Founder Playbook · Sub Club by RevenueCat
8 tactics from Luke Harris
How ElevenLabs Turns Feature Launches Into a Growth Engine
Watch the full episode“the heartbeat of the growth team is working hand-in-hand with our incredible product and research team such that when they're launching new features new models new capabilities we put it through this growth system and that normally translates into what's hopefully a really viral moment”
Every feature launch is a growth engine — not just a product update
ElevenLabs treats each model release as a coordinated growth event — not just a changelog. The moment engineering ships Scribe V2, the growth team has the tweet thread, the landing page refresh, and the paid ad variants ready in parallel. Most apps quietly push an in-app banner; treating the same engineering effort as a media moment is a free multiplier that almost every app leaves on the table.
“we built a custom GPT where we've analyzed what's all the top performing meta and Google copy and what's all our bottom performing meta and Google copy so we normally take what's our like brand X post and then we pass it through this custom GPT which will then iterate on it into like the high performing meta and Google copy”
Build a custom GPT trained on your best and worst ad copy
The ElevenLabs growth team built an internal GPT trained on their historic ad performance data — best and worst copy tagged. A launch post enters the GPT and exits as high-performing ad variants. The system makes institutional knowledge about what converts explicit and reusable, instead of locked in one person's head. Any team running paid ads can build this for a few hours of prompt engineering.
“we have two people which have the job title AI creative producers and their task is every day they're just obsessed they're using tools like motion they're obsessed with okay what's the data of what's working so far what are other companies in the space doing and then they're coming up with new creative ideas and then rapidly iterating and testing”
AI creative producers: motion design plus voiceover plus music for rapid variant testing
Rather than outsourcing creatives to agencies or relying purely on UGC, ElevenLabs hired two in-house 'AI creative producers' — motion designers who layer AI-generated voiceovers, music, and visuals to produce high-quality ad variants at scale. The combination keeps quality high while enabling rapid hook testing: swap the voiceover, change the opening 3 seconds, test a new music bed — all without new shoots.
“we actually ripped out all the localization infrastructure we built our own GitHub action and there's like a thin library which just extracts the strings or wraps all the strings and then each time we make a PR to our repo it just sends it with like a prompt per language about the localization guide and that means we're able to localize like way cheaper and much faster”
Replace expensive localization SaaS with a GitHub Action and LLM prompts
ElevenLabs was paying an agency and an expensive SaaS for localization until engineers noticed that ChatGPT consistently produced better translations. They replaced everything with a GitHub Action: on every PR, a thin wrapper extracts strings and sends them through a per-language LLM prompt. The result is faster, cheaper, and more natural-sounding copy — making full localization of ads and app store listings practical for any team.
“the important thing to do is to run the lift studies within the platform which actually doing AB tests with hold out groups to tell you because for a brand like 11 Labs where lots of people already know you you'll often end up with lots of people would have converted anyway and particularly as you start doing stuff like retargeting ads”
Run lift studies with holdout groups — not attribution reports — to know real ad impact
At scale, reported conversions are wildly inflated by people who would have converted without seeing an ad. ElevenLabs runs platform-level lift studies with holdout groups to measure true incrementality, then applies a quarterly adjustment to all channel attribution before making budget decisions. Any app spending into retargeting or branded keywords needs this discipline or it will dramatically overestimate paid channel ROI.
“when I think about scaling creative in order to get unlocked what you actually want to do is more segmentation so you're maybe using AI to help before maybe if you just had a general like here's how to make images and videos with AI but now you can go that one level deeper like here's how podcasters or here's how people making apps can create with image and video”
Segment creatives by persona to scale paid spend beyond a general message ceiling
Generic ads hit a ceiling quickly. The unlock at scale is persona-level creative: the same product feature gets a different hook for podcasters, a different one for app developers, a different one for musicians. AI makes producing these variants cheap; the strategic move is doing the segmentation work first, then commissioning the variants. This principle scales down — it applies even before reaching significant ad spend.
“our sense is probably each product will have a growth person who has like really great taste and really great ideas and then you work with the engineering team and you go okay I'm going to launch a product so say we're launching image and video creation in our app you then chat with an agent and you're like hey we're launching this product it then like literally checks through your codebase it checks through online like competitive landscape it reads through all your internal docs”
The future marketer is one person directing AI agents through the entire launch pipeline
Luke's prediction: within a couple of years, a single growth-minded person with taste will direct a suite of AI agents that handle the entire launch pipeline — competitive research, messaging, storyboard, video production, localization, ad variants, and performance monitoring. The job title shifts from 'marketer' to 'creative director of agents.' For solo founders today, this is already partially achievable by stitching existing tools together.
“we actually view yes we have the top down view but we're mainly thinking bottoms up and so we treat each channel as its own growth engine and actually our biggest one still is Google and Google search and so what we're doing is like that we're obsessed of okay how do we make really great ad copy how do we localize it to all the geos how do we have really high converting landing pages”
Treat each paid channel as its own bottoms-up growth engine with a specific bottleneck
Rather than allocating a $100M budget top-down across channels, ElevenLabs treats each channel as a standalone business with its own optimization loop. Google gets obsessive attention on copy, geo localization, and landing page CVR; Meta gets high creative volume plus quality control. The question is never 'what percentage goes to channel X' but 'what is the next specific bottleneck limiting this channel's 20% week-over-week growth?'