Analysis
I've sat through maybe forty pitches this year that describe themselves as AI-powered and are, underneath, the same SaaS product from 2019 with a chatbot bolted onto the sidebar. Crunchbase's Bob Morse put a name on it this week that I've been saying in less polished form for a year: AI-sprinkle versus AI-native -- read his piece. His argument, boiled down: sprinkling AI onto an existing workflow makes that workflow marginally faster; rebuilding the workflow around what AI actually makes possible changes the unit economics of the business. Those are not the same category of company, and investors keep pricing them as if they were.
Sprinkle companies are easy to spot in diligence. The AI feature sits on top of the product, usually behind a settings toggle, and the pitch deck slide about 'AI differentiation' describes a feature, not a moat. Ask what happens to retention if you turn the AI feature off -- if the answer is 'not much,' you're looking at a vitamin, not a business model change. AI-native companies look different from the org chart down: headcount per dollar of revenue is structurally lower, the product does work a human used to do rather than assisting a human doing it, and the AI isn't a line item, it's the reason the company can exist at a price point the old incumbents can't match.
“The AI feature sits on top of the product, usually behind a settings toggle, and the pitch deck slide about 'AI differentiation' describes a feature, not a moat.”
The founders getting this right aren't the ones talking about AI the most -- they're the ones who can tell me their cost to serve a customer and how it's changed over four quarters. That's the number I ask for now instead of the roadmap. If a company can't show a declining cost curve tied to AI adoption, I assume it's sprinkle until proven otherwise, no matter how good the deck reads.
Room for disagreement: the sprinkle-vs-native framing can turn into a purity test that punishes perfectly good incremental businesses. Plenty of large, durable companies win by shipping AI features into an existing distribution advantage rather than rebuilding from scratch -- Intuit and Adobe aren't rearchitecting their entire businesses around AI-native workflows, they're layering AI into products with real moats already, and that's a legitimate strategy, not a failure of imagination. Not every company needs to be AI-native to be a good investment; some just need to be smart about where they add AI without breaking what already works. Morse's framework is directionally right for new venture bets, less useful as a blanket rule for incumbents with existing distribution.
