money is not the thing that actually makes you successful. The money is ultimately a byproduct of your success. And if anything, we've seen both in our own experience and just in our space and other companies, companies that are almost killed by the amount of cash that they raise, they raise so much that it destroys their ability to focus. I think we had seen firsthand what a strong sense of urgency we got from not having too much money.

(32:08): I mean, we were going two or three times faster than we were when we had plenty of money, I'll just say that. And so we never want to lose that again. Now that we've tasted what it can feel like to be so lean and we really don't want below, we don't want to get to a point where you are hiring people just for the sake of hitting a number target. And especially the thing we've seen over and over again is you hire a bunch of people and just lay them all off again, and you're constantly doing this expanding and contracting thing that is hugely wasteful of everyone's time and a lot of money as well. And so we really have taken leanness and speed as a core, and that drove our fundraising strategy, which is raise enough money that we don't have to worry about money and we can attack all the new markets we want to, but no more than that.

(32:48): And we were excited to partner with you guys to be able to do that for us. In terms of roadmap, what I think that means is we still can't do everything because we didn't raise enough to hire a hundred person team. We don't want to actually grow our team by a ton. We want to keep that team small and focused, and we want to just take on the biggest opportunities. And for us, what that means is there's still a huge amount to do in presentations, but there's also this new opportunity when pursuing in websites. We just found this organic poll that many people who use Gamma as a hybrid document presentation realized they'd actually made a webpage and wanted to put it up online. And so we've been trying to really lean into those organic sources of poll, but not try to invent too many new ones that distract us from our core mission.

Vas Natarajan (33:30): Well, I couldn't be more thrilled to be working with you guys to join the board. Gamma, to me, is one of the canonical companies in general business productivity today, and there is such an enormous opportunity ahead of us and such an important and viable mission for us to fulfill. It's going to be exciting to watch this company grow over the next five to 10 years and can't wait to see all that's ahead of us. And John, thanks for walking us through these complex times, walking us through the story of Gamma and helping us to just understand this dawn of AI through the lens of our business. 

Meet your host

Partner
Focus
Cloud/SaaS, Enterprise IT, AI

Vas Natarajan

Vas Natarajan is a Partner at Accel focused on early-stage companies in the data infrastructure, API, and enterprise software spaces. Discover more.
Read more on Accel.com
Season 1 of the Spotlight On podcast, by Accel
Artificial Intelligence

Artificial Intelligence

Early Stage

Early Stage

Cloud/SaaS

Cloud/SaaS