Every IPO S-1 prospectus includes a section that lists all sorts of risk factors as to why the stock is not a buy. Many of them are routine warnings, warnings about financial performance, or about wars or acts of nature. However, some of them are company specific.
For cloud services startup ServiceTitan, which filed public IPO documents with the SEC on Monday, we have another boilerplate warning that the large-scale language models (LLMs) that help the business grow can also lead to harm. We may be witnessing the birth of
ServiceTitan had a 1,150-word risk factor about how the use of AI, specifically generative AI, could negatively impact its business. It warned that LLM's hallucinatory behavior could produce “inaccurate” information and lead to “discriminatory” behavior. That LLM may infringe the copyright or intellectual property of others. LLM also exposes more data to potential hacking and harm. However, if we do not have enough data, we may not be able to continue offering AI products or building new ones. The company also warned that employees and contractors could accidentally share customers' personal data with third-party systems, leading to security breaches or using that data to train models. There is.
ServiceTitan warns that the use of AI may someday conflict with social and ethical issues. Or that future regulations could be costly.
ServiceTitan is also concerned that it may not be able to hire the AI experts it needs for its products, and even if it does, they will be expensive. ServiceTitan warns that its AI products rely on third-party services it doesn't control, specifically naming Microsoft and OpenAI. Therefore, if it becomes unavailable or something goes wrong, that is also a risk.
These warnings are interesting because ServiceTitan operates in an industry nexus that is expected to be dominated by GenAI, particularly LLM-led AI agents. ServiceTitan provides software for businesses in the field service industry (think primarily small businesses, construction contractors, HVAC professionals, landscapers, etc.). Its software performs functions such as marketing, customer relationship management, customer support, and accounting. (ServiceTitan declined further comment.)
AI agents are now moving at a furious pace across sales, marketing, and CRM, with Salesforce CEO Marc Benioff, king of the CRM market, telling TechCrunch that his company alone will He said 1 billion AI agents could work for Salesforce customers. .
ServiceTitan has been offering an AI-powered service called Titan Intelligence since at least 2023. In October, it also introduced an AI agent suite for sales, customer service, and call centers.
But the LLM is designed to study words, images and data and create your own versions. In other words, they were designed to make something up. OpenAI scientists are busy racing toward general intelligence that mimics humans' better reasoning abilities. In other words, they are working on AI in hopes of further increasing their creativity.
Indeed, the credibility of LLM in business applications is something that will have to be worked out over time as more and more companies specialize in agents where reliability is more important than creativity. But it's worth noting that the companies selling AI bots today are actually saying the quiet parts out loud in legalese. At this early stage, AI adoption is likely to cause as many problems as it solves.