Sierra unveils Curie and Fleming models, Tandem Voice and Persona Studio at Summit

Two new models inside Agent OS

Sierra says its Agent OS relies on a "constellation of models" fine tuned to specific tasks, from customer acquisition to sales and service. The company points to earlier models such as Linnaeus and Darwin, which it says help understand what customers are asking and surface the right answers in milliseconds, without hallucinating.

At the Summit, Sierra added two more. Curie powers the core orchestration loop of a conversation: it decides which path to follow, which tools to call, and writes the replies customers see and hear. Fleming detects when the caller on the phone is another agent, which Sierra says lets businesses handle bots and fraud as well as personal agents.

According to Sierra, Agent OS provides state of the art compliance and adherence, low latency for voice, and a cost basis better than a company could build itself, so customers can focus on their agent while Sierra manages the token costs.

Voice: Tandem Voice and Persona Studio

Sierra describes voice as the place where model latency and quality are most noticeable. Its Tandem Voice lets an agent "think fast and think slow at the same time." A Conversation layer hears and responds in real time, picking up on mannerisms, tone and context. Underneath, a Reasoning layer uses the full audio and text transcript to map the path forward. While that layer works, the agent can keep the caller informed, for example by saying it is checking on an order.

Persona Studio lets companies upload their own voice samples and tune the personality of their agent directly in Agent Studio. Sierra says no engineering and no AI expertise is required. The company cites one large subscription business that saw its resolution rate improve by over 50% after deploying Voice Personas.

Personal Agent Protocol and Ghostwriter

Sierra and Meta are developing the Personal Agent Protocol, an open standard that defines how personal agents interact with businesses. Industry partners named in the announcement are Genesys, Instinct, Rocket, Shopify, Stripe and Walmart. Sierra says the standard is meant to address questions such as authentication and let companies decide what functionality is available to personal agents, via APIs or the company's own agent.

Sierra frames the interests this way: consumers want speed, dependability and trust, brands want visibility and control, and the companies building personal agents want efficiency and access.

The company also presented a reimagined Ghostwriter, described as an always-on teammate that monitors a customer's agent. According to Sierra, it can detect when APIs go down or metrics look off, evaluates the agent continuously, helps run and manage A/B tests, and uses context from the agent and from conversations in Slack or Teams to suggest opportunities teams had not considered.

Horizon, partners and what Sierra is pitching

Sierra recapped Horizon, launched in July, a platform for agents that pursue long-horizon goals such as originating a loan or getting prior authorization for a healthcare procedure. Horizon agents engage customers over days, weeks or months. A Context Engine combines information from a company's systems with memories of past interactions, and Horizon Advisors use that context to pick the next best action within goals and guardrails the company sets.

Sierra also introduced a partner ecosystem of technology platforms, marketplaces and service partners. The company says its customers include almost 50% of the Fortune 50, one in three of the leading banks, and 80% of the Fortune 50 healthcare companies.

Its central argument is that while any company can rent the same AI models, the customer context it builds is its own. In Sierra's words: "Rent the intelligence and own the context." The closing ask to businesses is to bring "your hardest problems," especially those that span many conversations, systems and weeks or months.