20VC: 7 Predictions for How AI Changes the World: Labour, Engineering, Social Media, GrokBots Buying Cybercabs and more with Matteo Franceschetti, Co-Founder @ Eight Sleep

Summary

Matteo Franceschetti says AI has already transformed Eight Sleep’s internal operating model: email marketing is fully automated, paid media is managed by a tiny team with agent support, and engineers no longer code in the traditional sense because AI coders now do that work. He argues growth decisions must be driven by incrementality tests, geo holdouts, and internal CAC models rather than platform-reported attribution, which he считает inflated. He also makes the case that brand awareness remains the key bottleneck for consumer hardware, especially when expanding into new geographies like China or the Middle East. Looking forward, he predicts AI will reshape labor, marketing, purchasing, and even company structure, with some businesses becoming holdings of multiple AI-enabled verticals. The conversation closes on his personal drive, skepticism toward university for founders, and a belief that UBI or similar redistribution will become necessary as AI reduces human labor demand.

Notes

• Hiring has swung back to 2021-like competition, with US engineer and PM compensation described as overheated.
• Retention is now harder than initial hiring because strong brands keep getting repeated higher offers.
• Angel investing is presented as market intelligence on valuations, multiples, hiring profiles, and channels.
• Early hardware founders routinely misestimate BOM, COGS, and return rates by at least 50%.
• Word of mouth is a major compounding channel and is said to be about 40% of revenue at scale.
• TikTok only started working for Eight Sleep in the past six months after iterative content development.
• Paid channels are managed with weekly budget unlocks and CAC caps.
• Eight Sleep runs incrementality tests every few months by turning channels off in selected states or geos.
• He does not trust platform-reported CAC because attribution is inflated.
• Meta-reported CAC is estimated to be roughly 20% below true CAC.
• The right approach is one channel at a time, starting with Meta before expanding spend.
• The company optimizes for immediate payback and day-zero contribution margin.
• Subscription revenue compounds on top of healthy day-one economics.
• Brand awareness remains the main bottleneck, so Matteo prefers more marketing over price cuts.
• Eight Sleep’s churn is described as low single-digit annual and world-class.
• AI search is now a tracked acquisition channel called AI SEO.
• AI-search traffic is still growing, and Google behavior appears to be shifting because of it.
• An out-of-home campaign underperformed in the Middle East because the brand was too new there.
• Brand awareness is measured with aided and unaided recall.
• The company built a truck with a bedroom in the back for live product trials.
• That experiential truck also visits events and weekend traffic hotspots.
• The marketing budget is in the hundreds of millions.
• Email marketing used to have a two-person team and is now fully AI-run.
• A cofounder built multiple email bots in three days after the team lead left.
• AI suggestions use historical data to decide email timing and content.
• Claude is likely the main tooling for email workflows.
• Eight Sleep likely has hundreds or thousands of internal agents.
• Counting AI workers, he believes the company is effectively 3-4x bigger.
• Morning reports from AI agents suggest changes to digital marketing, with humans approving or rejecting.
• Human discovery will be overloaded by content, so agent discovery needs its own strategy.
• Engineers stopped coding about a year ago because AI engineers now do the coding.
• The company claims scale across 35 countries with a tiny team.
• Claude spend is in the millions per month, but below $5 million.
• AI usage will rise while costs fall, so net spend may decline over time.
• He is less concerned about distributed intelligence than about how businesses change once intelligence becomes semi-unlimited.
• He expects a massive productivity jump comparable to the computer or industrial revolutions.
• AI-native companies may evolve into holdings with multiple businesses.
• Eight Sleep could eventually own 3, 4, 5, or 10 businesses.
• Chinese automakers pivoted toward electric technology about 10 years ago.
• He expects Chinese automakers to become robotics companies.
• Humanoids are framed as the next trillion-dollar market if Chinese automakers adapt.
• Chinese robot competitions already look clunky but may become mind-blowing within a year.
• He says frontier AI self-learning is the core risk because systems may start training themselves.
• He believes frontier labs may already see more dangerous capabilities than end users do.
• China is a major market for Eight Sleep and has required learning from scratch.
• Chinese consumers do not buy on websites; they buy through super apps like WeChat and RedNote.
• In China, websites are mostly browsing surfaces while transactions happen in super-app ecosystems.
• His own bot already buys through Amazon in a browser, so he does not yet share card data directly with AI.
• He expects trust in AI purchasing assistants to improve within one to two years.
• A bot buying a cybercab is framed as a future norm for autonomous commerce.
• Eight Sleep is starting to detect diabetes and hypertension likelihood from its data.
• He says the bed is becoming a medical device for sleep apnea.
• Retail is expected to matter only once brand recognition is high enough.
• China expansion also depends on brand maturity and local distribution understanding.
• Silicon Valley overweights product and underweights brand in consumer hardware.
• A great brand can support premium pricing if the product is already good.
• The Charles Leclerc relationship began after he already used the product for two years.
• He requires athletes and celebrities to already use the product, or to try it and give feedback.
• Athlete and celebrity deals are only worthwhile if there is a real relationship, not a pure transaction.
• He claims a very large share of Formula One drivers use the product, but the relationships are confidential.
• A famous buyer reportedly owns 55 units for different houses.
• Other buyers have ordered in the hundreds through embassies.
• Eight Sleep’s ambition is to democratize sleep while keeping a Tesla-like premium ladder.
• If the product cost $100, he believes they would have sold billions.
• He frames lowering costs as an engineering mission to make better sleep accessible.
• The company is only 160 people and claims much higher revenue per employee than Apple.
• Comparable companies at similar revenue scale are described as having around 1,000 employees.
• Paid media is run by just two people despite managing hundreds of millions of spend.
• He wants the organization to stay extremely small and flat.
• An AI internal tools team was created only six months ago.
• That team is building thousands of internal agents.
• The agents need direct access to revenue, CAC, and backend data pipelines.
• The paid-media agents are already said to generate hundreds of millions.
• Finance has been compressed to five people instead of a conventional larger team.
• China is the hardest market because sales cannot happen through the website.
• He uses a health-relationship-purpose hierarchy as his model of happiness.
• University is framed as useless for pragmatic founder careers.
• He expects AI-driven efficiency to make universal basic income necessary.
• He expects corporate bottom lines to improve enough to justify redistribution.
• He is undecided about allowing his children to use social media.
• He is interviewing parents of successful people to learn parenting.
• He is treating parenting as a due-diligence process.
• He believes childhood drive often comes from trying to please a parent.
• He says broken father-son relationships often drive ambition in boys.
• His father set a bar he still feels compelled to meet.
• He was still a lawyer in Milan when his father died.
• At that point his later life in the US, Silicon Valley, and entrepreneurship did not yet exist.
• His mother feared the entrepreneurship risk because she lacked financial cushion.
• He had a lawyer job to fall back on if the startup failed.
• He says he did not speak English until 24 and knew almost nothing about computers.
• He credits a small Italian startup and a father-backed loan for his move to San Diego.
• He went from average grades to magna cum laude after deciding to pursue elite law firms.
• The path to elite law firms required English and magna cum laude.
• A corporate box at Chelsea changed his perception of money and quality of life.
• He says he becomes relentless when he has a goal.
• He describes himself as unable to stop working and bored after a few days off.
• He admires people who can sit alone, without a phone, and enjoy a meal in peace.