One afternoon panel featured two hard-core enterprise data entrepreneurs: Alex Ratner of Snorkel AI, which “cleans” and otherwise structures data for effective use in LLMs, and May Habib of Writer, which creates generative AI content for enterprises.
Habib started out coining a cautionary phrase that resonated: avoiding the “over-chattification of things.” She was referring to the tendency of big companies to rely on chatbots as an AI strategy, rather than focussing on product-market fit for specific products.
She divided the enterprise AI product task into three components: a great LLM, a good scheme for connecting proprietary data and RAG solutions, and strong guardrails for managing accuracy and compliance. And developing enterprise AI solutions is not the same thing as deploying them effectively.
“There is a really long ‘last mile’ to building applications that get adoption,” she said.
Ratner, whose company has 10 years of experience in understanding enterprise data and how to make it useful, said specialty LLMs were destined to play a big role, because companies have highly particular needs and are worried about data security.
For Habib, at least, demand has not been a problem. When asked whether the Writer billboards on the 101 Freeway were really working, she said yes, in fact, they were, though that was no proof of the power of outdoor advertising. “In generative AI, everything is working.”
Add comment