How do you validate an AI startup idea before building?

To validate an AI startup idea before writing code, founders and investors emphasize focusing on the problem rather than the technology, using low-fidelity prototypes to gauge demand, and securing financial or data-driven commitments from early users.

1. Validate the Problem, Not the AI

The most common mistake is falling in love with a technical solution rather than a user problem. Uri Levine (Waze co-founder) argues that you must first validate people's "perception of the problem" through dialogue before thinking about the AI solution Lenny's Podcast — A founder’s guide to crisis management | Uri Levine (Waze co-founder, serial entrepreneur) @ 1:04:18. Andrew Ng adds that vague ideas (e.g., "AI for healthcare") are dangerous because they are easy to agree with but hard to build; instead, you must find a "concrete idea" that gives the team a clear direction to either validate or falsify quickly Y Combinator — Andrew Ng: Building Faster with AI @ 06:06.

2. The "Kickstarter" Validation Model

Brett Malinowski advocates for a "Kickstarter model" where you validate the market concept before tackling any engineering Brett Malinowski — Digital Dropshipping: The New Way to Start A SaaS Business @ 06:08. This process involves:
* Clickable Prototypes: Use Figma to create wireframes that look like a finished app. Hire a designer if necessary to ensure it looks professional Brett Malinowski — Start A SaaS in 5 Simple Steps (I did it) @ 03:02.
* Smoke Testing: Build a simple landing page with screenshots of the Figma design and drive traffic via social media or small ad spends ($50) to see if people sign up for a waitlist Brett Malinowski — Start A SaaS in 5 Simple Steps (I did it) @ 03:02.
* Pre-sales: The ultimate validation is getting someone to pull out their credit card. Malinowski suggests offering "lifetime access" for a steep discount to those who pay before the app is released Brett Malinowski — Start A SaaS in 5 Simple Steps (I did it) @ 03:02.

3. Defining the Value Equation (B2B)

For enterprise AI ideas, Y Combinator recommends defining a "value equation" with the customer. For instance, if you are selling customer service AI, you might claim it can solve 20% of inbound queries. Validation then becomes a "proof of concept" where you use a small sample of the customer's data to prove you can deliver that specific ROI Y Combinator — The Sales Playbook For Founders | Startup School @ 06:06. Another way to build trust is a "side-by-side" trial, where the AI's output is compared directly to the results of their current human-led process Y Combinator — From Idea to $650M Exit: Lessons in Building AI Startups @ 27:27.

4. Build a "Minimum Evolvable Product"

Where Brett Malinowski focuses on design-first validation, Y Combinator emphasizes building a "quick and dirty" prototype to get real user feedback immediately Y Combinator — How To Start A Dev Tools Company | Startup School @ 06:06. They suggest building a "Minimum Evolvable Product"—something simple that can survive contact with early users and adapt fast based on the pressures they apply to it Y Combinator — How To Get Your First Users @ 03:02.

Where they disagree:
While Brett Malinowski suggests not spending a dime on engineering until you have commitments Brett Malinowski — Digital Dropshipping: The New Way to Start A SaaS Business @ 06:08, YC partners note that many technical founders find their best ideas by "vibe coding" prototypes and iterating as they go Y Combinator — How To Start A Dev Tools Company | Startup School @ 06:06.

5. Validate Distribution and Marketing

Finally, consider if the idea has a natural marketing engine. Zach Yadegari notes that he validated his AI app idea by ensuring it could "go viral on social media." If there is no existing niche or influencer community to promote the product, validation will be significantly harder than if you have a built-in audience Brett Malinowski — The 17-Year-Old Who Built a $1.12M/Month Ai App | Zach Yadegari @ 15:18.

— Sources: 10 videos across 4 creators