YouTube Entrepreneurs Society
From product idea to full architecture:
ready to build
Salma Jafri, YouTube Strategist
"I've been working with AI tools for a while, but the architecture and technical side of building a product was a real gap in my knowledge. In one session, Sam gave me the full picture: the stack, the cost model, the build sequence. I left knowing exactly what to build first and confident I could actually do it. I didn't expect to come away with that level of clarity."
Salma Jafri
YouTube Strategist, Founder of YouTube Entrepreneurs Society
Background
Salma Jafri is a YouTube strategist and the founder of YouTube Entrepreneurs Society, a membership platform for online business owners who use YouTube to grow. She books an AI Strategy Sprint to explore how AI can serve her existing audience in a product that stands on its own.
Salma already has the audience, the credibility, and the content infrastructure. What she wants to figure out is what to build, what stack to use, what it'll cost to run, and how to sequence the build so she doesn't over-engineer before she knows what her members will pay for.
The opportunity
Salma's audience has a clear, specific problem: they want to use YouTube to grow their businesses, but the platform's feedback loop is slow. You post, you wait. There's no tool that gives you real-time, personalised guidance on what to improve, what's working, or how your channel compares to what's performing in your niche.
The Sprint works through what that tool could look like. The core concept becomes a YouTube tool suite -- several connected AI-powered tools that help creators analyse their channels, refine their brand voice, and generate content direction based on actual performance data.
What we designed
The Sprint produces a clear technical architecture for the product:
- Frontend: Next.js (React-based, good ecosystem, easy to deploy)
- Backend + deployment: Railway (simple, predictable pricing, no cold starts)
- Database: Supabase (Postgres with good free tier and auth built in)
- AI models: Dual-model -- Gemini 2.0 Flash for speed and cost on high-volume tasks, Claude Sonnet for depth on brand voice analysis and strategic recommendations
The dual-model approach is deliberate. Gemini handles the fast, cheap tasks (data parsing, formatting, structured extraction from YouTube API data). Claude handles the nuanced outputs -- brand voice work and strategic recommendations -- where quality matters more than throughput.
Cost model
Infrastructure costs come out at approximately $49-109/month at a reasonable scale:
| Item | Estimated cost |
|---|---|
| Railway (backend + deployment) | $5-20/mo |
| Supabase (Postgres + auth) | $0-25/mo |
| Gemini 2.0 Flash (AI calls) | ~$4/mo at moderate usage |
| Claude Sonnet (AI calls) | ~$20-40/mo depending on usage |
| Vercel (frontend) | $0-20/mo |
| Total estimate | ~$49-109/mo |
At a subscription price of $49-97/month, the product needs only 2-3 subscribers to cover its own infrastructure costs. That's a healthy margin profile even at very early traction.
The four-session build plan
Rather than trying to build everything at once, the Sprint produces a sequenced four-session plan. Each session produces something real and usable.
Brand Voice Agent
The first tool to build: analyses a creator's existing content and builds a detailed brand voice profile. Foundational because it feeds data into every other tool in the suite.
Channel Analyser
Pulls YouTube API data (views, CTR, watch time, audience retention) and surfaces patterns -- what's working, what isn't, and what to do next.
Thumbnail and title scorer
Rates thumbnails and titles against known performance patterns in the niche. Gives creators a score and specific improvement suggestions before they publish.
Content roadmap generator
Takes the Brand Voice output + Channel Analyser data and produces a personalised content direction -- topics, formats, and posting cadence tailored to the channel.
Return on investment
The Sprint costs $1,250. At $49/month per subscriber, Salma recovers that cost when six members join and stay for one month -- before accounting for any ongoing subscription revenue.
At 100 subscribers (a conservative target given her existing audience of YouTube Entrepreneurs Society members), the product generates $4,900-$9,700/month in recurring revenue -- at infrastructure costs of under $110/month.
The Sprint also saved months of exploratory work. Architecture decisions that would typically require multiple developer consultations -- which stack, which models, how to sequence the build -- were resolved in a single session.
What came next
Salma left the Sprint with a detailed technical roadmap specific enough to hand to a developer or begin building herself. The Brand Voice Agent was scoped as the first build -- a standalone tool with immediate value that also forms the foundation for everything else in the suite. The session resolved the stack decision (Next.js + Railway + Supabase + dual-model AI), produced a realistic cost model, and gave her a sequenced plan that doesn't over-engineer before she knows what her members will pay for.
Ready to design your own product?
In two hours, we'll go from scattered ideas to a clear architecture, a sequenced build plan, and a roadmap specific enough to act on.
Book your AI Strategy SprintQuestions? Get in touch at hello@samanthanorth.com