How to launch your first AI product using what you already know

Executive overview

Most businesses already collect customer data, hold expert knowledge, and provide some kind of interface. An AI product is simply those three things combined and routed through a large language model.

The window to act is now — building costs have dropped sharply, and the businesses that move early will be very difficult to displace.

The core insight: your existing intellectual property plus customer data plus a clean UI is all you need to build a scalable AI product.

The three-part product model

• User data — something unique the customer tells you about themselves (form, portal, connected system)
• Intellectual property — what you know that most people don't; your proprietary frameworks, databases, or expertise
• User interface — a well-designed portal that makes the experience feel effortless
• When all three intersect, feed the result to an LLM via an API call; the customer gets far more back than they put in

Real examples from the presenter's businesses

• ScoreApp — built from an agency that sold online assessments; AI added inside to scale value to customers
• Bookmagic.ai — distills publishing expertise into prompts and guidelines so authors write better books; growing ~8% per month
• Awards product (in development) — combines a global awards database with user business details to generate winning entries

What the next generation of SaaS looks like

• Old SaaS: a database with a nice UI wrapper
• New SaaS: a large language model with a nice UI wrapper
• The LLM wrapper is made of user data, intellectual property, and interface design
• API calls handle the communication — not advanced; any IT person can implement them

Economics and build costs

• Build a basic version for $20–30k
• Charge a subscription (e.g. $39/month); 500 subscribers = $20k/month
• Add tiers over time (e.g. $39 and $99 plans) to scale revenue
• Unit cost per LLM call is small and can be factored into subscription pricing
• Unlike old SaaS, you do not need massive scale to make this viable