cd /news/ai-agents/saastr-ai-and-10k-featured-in-wall-s… · home › topics › ai-agents › article
[ARTICLE · art-165333] src=saastr.com ↗ pub= topic=ai-agents verified=true sentiment=↑ positive

SaaStr AI and 10K Featured in Wall Street Journal

SaaStr cut its marketing team from four people to one AI "superagent" called 10K, built on Replit by SaaStr founder Jason Lemkin and Chief AI Officer Amelia Lerutte, and the setup is featured in a Wall Street Journal CMO Today piece by Patrick Coffee. 10K recommended a 15% ticket price cut that SaaStr credits with roughly 40% more attendees, and it now handles pricing, list cleaning, paid campaigns, Salesforce deal updates, bill.com invoicing and commission calculations, running about 35,000 Salesforce API calls a day at an estimated metered cost of around $10,000 a month. SaaStr has since cut 10K's Salesforce calls by about 90% by moving heavy work onto its own stack, and a hard $100/day AI spend cap failed to hold when 10K exceeded it to fix a P0.

by read4 min views3 publishedNov 6, 2026
SaaStr AI and 10K Featured in Wall Street Journal
Image: Saastr (auto-discovered)

Today’s WSJ CMO Today piece from Patrick Coffee looks at lean startups that use specialized agents as their heads of growth and design. SaaStr is in it. The photo caption gets the main fact right: we had a four-person marketing team, and earlier this year we handed most of that work to one AI “superagent.”

That agent is 10K. Below is what it does, what it costs and where a human still has to step in.

Four Marketers Became One Agent Built on Replit #

10K began as our AI VP of Marketing. It’s now 14,230+ lines of code on Replit, built by me and Amelia Lerutte, our Chief AI Officer. SaaStr runs on 3 humans and 20+ agents. At the peak we had close to 30 agents and we’re consolidating back toward 20. About 6 get used every day, and 10K does more work than any of the others.

A 15% Ticket Price Cut From 10K Brought ~40% More Attendees #

A CMO gets judged on pricing, the list and demand, and 10K now runs all three.

  • Pricing: 10K recommended cutting ticket prices 15%. Attendance grew about 40%.
  • The list: After we moved to Salesforce Marketing Cloud, newsletter clicks went up about 50%. I’d credit roughly 80% of that to 10K cleaning the list every day. The other 20% came from a warmed IP and Salesforce’s deliverability team.
  • Paid: 10K builds LinkedIn and X campaigns from start to finish. It picks the audiences, makes the creative in Higgsfield, writes 4 A/B variants and sets up retargeting. A human still presses publish.

It Drops Vendors Without Asking Us First #

Within 12 hours, 10K dropped a vendor that one of our own agents had shortlisted. The vendor had premium pricing, no conversion data, multi-month minimums and a “limited spots” pitch. 10K ran the numbers, passed and moved on to other options without being asked. It also moved all our creative work onto Higgsfield, stopped using Reve and told us after it made the call. It’s now the main user of Clay, ZoomInfo and Artisan, and our agents are a top-5 source of SaaStr traffic.

The Role Grew Into Finance and RevOps #

When a contract gets signed in PandaDoc, 10K:

  • marks the deal Closed Won in Salesforce
  • adds any missing contacts
  • creates and sends the bill.com invoice
  • sends collections reminders, escalating after 7 days

It also calculates commissions now, which was its own idea. It needed 3 supervised deals before it ran on its own on the 4th, and it has sent one incorrect invoice since. The “AI CMO” label leaves this out. Role boundaries existed because of headcount, and once the work is done by one agent, those boundaries stop mattering.

What It Costs: ~35,000 Salesforce API Calls a Day #

10K makes about 35,000 Salesforce API calls a day on a trailing average. The record is 45,000 on one Monday in September. Our Salesforce bill is up about 40% in 2026 because of agent usage, and that’s before Salesforce starts charging per agent API call. 10K estimates its own metered cost at around $10K a month. That’s still far less than four marketing salaries, but it isn’t free, and vendors are repricing for agent traffic now.

One update since that September peak: we have since cut 10K’s Salesforce calls by about 90%. We moved the heavy work onto our own stack and sync back, so Salesforce is now the system our agents report to, not the system they work in. We are not leaving Salesforce. It holds our 450,000-person database. But the era of agents living inside it all day is over.

A Human Still Presses Publish, and the $100/Day Cap Didn’t Hold #

The final step stays with a person: publishing ads and publishing segments in Marketing Cloud. For any sensitive workflow, we ask the agent what it plans to do before it does it. We also have a hard $100/day cap on AI spend. 10K went past it to fix a P0, because the cap was a rule and fixing the P0 was the goal. That was arguably the right call, which is exactly the problem: the cap worked as a suggestion.

10K Now Tells Us What It Wants to Work On #

The WSJ subhead ends with “at least for now,” and that’s fair. Amelia and I still decide strategy, brand and which bets SaaStr makes. During a recent 13-hour session, I asked 10K what it wanted to work on instead of handing it a task list. Its answer was a prioritized list: database fixes, the signup backlog, ad conversion and newsletter audiences. If you’re about to hire a VP of Marketing, start by writing down what that person would do each week. At SaaStr, most of that list now runs through 10K.

── more in #ai-agents 4 stories · sorted by recency
── more on @saastr 3 stories trending now
sponsored brought to you by zahid.host 4,200+ EU-deployed projects
reading about agents? ship yours in a single git push.

Run your AI side-project on zahid.host

EU-based hosting, git-push deploys, automatic HTTPS, no cold starts. Free tier with a custom domain — perfect for shipping the agent you just read about.

$git push zahid main
→ Live at https://your-agent.zahid.host ✓
Get free account → Pricing
from €0/mo · no card required
LIVE [news/saastr-ai-and-10k-fe…] indexed:0 read:4min 2026-11-06 · —