If you were forwarded this newsletter, and you'd like to receive it in the future, subscribe here.​

​The 4 Questions Startups Should Ask Themselves about Building with Generative AI​

There are 4 questions a startup should ask themselves about building a startup that uses generative AI.

I presented those questions & my views on their answers at Saastr’s Workshop Wednesday.

I had a blast putting this deck together. I started with a few sentences, uploaded them to gamma.app to outline the presentation, popped over to Midjourney to generate images along the story line, & published it in IA Presenter.

The video is here. The last slide contains the prompts for the images in the presentation.

The narrative is :

AI is a massive platform change that Goldman Sachs projects will increase GDP 300x more than the PC.

GS estimates a 1.5-2.9% increase to US GDP, doubling GDP growth, net of 7% job loss. The PC increased GDP by 0.006%, according to NBER

That alone should turn heads.

There are 4 questions startups should ask themselves about building with generative AI.

1. Layer : application, platform, or infrastructure? In the cloud, AWS, Azure, & GCP have created about as much market cap as all the top 100 B2B & B2C publics built on cloud (Netflix, ServiceNow, AirBnb, etc). But there are 100 applications compared to 3 infrastructure vendors.
2. Market : how to compete with incumbents? Startups have negative time to launch in many markets with Adobe, Microsoft, & Salesforce launching Gen AI enabled software in weeks.
3. Moats : how to develop competitive advantage? Competing on algorithms is possible but hard. Access to proprietary data provides a moat. Usage & distribution, like in classical SaaS, are likely the most sustainable & repeatable. Enterprise readiness will be an essential : ensuring buyers are safe from legal & compliance risks.
4. AI Depth : what level of technical sophistication to bring into the company? Startups can integrate with a plug-in, build a prompt-tuning engine (a little model on top of a bigger model), or develop & train their own models. Each subsequent choice is more expensive, but provides a deeper moat. It’s likely startups start at plug-ins & then move down with scale that affords more usage & more capital to invest.

If you’re building in the space,I’d love to hear from you.

​