20Growth: The $6.6B Growth Engine Behind ElevenLabs | Why ElevenLabs Do Not Have PMs | The 7 Part Launch Playbook to Crush All Launches with Luke Harries, Head of Growth @ ElevenLabs

Luke shares how he met ElevenLabs co-founder Matti at a Cambridge hackathon at age 19, initially declined to invest in ElevenLabs thinking their horizontal go-to-market was terrible, and eventually joined as Head of Growth after they hit 1M users. Includes lessons from his previous startup Fella (men's health/GLP-1) which now does $30M+ ARR.

• •Do hackathons with your smartest friends - Luke's led to his first Microsoft job and meeting ElevenLabs co-founder
• •Initially rejected investing in ElevenLabs due to horizontal 'sell to everyone' strategy - admits he was completely wrong
• •Key lesson from Fella: commit long-term and be patient - company hit $30M ARR after he left, proving market timing matters
• •Fella was early to GLP-1/Ozempic market but Luke left too early at $300K revenue due to impatience

Luke explains how ElevenLabs successfully operates a horizontal product strategy (consumer apps, creator tools, developer APIs, enterprise) by sharding the company into discrete teams with dedicated growth functions. Each product line has its own growth team with specialists, contradicting conventional wisdom about tight ICP focus.

• •Horizontal strategy works because they started with world-class AI models that attract users naturally, then raised capital to hire 'founder-type' people to own each product
• •Company is sharded with separate growth teams: consumer app team, creator growth, developer marketing, enterprise (20 people by end of year)
• •Horizontal growth team provides channel specialists (ex-Shopify performance marketing, ex-Canva SEO) who support product-specific growth leads
• •Each product has a 'CMO-like' product growth lead responsible for all metrics, activation, and awareness for their vertical

Detailed framework for scaling growth teams from 3 people to 50+. Luke recommends starting with one generalist growth marketer who understands the product deeply, followed by a front-end focused growth engineer. Emphasizes motion designers as underrated hires for creating launch videos that drive 200-700K views.

• •First hire: generalist growth marketer responsible for everything from positioning to channel testing - must deeply understand your product (technical for technical products)
• •Second hire: front-end leaning growth engineer for landing pages, mini tools, automated outreach, and rapid experimentation
• •Motion designers are critical but underrated - every major launch's keystone asset is a motion design video, not founder monologues
• •Video types: motion design (abstract, value-prop focused), founder-led (risky, can be ego-driven), screen share (works for technical audiences)

ElevenLabs' systematic approach to product launches using a three-tier system. Tier 1 launches (major products/models) follow a rigorous checklist: define audience and KPIs, nail core messaging, create tweet thread first, produce motion design video, then distribute everywhere with team amplification.

• •Three launch tiers: Tier 1 (major products, full playbook), Tier 2 (significant features), Tier 3 (changelog only)
• •Start with messaging: primary value prop (most accurate, fastest, cheapest - pick ONE primary) plus 2-3 secondary props
• •Always create tweet thread FIRST, then video that corresponds - thread sets the messaging foundation for all other assets
• •Distribution: cross-post everywhere (X, LinkedIn, BlueSky, Threads, Product Hunt, Reddit, Hacker News) with internal Slack amplification channel for team sharing
• •First 30 seconds of video are critical - get straight to core value prop, don't waste time on founder walking down staircases
• •Use 'Introducing' or 'Launching' as first word in tweets to signal clearly this is a launch
• •Blog posts still matter for technical audiences (show benchmarks, secret sauce) and SEO backlinks

Luke argues SEO isn't dead but is evolving. While long-form blog content will decline as LLMs scrape it, interactive tool pages (text-to-speech demos, mini product experiences) will remain valuable for 5+ years. Requires front-end growth engineers to build these experiences.

• •Blog-style SEO content is dying (LLMs will scrape and regurgitate) but tool pages will survive 5+ years
• •Tool pages require actual engineering work and proprietary data - example: text-to-speech Spanish with interactive text box ranks #1
• •Zapier still gets 70%+ of SEO traffic from blog content, so it's not dead yet despite predictions
• •Strategy: expose small bits of product value in mini tools outside login wall so users can experience 'wow moment' quickly
• •Hire front-end focused growth engineer early to independently build these tool pages and SEO experiences

Contrarian take on channels: organic LinkedIn is the most underappreciated growth channel because the content bar is so low compared to Twitter. LinkedIn ads are overpriced and polluted. Also covers why founders should shamelessly DM their entire network for launch amplification.

• •Most underappreciated channel: organic LinkedIn - competing against 'Freddie from Deloitte's promotion' vs Twitter's best writers
• •Most polluted channel: LinkedIn ads - high CPMs, tough ROI for enterprise products with long sales cycles
• •Don't be afraid to manually DM your entire Gmail contact list (3000+ people), Twitter followers, LinkedIn connections for launch boosts
• •Cross-post everywhere even if not primary channel - different audiences prefer different platforms, less competition on emerging ones
• •X targets creators/developers, LinkedIn targets future employees/partners/enterprises - write content that caters to both

Radical organizational structure: ElevenLabs has zero PMs. Instead, engineers own the entire product (they are 'product engineers') and partner with growth leads who handle marketing. This works because AI enables engineers to do more, and great engineers don't need PMs if you hire right.

• •Engineers own the product roadmap, speak to users, build features, analyze results - complete ownership without needing permission
• •Growth team (mostly ex-PMs) handles marketing, awareness, acquisition - partners with engineering lead on activation/retention
• •Product challenge in hiring: candidates must define features, create wireframes in Figma, AND design system architecture
• •Keep growth teams small and busy so they don't slip back into PM role doing wireframes and excessive user interviews
• •Prediction: PM role will split into either growth (PM + marketing) or product engineering (PM + coding with AI tools like Cursor)

Luke writes 100% of his code through Cursor by describing what he wants in chat. Core engineering team generates ~70% of code via AI. Only exception: research engineering for sensitive model code that represents half the company's enterprise value stays human-written.

• •Luke personally writes 20% of time, 100% through Cursor chat describing desired functionality
• •Core product engineering: ~70% AI-generated code across the codebase
• •Research engineering (AI models): 0% AI-generated due to sensitivity - Peter's models were half of initial $1B valuation
• •Growth features and add-ons are easiest to AI-generate since they're layered on top of core product

How ElevenLabs approaches enterprise with both top-down (ABM, executive dinners, webinars) and bottom-up (developer advocates, hackathons) strategies. North Star metric is marketing-sourced sales qualified leads. Webinars work but should be rebranded as 'Academy' or similar.

• •North Star for enterprise marketing: number of marketing-sourced SQLs (signed up, booked SDR call, SDR qualified)
• •Dual approach: enterprise marketing team (top-down ABM) + developer team (bottom-up awareness) with overlapping audiences
• •Webinars still work but rebrand them - OpenAI calls theirs 'OpenAI Academy' which sounds more valuable than 'webinar'
• •Enterprise marketing team will be 20 people by end of year, separate from mobile app's 5-10 person growth team
• •Focus on CAC to payback period (12-36 months depending on product) rather than CAC to LTV due to LTV uncertainty

Reframes how to think about retention for prosumer/consumer products versus enterprise. Individual user retention can be 85-90% but account-level NRR should exceed 100% when factoring in virality. Also discusses why VCs should be investing more aggressively in AI companies hitting $6-10M MRR quickly.

• •Consumer/prosumer: individual user retention can be 85-90%, but account NRR should exceed 100% when including viral growth from sharing
• •Enterprise: focus on account expansion (150% NRR) not individual seat retention - seats churn but accounts grow
• •If CAC to payback ratio is positive, put foot on gas immediately - don't just increase spend 20% weekly, grow as fast as possible
• •VCs are under-investing in AI companies despite seeing $6-10M MRR in first year - this is the RIGHT time to move fast, not 2021 COVID boom
• •Challenge: 10-15 competitors in every AI category vs 2-3 historically, making winner selection harder despite strong traction

Deep dive on counter-positioning using Ramp vs Brex as case study. Take competitor's core strength and position it as a weakness. Ramp made credit card points (Brex's strength) seem wasteful, positioning savings and software investment as superior. Applicable to any competitive market.

• •Counter-positioning: take competitor's core message and position the exact opposite as your strength
• •Ramp vs Brex example: Brex emphasized points/rewards, Ramp positioned points as wasteful and emphasized saving money via software
• •Once you establish counter-position, every competitor move gives you opportunity to reinforce your differentiation
• •Can apply to podcast format: 20VC was short/concise counter to long-form podcasts, now others counter-position against 20VC
• •Risk: as you become incumbent, you become target for counter-positioning (e.g., 'the 19 minute VC')

Founder brand is critical but can be distracting. Focus on channels that naturally resonate with founders rather than forcing Twitter activity. Video supply is exploding but still most effective medium. Hired creator whose ElevenLabs videos outperformed official content to run TikTok/YouTube.

• •Founder brand important but risky - can optimize for wrong metrics (LinkedIn likes) instead of building product
• •Matti focuses on in-person speaking events and fireside chats where he naturally excels, skips Twitter/X
• •Other team members (Luke, developer team) fill Twitter gap - don't force founders into channels they don't enjoy
• •Video supply increasing but still most effective medium - platforms boost it, people don't read blog posts
• •Hired creator who had most-viewed ElevenLabs video on YouTube (above official content) to create TikToks, YouTube Shorts, Instagram content

Common mistakes: doing paid marketing too early before PMF (especially B2B), wasting time on A/B testing instead of building better product. Biggest regret: not staffing proven channels sooner (affiliates set up 18 months ago by one engineer, untouched since, now generates tens of thousands MRR monthly).

• •Biggest mistake for B2B founders: paid marketing before PMF - wastes time on funnel optimization instead of product building
• •Exception: consumer products where distribution IS the risk, not just product quality
• •A/B testing messaging via paid ads rarely works - better to brute force more content like podcast episodes than optimize fonts/colors
• •ElevenLabs regret: not hiring dedicated people for working channels sooner - affiliates program untouched for 18 months despite strong ROI
• •Once you see signs of life in a channel AND have overall PMF, immediately staff one person laser-focused on that channel

Luke changed his mind on conversational AI agents - initially skeptical about latency, now believes they work better than human customer support. Infrastructure players like ElevenLabs can become application layer by being explicit about which areas they'll compete in. Inbound SDR role is dead, replaceable by AI agents doing BANT qualification.

• •Changed mind on conversational AI: sub-200ms latency now achievable, prefers AI customer support over humans (more knowledgeable, no fake politeness needed)
• •Infrastructure to application: works if you're explicit about competition areas - ElevenLabs builds conversational AI platform but won't compete with Synthesia/HeyGen on avatars
• •Inbound SDR role is dead: function is just data collection (BANT qualification), easily replaced by conversational AI agent
• •Can clone AE voices for AI agents to pre-screen their own deal flow before booking calls
• •Best growth strategy observed: Brian Johnson's 'Don't Die' brand - controversial, catchy messaging, constant Twitter engagement, authentic community building