从「听过最糟的点子」到一亿美元 ARR：Gamma 如何用三十人做出盈利增长

Lee starts with Gamma's founding in 2020, when he pitched investors over Zoom at night from a corner of a London flat after his kids were asleep. On his third pitch of the first day, an investor called it the worst idea he had ever heard, pointing out that Gamma was going up against incumbents with massive distribution, and hung up. Lee says the comment stuck, and the team promised itself that growth would have to be central from the beginning.

The first big test came with the public beta in August 2022. Gamma won Product Hunt's product of the day, week and month, but signups spiked and then flattened. There was no strong word of mouth, so the team concluded the product would not grow on its own. Running low on runway with just over 12 people, they treated it as a bet-the-company moment.

They spent the next three to four months rebuilding onboarding so that AI was part of it and every new user hit something magical in the first 30 seconds. After the relaunch in late March 2023, signups went from a few hundred a day to a couple of thousand, then 5,000, 10,000 and 20,000 a day, with no marketing or ads. Lee says spending on ads instead would have been a trap that led to brute-forcing a fleeting version of product-market fit.

His advice is to think of early work as building a word-of-mouth machine, even in B2B, where you should be telling colleagues about the product. If a healthy share of leads is not arriving that way, he says, go back and ask why. Gamma's internal mantra for the first few minutes was to make it dead simple to create content and dead simple to share it.

Lee argues that onboarding and the product experience are intertwined, comparing them to a restaurant visit from the door to the bill. Inspired by Scott Belsky's idea of the first mile, he suggests treating new users with a slightly cynical eye: they are selfish, vain and lazy, and have no desire to learn a new tool. The question becomes what you can give them in 30 seconds that earns the next 30.

With only 12 people, Gamma could not rebuild everything, so it put all its energy into that first moment. Lee says this lens helps pull forward the single most magical thing about a product, which may be only one of many features. He borrows a consumer-advertising image: throw someone one egg and they can catch it, throw five and they drop them all.

Gamma's egg is "create a slide in seconds." The idea for the company came from Lee's consulting days, when he spent hours formatting Google Slides instead of working on content. AI was not in the original vision, though the spirit of making content effortless was, and he calls it a magical gift that fit the building blocks the team had already designed.

Lee says the first users, friends, will be kind and then not come back, so the real early growth came after the AI launch, when the first 10,000 users arrived in a short period. Gamma skipped Product Hunt for that launch and posted on Twitter with an intentionally clickbaity line: the most valuable skill in business was about to become obsolete. A comment from Paul Graham, saying that the thing a slide deck describes is surely more valuable than the slide, boosted engagement, and Lee admits luck played a part.

He uses the episode to define founder-led marketing as controlling your narrative and learning to break through noise with copywriting, storytelling and visuals. Lee says this also helps when working with influencers, because you know what a good hook looks like. Founders who write something that reads like an ad and hand it to a creator, in his experience, never succeed.

On finding time, Lee suggests starting small with a running note of things learned or observed, then blocking a few hours a week to turn them into posts, after testing the ideas on teammates. He posts different material on LinkedIn and Twitter: more tactical, contrarian and technical on Twitter, with metrics, and broader and more inspirational on LinkedIn. He writes first thing in the morning and last thing at night, and reviews which posts get bookmarked or shared.

Lee frames founder-led marketing as an exchange of value, in which you give away useful content and bank goodwill rather than money. Later you can spend some of that goodwill announcing your product. Lenny recommends the book *Nobody Wants to Read Your Sh*t* as a reminder of how high the bar is.

At around $10 million ARR, Lee felt the initial brand was holding Gamma back. It had been a placeholder created when the company was incorporated, with limited art direction, voice and tone. The team spent many months on a rethink with its internal creative director and the agency Smith & Diction, which has also worked with Perplexity.

The point, he says, was replicability. As Gamma scaled it needed to create content for social, performance ads and influencer assets, and it now tests thousands of pieces of creative every week. That is impossible without a cohesive brand DNA, and he says the rebrand cost far more time and money than he expected but was the right investment.

Lee says founders assume influencer marketing is turnkey, but he onboarded all of Gamma's first influencers himself, on calls where he explained what Gamma was and how to use it. The aim is to have them tell your story in their own voice, not read a script. He compares it to onboarding a customer: treat influencers as an extension of your team and they will come back to post again.

He thinks the usual approach of spending a big budget on five or six million-follower creators is usually wrong, since it reads like an ad. He favors thousands of micro-influencers whose audiences genuinely use the product. Early on this meant educators, who make slides every day and love to tell other teachers about tools that save time, in what Gamma calls echo chambers.

Payments were a few hundred to low thousands of dollars per creator. Lee used manual cold outreach at first, then the YC-backed platform 1stCollab for automated outbound, plus young agencies such as AKG Media in the UK. More than 50% of new subscriber growth is word of mouth, and he says every influencer push lifts it further, with roughly 1.5 additional users arriving for each one from the campaign. He explains the effect through trust and the halo effect that comes when a recommendation feels like it comes from a friend.

Gamma started with about 20 creators a month and moved to 50, then 100. Lee says about 90% of the reach comes from under 10% of creators and he could never predict which ones, so you have to cast a wide net. He and Lenny agree that a budget of $10,000 to $20,000 a month is best spread across dozens of creators instead of a single big one, then doubled down on the personas that work.

To make posting easy, Gamma open sourced its brand at a dedicated site with its voice, art direction and even the Midjourney settings it uses. Lee says LinkedIn conversion is four to five times higher than other platforms, though creators there can cost more, while Twitter has not mattered much for Gamma. Sponsored posts are usually labeled as partnerships or with #ad.

Lee rejects the idea that brand and performance marketing are at odds, saying brand marketing is performance marketing that is simply harder to attribute. Investing in brand strengthens paid work by giving it more creative to test. Lenny's heuristic is that you are under-invested in brand if you are limited in how many ideas you can try in performance marketing.

He says brand matters most for consumer and prosumer products, where the experience begins with an ad or billboard and should feel cohesive when the user lands on the site. In B2B, where someone hears about a tool from a colleague and goes through procurement, it may matter less. Gamma's growth, in Lenny's summary, is over 50% word of mouth, then influencer and social, then paid.

For paid marketing, Lee gives two rules. Don't invest until word of mouth is working, and don't let more than 50% of acquisitions come from paid, since that signals a broken core growth engine and a leaky bucket. Relying on paid means expecting CAC to climb, and it is hard to get off that treadmill once you are on it.

Lee calls prototype testing a huge unlock. Friends will say they love the product and never return, so he recruits people who are legitimately good prospective users but have no stake in Gamma's success. He tells them it is an alternative to PowerPoint, asks them to narrate their thoughts and watches them struggle, which he says makes the pain immediate. Gamma has done this for landing pages, onboarding, launches, export and sharing.

To scale, Gamma uses panel platforms such as Voicepanel (in which Lee is an angel investor) and UserTesting. The team can have an idea in the morning, run a full-scale study in the afternoon and review results together by evening or the next day. Designers who code, plus AI prototyping tools, help build the prototypes, and Lee says about 20 people typically run through each one.

Gamma also built a program for power users called the Gammaster Program, in a separate Slack workspace where they see early wireframes and prototypes. Lee says this suits features for people who already understand Gamma. Ideas are kept small at first, so many die early and cheaply.

Lee ties this to his Optimizely background and the maxim "strong opinions weakly held": you can test even without statistical significance. He cites an early example in which the team spent six months on two ideas in parallel, the presentation tool and a virtual office called The Lobby. They chose presentations because the virtual office would always be competing with working side by side in person, while the presentation tool inspired a million ideas.

Asked what else drove growth, Lee points to adaptability and a small, lean team. He says experimentation is core infrastructure, including testing across 20 models in production, because the temptation is to throw the most expensive model at every job. That discipline, he says, gave Gamma strong unit economics and healthy margins from the beginning.

He also notes an unusual investment: at one point about a quarter of the team was product designers. Lee argues that AI companies trying to invent new interfaces need to get the foundation right and think deeply about user problems. He says it is not a common level of investment for an early-stage startup, but the team felt it was right.

Lenny raises the criticism that Gamma is a thin GPT wrapper. Lee answers that being a wrapper on a single model or provider may offer limited value, but Gamma uses more than 20 models, orchestrates them, and keeps testing newer models against cheaper older ones. The goal is to maximize value for users in a way that is sustainable for the business.

His first piece of advice is to start with a problem you care about enough to work on for five to 10 years, instead of chasing shiny objects. For Gamma that mission is democratizing visual storytelling, so that people who are not good at making slides do not lose out on deals or favor. He adds that you should own the end-to-end workflow so the product becomes the default tool someone thinks of when they need a presentation.

Different steps need different models: building an outline, producing a first draft, editing, reviewing a whole deck, finding the right image. Lee says the leaderboard keeps changing, and the harder question is which model suits a given user, since a consultant needs something more structured and information-dense than an educator. Early on, Perplexity worked well for outlines and web search.

Gamma had no monetization at the March 2023 AI launch. New users received 400 credits, and when they ran out, AI use simply stopped, so Intercom filled up with people asking how to buy more. The team spent April on a quick pricing exercise, using a Van Westendorp survey for willingness to pay and a form of conjoint analysis to learn which features users valued.

They launched a single plan at roughly $20 a month at the end of May 2023, partly anchored by ChatGPT's price and by the goal of keeping things simple. Lee says they monitored whether it was economical and found healthy margins that could be reinvested in headcount and inference costs. Lenny notes the ChatGPT team picked its price with a similar survey.

Lee says two checkpoints indicate product-market fit: organic growth, and willingness to pay. Runway was low, so profitability was not optional, and within a couple of months Gamma reached $1 million in ARR and became profitable.

Gamma's internal mantra is "hire painfully slowly." Lee says revenue per employee or profitability are side effects rather than goals, since a target such as adding 100 people leads you to settle for good enough. The team spent a lot of effort getting the DNA of the first 10 to 12 employees right, and all of them are still at the company five years later. He sees that continuity as a defense against a leaky bucket on the team side.

He describes this as the age of the generalist: people who are spiky in some areas but can wear many hats in a flat organization, with contractors or specialist agencies for areas such as influencer marketing. His head of marketing, for example, ran a drone show in San Francisco end to end while also leading the team. Gamma's people leaders are all player-coaches who still do hands-on work, which he compares to a quarterback adjusting a play on the field. Lee adds that this is where they are today and may evolve as the team scales.

When you find someone exceptional, he says, bet big on them. A-players want more playing time and harder problems, so you keep the hiring bar high and give strong people more resources. He also says founders should ask whether this could be their life's work, since candidates can tell missionaries from mercenaries.

Closing the main conversation, Lee returns to the investor who hung up on him as a low point. He describes a founder's job as increasing your luck surface area, and one dimension of that is the people you surround yourself with. He credits his two co-founders, with whom he has worked for five years, as essential to where Gamma is now.