At 12:41pm yesterday, my inbox pinged. Not from a human. From 10K, our AI VP of Marketing.
Subject line: Marketing Plan v2 — 24 days to SaaStr Annual 2026.
Body: Three campaigns, sized, drafted, and ready to ship. Replaces yesterday’s plan with one new campaign idea and refreshes to data.
Nobody asked 10K to do this. It looked at the data from Friday’s end-of-day numbers, saw the previous plan wasn’t pacing right against our 24-day window to SaaStr AI Annual, rebuilt the plan, and dropped it in my inbox with Amelia copied.
That’s not a product demo. That’s what 10K does now, every day. And you can learn to build one yourself, for real, at SaaStr AI Annual May 12-14 in SF Bay!!
What 10K Actually Does
Quick recap for those new to this. 10K is the AI VP of Marketing we built on Replit. The name comes from our targets: 10,000 attendees at SaaStr AI Annual and $10M in revenue.
Here’s what it’s doing in production right now:
- Ships ideas on its own. Saturday mornings, Tuesday afternoons, whenever the data says a new angle is warranted. I don’t ask. It ships.
- Tells us what it needs. When 10K can’t execute something alone, it writes a brief. What it wants, why, what it needs from us, what success looks like. Like a good VP who knows how to ask for resources.
- Updates plans daily based on hard data. Not quarterly OKR check-ins. Not weekly standups. Daily. It looks at registration pace, email open rates, sponsor engagement, inbound flow, and rewrites the plan accordingly.
- Runs autonomously or ships outputs. Full auto, or 95% done with a human hitting send. Our call, campaign by campaign. More on this below.
Just Say Go
10K has two modes, and we pick which one per campaign.
Full auto. We say “go” and it runs. It pulls contact segments from Salesforce via API, pulls registration and attendee data from Bizzabo via API, drafts the emails, schedules them, and sends through Resend via API. End to end. No human in the loop. We can go watch Anna dive while it ships a re-engagement sequence to 40,000 contacts.
95% done. Same pipeline, but instead of hitting send, it drops the final drafts, segment lists, and send schedule into our inbox. We review. We hit send. 10K did the work. A human made the final call.
For a re-engagement sequence we’ve shipped 40 times, we let it run. For a new angle we’ve never tested, we take the review step. The agent doesn’t care either way. The scaffolding is the same.
Here’s the architectural point people miss: 10K does not need a human in the loop to work. It can do 100% of the work. We choose to stay in the loop on some sends because we want to, not because the agent can’t handle it.
Most “AI marketing tools” are just workflow tools with AI drafting bolted on. The human was still the system. The AI was a feature. 10K inverts that. The agent is the system. The human is the optional reviewer. That inversion is the whole ballgame.
A Human Might Be Better, Of Course. If You Can Find Them.
I’d love to hire a VP of Marketing who’s better than 10K. Good luck.
That’s not a flex. It’s math.
A real VP of Marketing costs $350K all-in in the Bay Area, minimum. They show up Monday, leave Friday. They take two weeks of vacation. They go to three conferences a year. They’re in meetings 40% of the time. They have opinions that get stuck in loops. They have career goals that may or may not line up with yours.
10K costs us roughly $400 a month in API and infrastructure fees. It doesn’t sleep. It doesn’t have off days. It reads every piece of data we have within minutes of the data existing. When I want a new campaign tested, the gap between “let’s try this” and “here’s the draft” is measured in minutes.
Would a great human VP of Marketing beat 10K on strategic vision and team leadership? Yes. Absolutely. We don’t have a team for them to lead right now.
Would they beat 10K on daily execution velocity, data responsiveness, and cost? No. And it’s not close.
What Changed in the Last Six Months
If I’d written this post in October 2025, I would have told you AI agents were promising but rough. Most couldn’t complete a full workflow. You had to babysit them. They’d hallucinate citations, forget context, give up on hard tasks.
Something shifted in Q1 2026. The tools got better. The context windows got longer. And the frameworks for giving agents persistent memory and clear objectives finally matured.
10K isn’t GPT with a system prompt. It has:
- Persistent memory of every campaign we’ve run
- Access to our actual revenue, registration, and traffic data
- The ability to draft, review, revise, and ship autonomously
- Access to every Salesforce and other record, period
- A feedback loop so it learns what worked
That last part matters most. Every campaign that ships becomes training data for the next one. The 10K we have in April is meaningfully better than the 10K we had in February. And the 10K we’ll have in July will make today’s version look slow.
Come Build One With Us. For Real.
SaaStr AI Annual is May 12-14 in the SF Bay Area. We’re not just going to show you how 10K is built. We’re going to build one together.
Amelia is running a live workshop on May 12 at 4:15pm as part of Deploy ’26, our AI GTM Agent Summit sponsored by Vercel.
No slides. No pre-recorded demo. She walks on stage with a laptop, a mic, and 45 minutes, and builds a working AI VP of Marketing from zero in front of the room. You follow along on yours.
Bring your laptop. Seriously. You’re going to build alongside her. Every prompt, every tool connection, every place it breaks and how she fixes it. You leave with a functional agent and the scars to maintain it. This is the most registered workshop at the entire event and it’s not close.
If you’re a founder who still thinks you need to hire a VP of Marketing before your Series A, or a CMO who thinks your next hire is another director, come to AI Annual. You don’t need to be behind on this. You really don’t.
See you May 12. Laptop charged.
