Horizon 代理协调多天或多周的出站和入站互动——不仅仅是单次对话。上下文引擎和长远规划使每一次互动都成为叠加优势，因此随着客户关系的加深，你的代理也会变得更智能。这就是 Sierra 的成果导向大规模模型：你付费的是结果，而不是令牌。
Sierra 正在重新构想面向代理时代的软件——你只需描述目标，智能代理就会为你构建、执行并持续改进工作。认识 Ghostwriter，这个能够创建和优化其他代理的代理，将你的想法变成可投入生产的客户体验，无需点击、编码或复杂操作。
我们正在向新老投资者募集 9.5 亿美元，估值超过 150 亿美元。
Get notified about new product features, customer updates, and more.
The debate in enterprise software has shifted from what can I do with AI? to where does my competitive advantage lie? Or, as Satya Nadella, CEO of Microsoft, asked in June: how do companies avoid "ceding value" to a handful of models?
Meet Context Engine：/product/context-engine, which turns your customer relationships into your competitive advantage. It powers long-running Horizon agents：/blog/horizon that pursue business outcomes over days, weeks or months, with every interaction adding more context and making the next one smarter.
When a customer calls to cancel, the context you need to make the right offer probably already exists somewhere in your business. The question is whether your agent can find it in the thirty seconds that matter. That context comes from two places:
Having access is not the same as having context. Hand an agent every record in your business and you've only moved the problem one step closer to the customer. Most data means nothing in isolation. Device telemetry is just a health ping until it's connected to the order that shipped the device, the customer who bought it, and the setup ticket that's still open. Then it becomes the right moment for the agent to step in.
With all this information, knowing what matters when is the hardest problem. Every business has millions of potential relationships across customers, products, transactions and interactions. Context Engine learns which pieces of context matter for which decisions, using the outcomes of every interaction to get better over time at surfacing what matters.
Context tells an agent what is true. It does not tell it what will work. A subscription service, for example, has to detect rising cancel intent — like a price increase paired with a drop in engagement — and decide which save offer to lead with, or whether the customer needs one at all.
A retailer has to decide what to recommend, when to hold back, and when an apology will do more than a lower price. Those answers come from outcomes. Every decision an agent built on Sierra makes becomes evidence about what worked, for which customer, in which situation.
Context Engine treats every decision as an experiment. Most decisions are based on what has worked best for similar customers in similar situations. But a small number deliberately explore promising alternatives, because the only way to discover something new is to try. Exploration may cost a little in the moment, but it's also what keeps the agent learning instead of settling for the first decent answer.
The self-improvement loop looks for the next opportunity to improve outcomes. Sometimes that means discovering new context. Customers with large unspent loyalty balances, for example, keep declining discounts and churning anyway. Intuition would suggest a growing balance signals loyalty, but the evidence shows it’s a sign of disengagement. So the agent learns a new rule: lead with re-engagement instead of price.
Other improvements are statistical. As more evidence accumulates, Context Engine can train models that help agents make better decisions: predicting which customers are likely to churn, which offers are most likely to be accepted, or which leads are most likely to convert.
The business decides the outcomes that matter: saves, lifetime value, qualified leads. Context Engine does the rest. By morning, it knows a little more about what keeps your customers than it did the night before.
The foundation models cannot be your advantage. Off the shelf, they’re identical to what your closest competitor has, and you both get the next release.
What is unique to your business — and what you own entirely — is the observations your agents collect, the evidence of what works and for whom, and the decisions that lead to better outcomes. That record is your moat, one your competitors cannot buy or shortcut. All they can do is start building their own one outcome at a time.
Over time, it makes every part of your business smarter, delivering better customer experiences, improved outcomes and stronger growth. That's the advantage Horizon agents act on — turning what Context Engine knows into outcomes pursued over days, weeks, and months.
Find out how Sierra can help you deliver better outcomes with AI.
Horizon agents orchestrate outbound and inbound interactions over days or weeks — not just single conversations. A context engine and long-horizon planning turn every interaction into a compounding advantage, so your agents get smarter as your customer relationships deepen. It's Sierra's outcomes-based model at scale: you pay for results, not tokens.
Sierra is reimagining software for the agent era—where you simply describe the outcome, and intelligent agents build, execute, and continuously improve the work for you. Meet Ghostwriter, the agent that creates and optimizes other agents, turning your ideas into production-ready customer experiences without clicks, code, or complexity.
We’re raising $950 million from new and existing investors, at a valuation of over $15 billion.
扫码直达文章详情