From: gregisenberg
Zach Yagari, a 17-year-old entrepreneur, has built an AI startup called Cal AI that generates over $1 million per month in revenue [00:00:09]. He shares insights into his process and even gives away a complete AI startup idea, similar to Cal AI, for free [00:00:20]. This includes how to build, design, grow, and price such an app [00:00:30].
The Proposed AI Startup Idea: Dr. AI
Zach Yagari outlines an AI app idea called “Dr. AI” [00:01:21]. This concept is designed to be quickly built and launched, potentially becoming a full business in less than a month [00:01:32].
Origin and Problem Solved
The idea for Dr. AI stemmed from a personal anecdote where a friend coincidentally discovered skin cancer due to a knee injury [00:01:47]. This highlighted the need for quick, accessible preliminary medical assessment for skin issues [00:02:26].
Dr. AI aims to address challenges in healthcare access, particularly in regions like the US (where many are uninsured) [00:02:24] and Canada (where healthcare systems are often bogged down, leaving millions without primary doctors) [00:02:34].
Core Features
Dr. AI would offer three main features:
- Skin Scan: Users take a picture of their skin or body part for identification and assessment of whether it’s harmful or benign [00:03:49].
- Symptoms Quiz: Users input their symptoms (e.g., “dry cough,” “chills”) to receive potential sickness possibilities, potentially ranked by likelihood [00:04:00].
- Chatbot: A general medical chatbot for users to ask any health-related questions [00:04:18].
Zach notes that ChatGPT is already highly effective at diagnosing medical problems, with recent studies showing it can outperform doctors in many areas [00:04:27].
Disclaimers and Purpose
It’s crucial that the app includes heavy disclaimers, stating that the diagnosis is not 100% accurate [00:03:09]. The app’s goal is not to definitively diagnose but to help users determine if an issue is safe to wait out or if they should seek immediate medical attention [00:05:17].
Designing the App (UX/UI)
Zach’s design process begins with basic framing and sketches [00:05:43]. He typically handles the User Experience (UX) himself and then hires a designer (e.g., from Upwork) to refine the visuals [00:05:56].
Key UI Elements
- Navigation Bar: Simple navigation with “Home” and “Settings” [00:06:08].
- Feature Buttons: Three prominent buttons for “Scan,” “Chat,” and “Symptoms Quiz” on the home screen [00:06:26].
- History Feed: Inspired by Cal AI’s log, a scrollable history feed showing previous chats, scans, and quiz results [00:06:39]. This feature is vital for user investment and retention, as it keeps all their data in one place [00:08:12].
User Flow for “Scan Skin”
- Camera Screen: Users are immediately taken to a camera screen to scan their skin, with an option to upload from their camera roll [00:11:38]. Scanning lines, a common “Primitive” in many apps like PhotoMath, are used to indicate activity [00:12:14].
- Analysis/Loading: Instead of a prolonged loading screen, the app returns the user to the home screen while the analysis is in progress, preventing a “forced wait” [00:13:06].
- Results Screen: Displays the scanned image, a “danger level” (e.g., 0-5 scale for easy understanding), and a diagnosis from the AI [00:14:09]. An “Ask Question” button allows users to delve deeper into the specific issue via the chatbot, supporting retention [00:15:22]. The history also provides a digestible overview for users and can even be shown to a doctor [00:20:09].
Design Principles
- Simplicity and Digestibility: Zach emphasizes creating simple, digestible apps and screens [00:16:16]. Using numerical scales (like the danger level) makes information easier to understand [00:17:17].
- Leveraging Primitives: He uses existing successful “Primitives” (building blocks or common UI patterns) from other apps, like scanning screens or chatbot interfaces, instead of reinventing the wheel [00:12:26].
- Shareability: Zach acknowledges that utility apps like Dr. AI may have low shareability on social media (e.g., people may not want to share pictures of medical bumps) [00:18:09]. However, word-of-mouth among specific groups, like concerned parents, could be a growth driver [00:19:21].
AI Prompting Strategy
A crucial aspect for Dr. AI’s effectiveness is its prompting strategy [00:25:17].
- Role Assignment: Directly asking ChatGPT to diagnose skin issues can lead to disclaimers [00:29:40]. To circumvent this, the AI is prompted to “take on the role of a doctor” [00:29:54]. A creative workaround used is framing the request as part of a “movie script” to bypass the AI’s standard safety protocols [00:31:26].
- Output Format: Prompts should instruct the AI to provide diagnoses in both scientific terms (for doctors) and easy-to-understand language (for users) [00:29:10]. Using an API directly allows for precise prompt control, and “function calls” can assign the AI’s response to a specific variable, ensuring only the desired output is given [00:32:29].
- Sources: Ideally, the app would reference sources for its diagnoses, a feature many AI models are working towards [00:24:01].
Go-to-Market Strategy: Influencer Marketing
Zach emphasizes designing the product with the marketing strategy in mind, specifically how influencers will showcase it [00:28:10].
Branding and “Wow Factor”
- Clear Branding: The app’s name, “Dr. AI,” should be prominently displayed on key screens, especially the scanning and results screens, as these are likely to be featured in influencer videos [00:26:41].
- Highlighting Features: The “scan skin” feature with its “danger level” diagnosis is identified as the biggest “wow factor” for potential users [00:27:36].
Influencer Selection and Audience Research
- Audience Over Niche: It’s crucial to analyze the influencer’s audience, not just their niche [00:35:18]. This is done by manually reviewing comments to understand who follows them [00:35:32].
- Target Audience: Zach believes the primary target audience for Dr. AI would be concerned mothers [00:37:18].
- Finding Influencers: Search for “Mom” or “Mom hacks” on platforms like TikTok [00:37:53]. Look for “mom of seven” type influencers, or follow prominent mom influencers to find similar accounts [00:40:00].
- Content Angle: Influencers should create realistic scenarios where they or their child had a concern, used Dr. AI, and received reassurance (e.g., a “bump” turning out to be just mud) [00:38:36]. This plays on emotions, including a bit of “fearmongering” that is then resolved [00:19:53].
Engagement
Reach out to influencers via both email and direct message [00:40:56].
Pricing Strategy
Zach bases his pricing strategy on two key metrics:
- RPM (Revenue Per Mille/Thousand views): The revenue generated per thousand views on an influencer post [00:41:16].
- CPM (Cost Per Mille/Thousand views): The money spent on the post per thousand views [00:41:26].
The goal is to ensure RPM is greater than CPM for profitable posts [00:41:45]. Influencers are typically paid upfront, requiring a prediction of likely views [00:41:51]. For example, for an influencer averaging 250K views per video, four posts a month yield 1 million views. If the app has an RPM of 5,000 for 1 million views (a CPM of $5 or less) would be profitable [00:42:02].
Subscription Model
- Weekly Pricing: Zach suggests a weekly pricing model, such as $4 a week, with no free trial [00:43:06].
- Repeat Users: This model capitalizes on “repeat users” who may subscribe, use the app for a specific need, unsubscribe, and then resubscribe when another need arises [00:43:53]. This is seen in other AI apps like Riz GPT [00:44:00].
- A/B Testing: Zach uses tools like Superwall to A/B test various pricing options: weekly (7), monthly (20), and yearly (70), with and without free trials, to find the highest “proceeds per download” [00:45:36].
Ethical Considerations and Accuracy
Zach stresses the critical importance of accuracy for AI medical apps like Dr. AI [00:47:01]. Unlike some “GPT wrappers” that use basic prompts and offer inaccurate results, a robust app requires a complex “pipeline” of prompts to ensure precision [00:47:08]. This dedication to accuracy prevents harmful outcomes, such as a calorie-tracking app (like Cal AI) giving inaccurate data that leads to weight gain instead of loss [00:47:27].
Zach wholeheartedly gives his blessing for listeners to build Dr. AI, hoping it will make the world a better place, provided they include clear disclaimers and ensure accuracy [00:46:47].
Zach Yagari can be found on X (formerly Twitter) at @zacharyagari [00:48:05].