This week the SaaStr AI YouTube channel passed 200,000 subscribers
11.7 million views and 1,022,350 hours watched, which is roughly 116 years of continuous watch time. Average view duration sits at 5 minutes 50 seconds. People aren’t bouncing.
A year ago, the most-watched content on this channel was about how B2B companies scale. Pricing, sales motions, our classic content. That content still performs. But it’s no longer what pulls the biggest audiences. The videos doing the biggest numbers now are the ones documenting what it actually looks like to run a company on AI agents in production. Not the theory. The mess.
The Growth Is Gaining
Look at the last 90 days against the prior 90:
- Views: 1.47 million, up 167%.
- New subscribers: 38,437, a 56% jump in the rate of growth.
- Watch time: 19,381 hours, up 65%.
The split between those two numbers is the whole story. Views nearly tripled, but watch time grew slower than views. That gap tells you the new audience is finding us through Shorts and quick-hit clips, then converting into the people who sit through the long agent breakdowns. The top of the funnel got wider and the bottom got deeper at the same time. That only happens when the content people discover you for is the content they actually wanted.
For the record, total impressions are now at 65.7 million. The content has shown up on screens tens of millions of times. The subscriber number is just the slice of that audience who decided to come back.
The Agents Became the Flagship
If you want one explanation for the last 90 days, it’s a single series: The Agents.
This stopped being a video series a while ago. It’s the running record of what happens when you actually hand AI agents the keys to the company instead of speculating about the future of work in a keynote. We run 21+ agents in production. We film what they do, including the parts that go wrong. In the last 90 days that one series contributed more than 125,000 views, and a far larger share of the watch time that actually turns viewers into subscribers, which I’ll get to below.
The arc people followed:
- The scale. We documented building a $10M+ AI agent stack that automated more than 600 meetings. This is the video to send anyone who still thinks “AI agents” means a chatbot. It’s an operating model, not a feature.
- The unfiltered version. The breakout was the week an agent negotiated a vendor renewal and effectively became our CFO. That one video was among our top subscriber drivers this month. It worked because we left in the parts that didn’t, which is the whole reason people trust the series.
- The costs. We didn’t bury the bills. The $500K AI bill video and the running thread on “lazy agents” that quietly burn money exist because the cost side is where most teams are about to get surprised.
- The org questions. The series moved past “can it work” into the harder stuff: managing QBee, our AI VP of Customer Success, and the now-recurring debate in the comments on whether your next human hire should report to an agent. QBee in particular set off a wave of technical questions about vibe coding and where you put the guardrails. That is the clearest signal we get that this audience wants to copy the build, not just watch it. The episode where our AI started doing the hiring pushed that conversation from theoretical to operational.
The pattern across all of it is the same: engagement spikes every time we talk about a real agentic workflow with a real number attached, and it flattens the moment anything drifts toward hype. The audience has already moved past “is this coming.” They’re trying to deploy.
The Agents Outperforms
The Agents is a new series, only nine major installments in, and it’s already outperforming long-form strategy sessions we’ve been refining for years. The relevant signals:
It builds loyalty faster than it builds reach. The 125,000 views the series pulled in 90 days is a modest slice of our 1.47 million total. But it accounts for roughly 25% of all watch time from new subscribers in that window. So it isn’t what most people find first. It’s a big part of what makes them stay once they do. People show up for a Short and end up watching the agent builds.
People watch the dense somewhat technical episodes to the end. Against our standard event footage, The Agents holds about 15% higher retention through the middle of a video, the exact stretch where most long-form loses people. On the heaviest episodes, like How We Built a $10M+ AI Agent Stack and Our $500K AI Bill, average view duration runs up to 60% above the channel average.
The series converts viewers to subscribers better than our interviews. Per 1,000 views, The Agents #005 brought in 32% more new subscribers than our channel average. People don’t subscribe to be entertained. They subscribe because they expect the next episode to be useful to a decision they’re about to make.
Transparent thumbnails win the click. The videos that name the actual number on the thumbnail, like Our Salesforce Bill Jumped 80%, are pulling click-through rates above 4%, well over our 90-day average. The lesson for anyone publishing in B2B: put the real number on the thumbnail. The specific figure earns the click better than any clever tease.
Beyond the Agents: The GTM and Event Content
The agents series carries the channel, but two other things hold the deeper watch time.
Eleanor Dorfman’s breakdown of AI-native sales at Anthropic became the reference video people send to their teams. The comments aren’t “great talk.” They’re “I’m rebuilding my sales process around this.” Information density is the whole product.
And the live SaaStr AI Annual coverage proved the demand is global, not Bay Area. The Snowflake CMO’s session on the death of the dashboard is approaching a million views on its own. Nobody watches a million minutes of a CMO talk unless they’re trying to make a decision Monday morning.
The Stuff That Performed Best Was the Stuff We Almost Didn’t Post
The counterintuitive lesson from a year of this: the videos where we showed the failures outperformed the ones where we showed the wins.
We posted the time our Salesforce bill jumped 80% because of agent activity. We posted the conversation about AI deception and where the ethical line actually is, which is not a comfortable topic for a team building and sharing in public.
Every one of those did better than the equivalent “here’s how to do it right” video would have. A recurring comment is some version of “this should be behind a paywall.”
Thank You
We’re not diversifying away from this. We’re concentrating harder on it:
- The agent stack, with the bills attached. The honest version of which tools and APIs actually return ROI, including cost. Claude, Replit, the vibe-coding workflow, and what each one costs to run at scale. Numbers, not logos.
- Founder-led AI. More pitch clinics and more operators showing exactly how they’re rebuilding their companies in real time, with the receipts.
- The human side of the math. The harder question underneath all of this: how to augment a team with agents without quietly destroying accountability inside the org. We don’t have this fully solved. We’ll keep showing the work.
We really appreciate your support. Our YouTube may not be as flashy as some other content out there. But we think it’s some of the most practical and useful in AI + B2B
