Serial entrepreneur Bret Taylor is ready for his next act. More than a year since he stepped down as co-CEO of Salesforce, on Tuesday, he formally launched Sierra, a startup helping businesses develop conversational AI chatbots. And he has a $110 million war chest to help him tackle that problem, thanks to Silicon Valley investors Sequoia Capital, Benchmark and others.
Can Sierra Stand Out?
Sierra’s emphasis on catering to business customers shouldn’t come as a surprise. This direction aligns with Taylor’s background, evident from his previous venture Quip, which aimed to compete with Microsoft Office, and his tenure at Salesforce. It appears that, in line with his commitment to aiding companies in their modernization efforts, he teamed up with his longtime Google colleague Clay Bavor to establish Sierra and help businesses embrace AI.
“Since the advent of the Internet, technology has transformed how businesses interact with their customers. A company’s website in 1995 may have simply been a digital business card, directing customers to a phone number or retail location. Today, that same site likely encompasses the entire customer experience – from commerce to support.
Each wave of new technology since has dramatically changed consumer behavior and expanded companies’ digital footprint. Social networks enabled consumer brands to connect with customers via their profile page. Later, as smartphones became ubiquitous, many companies’ mobile apps became their primary customer experience.”
— Sierra blog post
The problem with all of that is all of these communication methods — websites, social media and mobile apps — create some form of friction, preventing customers from getting answers. But not with conversational AI, according to Taylor and Bavor:
“Conversational AI is so intuitive and frictionless, we believe it will have an impact on your customer experience on par with the Internet. In the age of conversational AI, the best customer experience is not installing an app or clicking a link, but simply having a conversation.”
As noble of a mission that is, Sierra enters a marketplace already filled with established competitors, some Taylor knows very well: Twilio, Zendesk, Intercom, Oracle, LivePerson and Salesforce. What’s more, these rivals have a larger platform offering that can scale for the enterprise.
Undeterred, Sierra says it has signed up noteworthy brands as customers including WeightWatchers, SiriusXM, Sonos, and OluKai.
Introducing the Agent OS
One factor that might distinguish Sierra from its competitors is its approach to artificial intelligence. Unlike others in the field, Sierra isn’t simply adding AI onto its current platform. Instead, it operates as an AI-centric business from the start. This means it might avoid the challenges associated with integrating AI into existing systems, dealing with extensive workflow adjustments, or addressing other legacy issues.
Furthermore, Sierra doesn’t seem to be showing a preference for one Large Language Model over another, despite Taylor’s position on OpenAI’s board. According to Fortune, the company clarified that a standard agent on its platform would utilize “four or five different AI models” during user interactions. While some of these models are exclusive to Sierra, the platform also incorporates open-source Large Language Models.
All of this leads to the creation of what Sierra is calling its Agent OS. It’s a development platform for businesses to quickly and easily build an AI agent. At the core of this enterprise-grade conversational AI platform are three things: sophistication, authenticity and trustworthiness.
The company claims that its AI agents can effectively manage any workflow or task you assign to them. These agents differ from the typical Q&A chatbots commonly found on websites. Sierra seems to employ retrieval augmented generation (RAG), enabling the agent to take action by retrieving information from your systems to fulfill its tasks.
Agents can reason, problem solve, and make decisions. With Sierra, you set goals to guide your agent towards the right solutions and guardrails to ensure your agent stays on-point and aligned with your policies.
Sierra states its agents utilize natural language and “sophisticated” reasoning to generate interactions customers will find “authentic, satisfying, and on-brand.” That means it’ll understand jargon, typos and context.
The company has auditing and quality assurance tools to ensure the correct answers are provided. The platform also has a “strict” data governance policy to protect personal information and keep all internal data safe.
A Generic Idea But Only the Start
As mentioned earlier, Sierra’s idea for chatbots isn’t unique. However, it’s worth paying attention to the company’s approach in the future because that could be a real differentiator, one that leaves incumbents like Salesforce scrambling to retool to catch up.
Sierra’s work bears some similarities to Google’s Duplex tool, albeit customized for the enterprise. Moreover, by having multiple AI models, Sierra is able to reduce the chances of hallucinations happening and also likely preventing people forcing the chatbot to say things a brand may not want it to say.
As it continues to refine its platform and attract notable clients, Sierra’s focus on sophistication, authenticity, and trustworthiness could indeed set a new standard in enterprise AI solutions.