A user needs direct AI analysis capabilities within their ERP system. The current barrier is the concern over unpredictable and high token costs. They currently use a manual workaround (prompt generator for external AI tools), highlighting a clear need for a seamless, integrated solution where token costs are either transparent, managed, or predictable within the ERP.
I saw a post a few days ago on how AI companies are totally different to SaaS and I can’t stop thinking about it… (or find it) It was basically saying that the whole idea of SaaS businesses is that you build it once and then every use is a marginal cost (with 75-99% margins). But AI is different. Every time your customer takes an AI action in your product, you’re spending tokens. All of a sudden, your margin varies massively and if you charge by use you’re locking in a fixed margin - that will constantly get squeezed. Th post went on to compare his AI company to operating more like a restaurant with super thin margins that vary massively by ingredients (i.e. AI functions and context size for tokens.) But it got me thinking about the massive spike in VC investment that has gone to AI and has me wondering if the VCs have adjusted their expectations or if they’re still applying traditional SaaS models to it. While I don’t think AI co’s are comparable to restaurants, I do think that it’s going to take a while to adjust our world view to the new normals. Software companies with non-marginal costs is a very different mental model. EDIT: Thanks to Geoffrey Forbes for finding the initial article: https://lnkd.in/drSssWzd