Eight Sleep Replaced Its Email Marketing Team With AI Bots Built in Three Days
A co-founder used Claude to build helpers that pick who gets each email.
What happened When the lead of Eight Sleep's email team left, a co-founder spent three days building bots with Claude that decide when to send each email, to whom and with what copy. The CEO says email now runs with zero staff and makes "close to $100 million."
Who this is about: Sleep tech hardware and subscription company (Eight Sleep, about 160 employees)
What it takes
- Cost
- Free planClaude: Free plan. Pro is $20 a month billed monthly ($17 a month billed yearly), and Max starts at $100 a month. Team seats are $25 a month billed monthly.Price checked Oct 24, 2026 on claude.com. Prices change; check before you buy.
- Setup time
- A few daysOur estimate from the source, for one business.
- Skill needed
- Some tech skillsComfortable installing an app, pasting an API key or running a command.
Explain it simply
Eight Sleep sells a smart mattress cover, and email is one way it sells. When the email lead left, a co-founder used Claude, an AI chat tool, to build small helpers. They look at which old emails worked. Then they pick when to send, who gets each note, and what it says. The company says email still brings in a lot of money with no staff on it. That figure is the CEO's own claim.
What was done
Eight Sleep had a two-person email marketing team. When the person leading it left, co-founder Alexandra Zatarain decided to see whether AI could do the job instead of hiring a replacement. Within three days she used Claude to build multiple bots that now run the whole email calendar: when to send, to whom and with what copy, based on historical performance data.
The company also has an internal AI tools team that sits close to its data engineering team, because its agents need reliable access to the same backend numbers.
How it went
On the 20VC podcast, Franceschetti said the email team is now "a team of zero" and that "email marketing makes close to 100 million." That figure is the CEO's own statement in an interview, not an audited number. Critics on another podcast pointed out that email mostly collects demand that other paid channels created, so not all of that revenue is the bots' doing.
- This is also a large brand with a big existing customer list, so a small store shouldn't expect the same scale. No details of the bots' prompts, tools or review process were shared.
Prompt
Inputs: past campaign results (send time, segment, subject line, opens,
clicks, revenue) exported from your email platform.
Bot 1 - Scheduler: "From this history, propose next week's send calendar:
which days and times performed best for each segment, and why."
Bot 2 - Audience picker: "For each planned send, choose the segment most
likely to buy, and list who should be excluded (recent buyers, unsubscribes)."
Bot 3 - Copywriter: "Write subject lines and body copy for each send in our
brand voice, using the messages that performed best before as examples."
Start with one flow (for example, abandoned cart or win-back) and compare
results with your current version before handing over the full calendar.Our starting outline, not Eight Sleep's bots.
Business fit
Who this helps, and how:
Ecommerce brands with an email list: consider rebuilding a function AI-first when someone leaves, instead of hiring into the old setup.
Shopify & ecommerceSmall marketing teams: let AI handle the routine decisions (send time, audience, copy) and base them on your own past results.
StartupsFounders: connect your agents to clean, shared data first, so every bot works from the same numbers.
Solo founders
Your turn
Ideas to try on your own work. These are starters, not results anyone has measured.
- Which email do you send on a schedule that a helper could draft?
- What would you still want a person to approve before an email sends?
- Which past emails would you show a helper so it can match your tone?
Related
- Evidence grade BClaude in Chrome Set Up Four Google Ads Campaigns From a Keyword ListStrong evidence, using Claude, Google Ads and Semrush
- Evidence grade CAn Apparel Brand Sends 40% Fewer Emails Using AI Segmentation, With Claude Building SegmentsVendor or secondhand, using Claude, Klaviyo and Orita (ML segmentation)
- Evidence grade DClaude Picked a Slow Product and Segment From Omnisend Data, and the Email Brought In $26,619Light evidence, using Claude, Omnisend and Shopify
- AI marketing use casesDepartment hub: AI for marketing, 26 examples
Similar tools
Other tools for this job, and how they differ. Chosen by our editors, never paid for.
- Klaviyo
An email platform with built-in AI for send times, segments and copy.
- Braze
A customer engagement platform with Intelligent Timing for per-person send times.
- Iterable
A marketing platform with send-time optimization and AI features.
“Light evidence” (grade D) means: A brief, early or in-progress claim. Treat it as an idea to test, not proof. How we grade. Spotted a mistake? Suggest a correction.