Sep 8, 2025 · 43m · startup-ideas
Screensharing How This Guy Makes Millions with AI Agents (Rowan Cheung)
0:00 / 0:00
▶ Watch on YouTube →
gold bands on the timeline = statements, start to end. Hover to read, click to jump. CC turns on captions
In this masterclass on AI-driven business leverage, The Rundown AI founder Rowan Cheung demonstrates seven practical AI workflows—spanning synthetic video avatars, Claude editorial prompts, and autonomous Lindy and Zapier agents—that allow a lean 15-person team to operate with the output of a 50-person media enterprise.
How this conversation actually went
Every chapter scored 0–10 on four independent dynamics. Hover any point for the reasoning behind the score. Greg holds 27.9% of the talking time here. How this is scored →
Greg as informed peer 4.5 Guest teaching 5.4 Guest disagreement 0.2 Greg pushing back 0.8
speaking balance: gold is Greg, purple is the guest (3 minute bins)
Sharpest disagreement ▶ 33:32 Rowan dismisses Dia browser in favor of Perplexity
When Greg suggests Dia browser has a comparable assistant feature, Rowan rejects switching and defends sticking with Perplexity Comet because it is already sticky and functional.
Hardest push from Greg ▶ 9:55 Greg challenges the applicability of AI avatarsGreg stops the positive momentum to demand why founders building cash-flowing companies shouldn't use AI avatars, forcing Rowan to articulate the authenticity risks.
Biggest teaching moment ▶ 20:20 Rowan reveals Claude context-length workaroundRowan demonstrates an exact UI workaround to bypass Claude's maximum conversation length by branching edits from earlier prompts, catching Greg completely by surprise.
Greg holds their own ▶ 9:03 Greg proposes advanced Gumloop and Reddit sourcing workflowsGreg demonstrates technical domain expertise by outlining how founders can integrate Gumloop and Lindy to pull viral Reddit trends directly into Slack to fuel avatar scripts.
the scores for every segment, with the reasoning behind each
| Chapter | Topic | Greg as informed peer | Guest teaching | Guest disagreement | Greg pushing back | Why |
|---|---|---|---|---|---|---|
| Scaling The Rundown AI with Autonomous Workflows | 4 | 3 | 0 | 1 | Greg sets the stage by framing how autonomous AI systems allow lean teams to operate like large enterprises. Rowan agrees and clarifies their team of 15 operates with the speed of a 50-person organization. | |
| AI Video Avatars: Repurposing Text into Short-Form Video | 1 | 5 | 0 | 0 | Rowan walks through his video avatar workflow using HeyGen and ElevenLabs to repurpose written newsletter content into short-form videos. Greg listens attentively with minimal interruption. | |
| Idea Browser Platform for Startup Market Trends | 6 | 5 | 0 | 1 | Greg identifies the specific viral video Rowan displays and suggests using tools like Gumloop and Lindy to scrape Reddit for trending story ideas. Rowan explains separating ElevenLabs audio from HeyGen video for higher production quality. | |
| Evaluating Avatar Use Cases, Authenticity, and Audience Moats | 6 | 4 | 1 | 2 | Greg pushes Rowan to explain why founders shouldn't use avatars, prompting Rowan to discuss audience authenticity and human moats. Greg contributes an analysis of the psychological barrier of 'cringe mountain' in content creation. | |
| Voice-Dictated Ideation and Claude Style Training for Social Media | 2 | 6 | 0 | 0 | Rowan demonstrates his method of dictating raw thoughts via Whisperflow while walking and feeding them into Claude trained on his top historical tweets. Greg observes the prompt structure. | |
| Brainstorming Walk Routines and the Final Ten Percent Polish | 5 | 5 | 0 | 2 | Greg probes Rowan's brainstorming routine and specifically challenges him on how to bridge the final 10% polish gap that AI struggles to handle. Rowan explains his iterative prompting process. | |
| Claude as Newsletter Editor-in-Chief and Context Management Hacks | 4 | 8 | 0 | 1 | Greg asks about handling Claude's context length limitations in extended editing threads. Rowan demonstrates an editing hack where branching off an earlier message preserves context while bypassing thread-length caps, visibly surprising Greg. | |
| Deploying 'Ava': Autonomous Meeting Scheduling via Lindy AI | 6 | 6 | 0 | 0 | Rowan demonstrates 'Ava', his Lindy AI meeting scheduler integrated into workspace email. Greg highlights the conversion advantage of one-minute response times for inbound service leads. | |
| Inbound Lead Qualification and Lindy Agent Configuration | 5 | 6 | 0 | 1 | Rowan explains configuring inbound lead qualification agents and notes why he paused it for enterprise clients due to risk of hallucination. Greg notes the high revenue stakes of enterprise sponsors. | |
| Perplexity Comet as an On-Demand Interactive AI Video Tutor | 6 | 5 | 1 | 2 | Rowan explains using Perplexity Comet as an interactive video tutor grounded in full YouTube transcripts. Greg suggests the competing browser Dia, but Rowan firmly defends staying with Perplexity. | |
| Custom GPTs for Frictionless Employee Onboarding and SOPs | 5 | 6 | 0 | 0 | Rowan shares his internal onboarding GPT trained on SOPs and meeting notes under a 'GPT first' policy. Greg validates the system as reducing anxiety for junior hires asking sensitive administrative questions. | |
| Automated Email Inbox Triage and Routing with Zapier Agents | 4 | 6 | 0 | 0 | Rowan explains how Zapier Agents triage support and newsletter inboxes, route key signals to Slack, and draft replies, resulting in landed sponsorships. Greg connects this to his own unread YouTube inbox problem. |
Statements from this episode (15)
Cheung: The Rundown's 15-Person Team Operates Like 50 People
“We have a team about 15. I think we operate as like a 50 person team in terms of speed, but in some cases we might actually operate faster because like the headache of having 50 people is like really burdensome.”
Cheung: The Rundown Gained 150K+ Followers In One Year Using AI Avatars
“We started doing it and it's been under a year and we've grown to over a 150,000 followers.”
Cheung: Limiting AI Avatars to Three Seconds Maximizes Visual Realism
“The avatar is only really shown for the first like three, four seconds. Cause the avatar portion of it isn't that great, but if you just show like 3:04 seconds, it looks more real.”
Cheung: ElevenLabs Direct Generation Outperforms HeyGen's Native Voice Integration
“HeyGen has like an 11 Labs integration. But you actually want to do it separately, because if you do it on 11 Labs, you get higher quality and you just stitch it together.”
Rowan Cheung: Human Authenticity Is the Moat Against AI Content
“Because like, as we get into the age of AI generated content and there's just like seas of AI generated content, you want to push for like that human layer. Cause that's really your moat.”
Cheung: AI Exaggerates And Ultimately Fails At Delivering Authentic Human Insight
“The insight part, AI sucks at giving like human insight and experience, comparing certain things to like other models, for example, and just like tweaking up little things that just don't seem right. Like, Claude likes to exaggerate a lot. Like I say, like wor…”
Cheung: The Rundown drafts newsletters manually, using Claude as editor-in-chief
“When we write our newsletter, we write it all by hand, But we like to use Claude as like an editor in chief.”
Cheung: Co-editing with AI increases editing time but produces superior content
“It actually takes more time to edit with AI as a co-editor, but your content gets so good. Because it's trained on your best writing and you're just going to find these small details that you wouldn't have otherwise seen.”
Cheung: The Rundown built custom Lindy scheduling agent Ava
“Lindy actually provides like a template for this, but I've modified it a little bit because I wanted like my workspace email. So I actually created an agent and I called it Ava. So Lindy has this, like they give this own template and they use like their email …”
Cheung: Sending Calendly links is often perceived as disrespectful
“And I know there's this weird stigma around Calendly. I don't know how many people who watch your podcast know about this. I don't know why. I don't agree with it. But like if you send a Calendly link, it's kind of known, I don't know, it's kind of disrespectf…”
Cheung: 1% AI Agent Error Rate Is Too Risky for Enterprise Inbound
“I'm just a little weary of using agents right now because There are like occasional times where like it could go wrong, and just that like one percent Knowing that it could go wrong one percent just worries me for those like really high tier clients.”
Rowan Cheung Uses ChatGPT Voice While Reading To Boost Concept Retention
“So for example, I use ChatGPT a lot or ChatGPT voice when I'm reading. So I'll just have it open next to me. And whenever there's a topic I don't really understand, I'll just ask it, oh, hey, what does this mean? And I'll go deeper with it before I move on to …”
The Rundown Built A Custom Internal GPT Trained On HR And SOPs
“We built this GPT that's trained on SOPs literally loom transcripts, meeting notes, our policies on AI usage, HR docs, any like admin really related question. And it also knows everyone on the team and their role.”