Eight Sleep now runs its entire email marketing operation with zero human staff, generating close to $100 million in revenue through AI bots built in three days. Co-founder Matteo Franceschetti revealed on the 20VC podcast that the company has structurally replaced human teams with AI agents across engineering, paid media, and finance - operating in 35 countries with 160 people at a revenue-per-employee figure he claims is "way higher than Apple."
The transformation began when the email marketing lead departed. Co-founder Alexandra Zubillaga stepped in to test whether bots could handle the function. "Within 3 days she was able to build multiple bots that now run all our email marketing and so now we have a team of zero and email marketing makes close to 100 million," Franceschetti said. The broader engineering shift is even more sweeping: "Our engineers stopped coding around a year ago. What they have is hundreds of AI engineers that code for them."
The headcount numbers are stark. Engineering runs on a small human team directing hundreds of AI agents, primarily through Claude. Paid media - managing hundreds of millions in spend - operates with two people. Finance runs on four people where comparable firms need twenty. A new internal AI tools team, formed roughly six months before the conversation, builds thousands of agents for internal use. Franceschetti estimates the company runs "probably by now hundreds if not thousands of agents" and has begun measuring AI employees alongside human ones. By his math, the company is three to four times larger when AI headcount is included.
The $200 million attribution trap
Franceschetti's most pointed operational warning for marketers concerns platform-reported customer acquisition costs. "You cannot trust there will be an excess of attribution from the same platform. If you look at the CAC on Meta it is probably 20% lower than what the true CAC is," he said. Platforms have every incentive to overstate their contribution, and companies that accept those numbers uncritically are building their growth strategy on inflated math.
The fix is incrementality testing every six months. His method: pick two comparable regions, turn off a channel in one, and measure the revenue difference. If Texas and California are both growing 50% year over year and you kill Meta in one state, the delta reveals Meta's actual contribution. "Every brand needs its own models because otherwise they just fool themselves," he said. This discipline extends to the company's entire growth posture. Eight Sleep targets immediate payback and healthy day-zero contribution margins. AI for Marketing tools can accelerate execution, but without independent measurement, the numbers are fiction.
The cost of that discipline is real. "We could be growing 50% more if we wanted year over year this year just by spending more. But then next year you're in trouble because your CAC is upside down," Franceschetti said. Channel expansion must be sequential - start with Meta, set a hard CAC cap, and scale only within it. Overspend, pull back, and enter the next year with no year-over-year growth, and "good luck fundraising when you're going from 400 to 70."
Word of mouth now accounts for roughly 40% of revenue. TikTok only started working in the six months before the episode, driven by influencer content volume. The company is also tracking "AI SEO" - how it ranks in AI-powered searches - which Franceschetti estimates falls between 1% and 20% of traffic and is growing.
How AI agents replaced whole functions
The organizational model at Eight Sleep is built on what Franceschetti calls "teams of two" - small, flat groups with redundancy and world-class oversight. Marketing staff receive a morning report from AI agents suggesting changes to make or reject. The growth function is run by "a few people." The AI internal tools team stays tightly coupled to data engineering so agents have accurate access to revenue-by-country and CAC-by-country data. AI Agents & Automation at this scale requires infrastructure discipline, not just tooling.
Franceschetti's monthly AI spend is "in the millions" - under $5 million, though growing fast enough that he no longer tracks it precisely. He predicts two opposing cost vectors: usage will increase dramatically, with AI spend potentially rising from 5% to 50% of engineering salary budgets at large companies, but unit costs will fall. Net spend, he argues, goes down.
Celebrity deals with two non-negotiable conditions
Eight Sleep's sports partnerships are extensive. Charles Leclerc used the product for two years before meeting Franceschetti and later invested. Roughly 80% of Formula 1 drivers use it. Tadej PogaΔar installed the unit himself at every hotel on every Tour de France stage. Top tennis players sleep on it at every Grand Slam. One prominent actor bought ten units for family. Another individual owns 55 units across multiple houses.
Franceschetti imposes two conditions on any athlete or celebrity deal. First, they must already use the product or try it and give feedback - pure business transactions are rejected. Second, he requires a direct relationship with the athlete, not through managers, because the real value is word-of-mouth among their peers. A financial stake alone "goes nowhere."
Why this matters for marketers
The Eight Sleep model has immediate implications for marketing leaders. Platform-reported attribution is systematically inflated, and companies that do not run their own incrementality tests are optimizing against false numbers. Email marketing - a function many organizations still staff with dedicated teams - can be automated with AI agents that handle the full workflow. The channel sequencing lesson is equally concrete: new channels fail not because the channel is bad, but because brand awareness is insufficient. Franceschetti spent heavily on out-of-home advertising in the Middle East before the brand was known there, and it failed. The same campaign in New York, where awareness is high, would perform differently. For marketers managing growth budgets, the takeaway is that measurement independence and channel patience are not optional - they are what separates companies that stay "public ready" from those that overspend into a corner.
Your membership also unlocks:
