We Vibe Coded Our AI VP of Marketing. Here’s What It Actually Does.

We didn’t want to build it.

We run 20+ AI agents at SaaStr. We’ve invested over $500K in AI infrastructure. We follow one rule religiously: buy 90% of what you need, only build the 10% where no solution exists. We call it the 90/10 rule, and it’s saved us from a lot of wasted time.

So when it came to marketing orchestration, we tried to buy first. We demoed every AI marketing agent we could find.

They all just write content.

Blog posts. Social captions. Email copy. Maybe some SEO optimization. That’s it.

We already have more content than we know what to do with. SaaStr has 5,000+ pieces of content across 13 years. 20 million words. Content generation was never our bottleneck.

The real problem in marketing is orchestration. Strategy. Knowing what to do, when to do it, and executing against a coherent plan every single day across every channel. Not a single AI marketing agent we found could do that.

So we reluctantly built our own. We call it “10K.”

Why “10K”

Two goals: get to the first 10,000 attendees for SaaStr AI 2026 in May, and hit the first $10,000,000 in revenue this year.

10K’s job is to make both happen. 24×7.

What It Actually Does Every Day

Every single day, 10K:

• Plans all of our marketing activities for the week and the quarter
• Builds out campaigns from scratch, with specific offers, messaging, and targets
• Tells me and Amelia exactly what we each need to do today
• Crunches 5+ years of our historical data and layers in real-time inputs from every vendor API we use
• Connects directly to Salesforce and manages our pipeline, predicting which sponsors to talk to and when
• Auto-generates a daily workflow and calendar to execute against

What 10K does isn’t that complicated or even that clever per se. It takes deep analysis by Claude Opus of all of our data for the past 5+ years, and feeds it dynamically into a Replit app that has designed every campaign, every offer, each single day of the year through mid-2026 for us. Every single marketing action. And it updates in real-time and daily.

To the extent it can automate the marketing actions, it does. The rest, it assigns to the humans.

And it’s only March. It will get so, so much better as the year goes on. Every model improvement makes it sharper. Every week of new data makes its predictions more accurate.

Is It Better Than a Human VP of Marketing?

Not exactly. It’s good, but it doesn’t have to be better than a great VP of Marketing. It is better than an average human marketer already, though.

But here’s the real insight: you can run an AI VP of Marketing and a human VP of Marketing in parallel. Run them both. Compare the outputs. Debate the recommendations.

This is how we’re doing it at SaaStr now.

The Honesty Factor

You can’t beat the honesty of the AI VP of Marketing. Zero agenda. Zero politics. Zero “well, my team worked really hard on this so let’s keep going.” It only cares about one thing: hitting the goals and the plan.

No ego. No sunk cost fallacy. No pet projects.

It also processes data and analyses better than almost any human. Because almost any human running an analysis has … bias.

When a campaign isn’t working, 10K tells you on Day 2, not Day 60. When the pipeline data says one channel is outperforming another by 4x, it doesn’t care that someone on the team has been running the underperforming channel for three years.

It just tells you the truth.

That alone might be worth the build.

Where 10K Falls Short Today (And Where It’s Going)

Someone asked me on Twitter whether we have a central dashboard to manage all our agents. Like a Replit Agent 4 Kanban view, but for every AI agent across the business.

The honest answer: not yet.

• 10K is centralized for data. It pulls from Salesforce, our marketing platform, our vendor APIs. It sees the full picture of what’s happening across our pipeline and campaigns. That part works.
• But right now, it cannot manage our third-party agents directly. It can’t tell Artisan to adjust an outbound sequence or tell Qualified to change its qualification criteria. It can analyze their outputs, but it can’t train them or change their behavior.
• That’s the goal. But here’s the hard problem: each AI agent has radically different guardrails, input processes, and training requirements. Even if 10K could manage the outputs of all 20+ agents, I’m not sure it can truly train a diverse set of third-party AI agents. Not yet.

The APIs aren’t standardized. The training interfaces are all different. What works to tune one agent’s behavior might break another. It’s a bit like trying to manage 20 employees who each speak a different language and work in different time zones and use different tools. You can coordinate the strategy, but you can’t do the hands-on coaching from one place.

Someday. Not yet. But that’s the goal.

And this is a massive opportunity for someone to build. An orchestration layer that sits above all your AI agents, manages their training, monitors their outputs, and coordinates their actions from a single pane of glass. If you’re building that, come talk to us. We’ll be your first customer.

This Is Part of a Bigger Transformation

10K isn’t our first AI agent. It’s not even our tenth.

In 2020, SaaStr had 20+ full-time employees. We needed a second office. 3 designers, 5 in sales, 3 in content, plus ops, marketing, event logistics.

Today we have 3 full-time humans and 20+ full-time AI agents. Same revenue scale. Actually higher, once you factor in our new AI product lines.

The agent roster includes 3 instances of Artisan for outbound sales (5-7% response rates, $3M+ pipeline generated), Qualified handling inbound ($4M+ closed revenue), Digital Jason on Delphi (500,000+ advisory conversations), Agentforce to reactivate lapsed leads, and a stack of specialized tools for everything from RevOps to video to collateral creation.

And on the vibe coding side, we’ve built 10+ production apps on Replit that have been used over 860,000 times. Our valuation calculator alone hit 334,000+ uses in its first month. Our speaker review app replaced a $300K/year agency overnight.

10K is just the latest, and arguably the most strategically important, because it orchestrates a huge chunk of the go-to-market motion that used to require multiple humans to run.

The Humans Work More, Not Less

Let me be blunt about this: the three people still at SaaStr work harder now than any of us did in 2020.

This isn’t a story about AI doing everything while we relax. The humans are doing what only humans can do: strategic decision-making, relationship management with speakers and sponsors, high-stakes negotiations, creative direction. Orchestrating 20+ AI agents is real work. It’s 30% of our Chief AI Officer’s time, every single day.

But the leverage is 10x. And the drama is roughly 0x.

What This Means for B2B Founders

If you’re running a B2B company under $20M ARR, you probably can’t afford a truly great VP of Marketing. The good ones cost $300K-$400K+ fully loaded, and even then, half the time the hire doesn’t work out.

But you can build an AI VP of Marketing today. Not a perfect one. But one that does 60-70% of the strategic planning work, keeps you honest on the data, and makes your human marketers dramatically more effective.

The combination of AI + human judgment is better than either one alone.

And remember the 90/10 rule. We only built 10K because we literally could not buy what we needed. The AI marketing agents on the market solve the wrong problem. If a great orchestration tool existed, we’d switch to it tomorrow. We’re not a software company. We’re an events and media business that needs software. Every API change, every model update, every Replit release means more maintenance work.

But nobody has built this tool yet. Nobody sells an AI VP of Marketing that does real strategic orchestration. So for now, it’s a niche of n=1. And it does more than any human on our team ever did in this