Jon Noronha
Cofounder and Chief Product Officer (Head of Product), Gamma
Where to hear and read about Jon Noronha1 episode or test
Jon Noronha is cofounder and Chief Product Officer of Gamma, an AI presentation tool that reached 50 million registered users and $50 million in ARR in about two years. The thing most worth learning from him is how to build a product whose shared output brings in new users, backed by pricing tested against real AI costs.
We want PowerPoint to feel broken after you've used Gamma for a bit.
How Gamma Grew to $50M in ARR in 2 Years
Jon Noronha: the story and the lessons
1Jon Noronha's path to founding Gamma
Noronha learned product work through experimentation at large scale. He started as a product manager on Bing image search at Microsoft, with hundreds of millions of users and heavy A/B testing. He then joined Optimizely after its Series A, when it grew from about 100 to almost 500 people.
He met his Gamma cofounders and much of the early team at Optimizely. He left Microsoft because he felt big companies were not for him.
Gamma began in 2020 as a remote-work idea with no AI. Cofounder Grant floated the idea at a London dinner in March 2020, just before lockdowns. It was a format between a presentation and a document, something you could send before a meeting or read after it.
The first fundraise rested on that future-of-work thesis. An early slogan, roughly presentations you'll actually want to sit through, did not stick.
By 2022 the beta had happy users but not enough to raise again. The team turned to AI only after the beta stalled. A few months before the AI launch they had been considering shutting down to raise money.
2How Noronha bet Gamma on an AI launch in 2023
Gamma launched with a deliberately provocative tweet and a polished video. In an "existential" mood, Grant wrote the most clickbait tweet the team could, claiming presentations were about to be obsolete. Day one brought only hundreds of likes. Three days later Paul Graham retweeted it, calling it stupid, and other VCs piled on, which drove views and signups.
Noronha privately gave the bet 30 to 40% odds, yet thought "how could it not work?" Graham's point was that presentations are not the most important part of business.
Gamma used an AI first draft to fix the blank-page problem. 90 to 95% of signups tried the blank editor for a few minutes and never came back. The bet was that the first five minutes could start with a rough draft instead.
Activation barely moved, but daily signups still climbed from 1,000 to 5,000 as the minority who activated told others.
Steady daily growth convinced Noronha more than the first spike. The first spike faded and the team feared the launch had failed. Then signups rose every day with surprising regularity.
He projected the line to about 50,000 signups a day within 3 months. He did not believe it, but the pattern showed something self-sustaining.
3How Noronha listened to early users and found product/market fit
Noronha emailed every new signup with a note that looked hand-written. An automation asked each new user for feedback on their experience. It drew one or two replies a week for a year, then 30 to 100 a day after the AI launch.
Answering them took a large part of his day. The frustrated tone meant users cared.
He missed that consultants, not product managers, were the real signal. Noronha over-weighted product managers like himself, who explained why they could not use Gamma internally. Consultants were already using it for external sales proposals and client pitches.
He wishes he had spent more time on the work behind each request than on the request itself. His lesson: know your three to five key segments and do not over-index on feedback that matches your own background.
He read product/market fit from things breaking under usage. Two to three weeks after launch he was still unsure. Then Intercom went from 3 messages a day to 50 and then 300, with no support hire, and the team spent its time answering instead of building.
He had found "you'll just know" a frustrating definition of fit, but it matched what happened.
4How Gamma priced and monetized AI
Gamma launched the AI with no way to pay for it. The team rushed the launch and skipped monetization. Credits capped AI use but users could not buy more, and more than half of incoming messages asked how to pay.
The first build was a Stripe integration for more credits or unlimited AI. Monetization launched in May 2023 and Gamma was cash flow positive by August.
Noronha A/B tested price and found $20 beat the $10 norm. He tested $8, $10, $12 and $20 against conversion and revenue. $20 won, because users understood AI has a cost and could justify it by the time saved.
"I was shocked to find out that 20 bucks won the test." Gamma launched Pro at that price.
Gamma built on cheap, fast models first, then reversed course. With thin runway and a need to deliver value in the first five minutes, the question was always whether a cheap, fast model could solve the problem. People assumed Gamma spent millions on AI, but it did not.
Gamma later moved off this approach because underpowered AI was underserving customers.
Gamma moved value into its $10 Plus plan after seeing retention. Plus launched limited, for fear of cannibalizing the $20 Pro plan, and buyers split one third Plus and two thirds Pro. Micro A/B tests on feature placement and price flipped it to about two thirds Plus.
The lower plan retained as well or better, partly because net retention lets low-plan users expand later.
Noronha chose simple tiers over usage-based pricing and accepts some losses. Gamma tracks AI cost per user and per plan and needs each plan to be profitable on average. Losing money at the 90th percentile of heavy users is fine if it is not outrageous.
The goal is ubiquity and the organic loop, not maximum revenue. Many other AI tools chose usage-based pricing and gave up simplicity.
Gamma customizes prices for only about 8 to 10 countries. Those countries get local-currency prices and everyone else pays a general dollar rate. Users paying in US dollars retained much worse, and local prices fixed that.
Some markets now retain better than the US. Noronha admits some people in unpriced countries are probably missing out, since custom pricing for every country needs more bandwidth than the team has.
5Gamma's growth loops: badges, export, publishing and localization
A badge on every output doubled as upsell and viral hook. After monetization launched, Gamma put a 'Made with Gamma' badge on everything created, and paying removes it. People still click badges on PDFs made two years ago.
Polished output was already getting people to ask what it was made with before any badge existed.
PDF and PowerPoint export built trust and spread the brand. Export let people work in Gamma without feeling locked in. Noronha calls them probably the two biggest features that unlocked growth.
Every exported file carries the badge and spreads as a casual contact loop, where nobody has to promote anything.
Gamma pairs each free viral feature with a paid premium version. Users turned pitch decks into websites, so Gamma built publishing on free gamma.site subdomains, with a custom domain as the paid option. The published pages carry the badge and rank in search.
Noronha calls SEO probably the next big loop and plans to feature user presentations and templates on Gamma's own site.
One prompt line, "write in Portuguese", opened international markets. Adding it to the end of the prompt let people in Brazil make perfect Portuguese decks, and Noronha says it 10xed the market overnight. 90 to 95% of free users and over three quarters of paying customers are outside the US.
Year one localized content, year two the interface with machine translation, year three finer work like right-to-left editing and small local ad campaigns.
6How Noronha plans to keep Gamma ahead and runs its culture
He wants lasting advantage from core primitives, not from AI alone. Noronha expects AI to help at every stage of making a presentation, until it is invisible and tools without it feel broken. The work ahead is visual building blocks and easy export, so AI has good tools to work with.
He pictures having the designer behind a Steve Jobs keynote for $20 a month, with no sparkle buttons on every action.
Gamma dogfoods its product daily and invents reasons to use it. Internal docs are written in Gamma, and every Friday one person gives a TED-style 'Gamarama' talk that the company critiques. Bugs reported in Slack get fixed about 20 minutes later because they bother staff too.
Noronha refused to found a company around a product the team could not use daily, which he says rules out about 98% of good products.
He names incumbents, startups and the models themselves as threats. PowerPoint and Google could build good AI, tiny startups could leapfrog Gamma, and models may get good enough to operate PowerPoint for you.
His answer is to try to stay ahead of all of it. The fast, loose, in-person culture also strains as the team scales, so the aim is to add structure without breaking the magic.
How Gamma Grew to $50M in ARR in 2 Years
Gamma's cofounder explains how a bet-the-company AI launch in early 2023 led to product/market fit, viral sharing loops, and a pricing model that covers AI costs. The company hit cash flow positive three months after launching monetization and grew to 50 million registered users and $50 million in ARR.
Noronha learned product work through experimentation at large scale. He started as a product manager on Bing image search at Microsoft, with hundreds of millions of users and heavy A/B testing. He then joined Optimizely after its Series A, when it grew from about 100 to almost 500 people.
He met his Gamma cofounders and much of the early team at Optimizely. He left Microsoft because he felt big companies were not for him.
Gamma began in 2020 as a remote-work idea with no AI. Cofounder Grant floated the idea at a London dinner in March 2020, just before lockdowns. It was a format between a presentation and a document, something you could send before a meeting or read after it.
The first fundraise rested on that future-of-work thesis. An early slogan, roughly presentations you'll actually want to sit through, did not stick.
By 2022 the beta had happy users but not enough to raise again. The team turned to AI only after the beta stalled. A few months before the AI launch they had been considering shutting down to raise money.
Gamma launched with a deliberately provocative tweet and a polished video. In an "existential" mood, Grant wrote the most clickbait tweet the team could, claiming presentations were about to be obsolete. Day one brought only hundreds of likes. Three days later Paul Graham retweeted it, calling it stupid, and other VCs piled on, which drove views and signups.
Noronha privately gave the bet 30 to 40% odds, yet thought "how could it not work?" Graham's point was that presentations are not the most important part of business.
Gamma used an AI first draft to fix the blank-page problem. 90 to 95% of signups tried the blank editor for a few minutes and never came back. The bet was that the first five minutes could start with a rough draft instead.
Activation barely moved, but daily signups still climbed from 1,000 to 5,000 as the minority who activated told others.
Steady daily growth convinced Noronha more than the first spike. The first spike faded and the team feared the launch had failed. Then signups rose every day with surprising regularity.
He projected the line to about 50,000 signups a day within 3 months. He did not believe it, but the pattern showed something self-sustaining.
Noronha emailed every new signup with a note that looked hand-written. An automation asked each new user for feedback on their experience. It drew one or two replies a week for a year, then 30 to 100 a day after the AI launch.
Answering them took a large part of his day. The frustrated tone meant users cared.
He missed that consultants, not product managers, were the real signal. Noronha over-weighted product managers like himself, who explained why they could not use Gamma internally. Consultants were already using it for external sales proposals and client pitches.
He wishes he had spent more time on the work behind each request than on the request itself. His lesson: know your three to five key segments and do not over-index on feedback that matches your own background.
He read product/market fit from things breaking under usage. Two to three weeks after launch he was still unsure. Then Intercom went from 3 messages a day to 50 and then 300, with no support hire, and the team spent its time answering instead of building.
He had found "you'll just know" a frustrating definition of fit, but it matched what happened.
Gamma launched the AI with no way to pay for it. The team rushed the launch and skipped monetization. Credits capped AI use but users could not buy more, and more than half of incoming messages asked how to pay.
The first build was a Stripe integration for more credits or unlimited AI. Monetization launched in May 2023 and Gamma was cash flow positive by August.
Noronha A/B tested price and found $20 beat the $10 norm. He tested $8, $10, $12 and $20 against conversion and revenue. $20 won, because users understood AI has a cost and could justify it by the time saved.
"I was shocked to find out that 20 bucks won the test." Gamma launched Pro at that price.
Gamma built on cheap, fast models first, then reversed course. With thin runway and a need to deliver value in the first five minutes, the question was always whether a cheap, fast model could solve the problem. People assumed Gamma spent millions on AI, but it did not.
Gamma later moved off this approach because underpowered AI was underserving customers.
Gamma moved value into its $10 Plus plan after seeing retention. Plus launched limited, for fear of cannibalizing the $20 Pro plan, and buyers split one third Plus and two thirds Pro. Micro A/B tests on feature placement and price flipped it to about two thirds Plus.
The lower plan retained as well or better, partly because net retention lets low-plan users expand later.
Noronha chose simple tiers over usage-based pricing and accepts some losses. Gamma tracks AI cost per user and per plan and needs each plan to be profitable on average. Losing money at the 90th percentile of heavy users is fine if it is not outrageous.
The goal is ubiquity and the organic loop, not maximum revenue. Many other AI tools chose usage-based pricing and gave up simplicity.
Gamma customizes prices for only about 8 to 10 countries. Those countries get local-currency prices and everyone else pays a general dollar rate. Users paying in US dollars retained much worse, and local prices fixed that.
Some markets now retain better than the US. Noronha admits some people in unpriced countries are probably missing out, since custom pricing for every country needs more bandwidth than the team has.
A badge on every output doubled as upsell and viral hook. After monetization launched, Gamma put a 'Made with Gamma' badge on everything created, and paying removes it. People still click badges on PDFs made two years ago.
Polished output was already getting people to ask what it was made with before any badge existed.
PDF and PowerPoint export built trust and spread the brand. Export let people work in Gamma without feeling locked in. Noronha calls them probably the two biggest features that unlocked growth.
Every exported file carries the badge and spreads as a casual contact loop, where nobody has to promote anything.
Gamma pairs each free viral feature with a paid premium version. Users turned pitch decks into websites, so Gamma built publishing on free gamma.site subdomains, with a custom domain as the paid option. The published pages carry the badge and rank in search.
Noronha calls SEO probably the next big loop and plans to feature user presentations and templates on Gamma's own site.
One prompt line, "write in Portuguese", opened international markets. Adding it to the end of the prompt let people in Brazil make perfect Portuguese decks, and Noronha says it 10xed the market overnight. 90 to 95% of free users and over three quarters of paying customers are outside the US.
Year one localized content, year two the interface with machine translation, year three finer work like right-to-left editing and small local ad campaigns.
He wants lasting advantage from core primitives, not from AI alone. Noronha expects AI to help at every stage of making a presentation, until it is invisible and tools without it feel broken. The work ahead is visual building blocks and easy export, so AI has good tools to work with.
He pictures having the designer behind a Steve Jobs keynote for $20 a month, with no sparkle buttons on every action.
Gamma dogfoods its product daily and invents reasons to use it. Internal docs are written in Gamma, and every Friday one person gives a TED-style 'Gamarama' talk that the company critiques. Bugs reported in Slack get fixed about 20 minutes later because they bother staff too.
Noronha refused to found a company around a product the team could not use daily, which he says rules out about 98% of good products.
He names incumbents, startups and the models themselves as threats. PowerPoint and Google could build good AI, tiny startups could leapfrog Gamma, and models may get good enough to operate PowerPoint for you.
His answer is to try to stay ahead of all of it. The fast, loose, in-person culture also strains as the team scales, so the aim is to add structure without breaking the magic.
Gamma