The job
When you're a giant company, customer support is a monster. Millions of customers, questions at 3 AM, returns, billing disputes, account changes — and every interaction either builds loyalty or burns it. Hiring enough humans to cover it all, in every language, around the clock, is basically impossible at scale. Sierra was founded to solve exactly this: AI agents that don't just chat, but actually do things — check an order, process a return, change an account, book the appointment. So basically, the difference between a chatbot that says "I understand" and an agent that fixes the problem while you're still on the line.
The agent
Sierra is an enterprise conversational AI platform founded by Bret Taylor and Clay Bavor. Companies build AI agents on it that handle voice and chat across the whole customer journey — not just answering questions but completing multi-step tasks inside the company's real systems, with guardrails on every action. It charges per resolved outcome, like Intercom's Fin, so the pricing follows the results. The company has raised $950M at a $15.8B valuation and says over 40% of the Fortune 50 use its software (company figures, 2026 — vendor-reported).
Results
The named deployments are the strongest evidence. SoftBank deployed Sierra's AI at LINEMO, its online-only mobile brand in Japan: inquiry resolution rose from 83% to 97% and customer satisfaction from 74% to 93% (TechInAsia, July 2026, sourced from SoftBank). Virgin Media O2 launched a Sierra-powered AI voice agent for broadband fault calls in 2026, rolling it out gradually while keeping human transfer always available (Customer Service Manager, 2026). Sierra also published a "Personal Agent Protocol" with 35 design partners including OpenAI, Visa, Target, and United Airlines (PYMNTS, Oct 2026). These are real, named, recent deployments — the strongest sourcing in this entire directory.
The catch
Enterprise agent deployments are expensive, slow, and complicated — this isn't a $20/month tool you turn on. Weeks-to-months of integration work, journey mapping, and guardrail design before anything goes live. The outcome-based pricing sounds fair but shifts risk in ways procurement teams have to model carefully; when volume is high, the bill is high. And the LINEMO numbers, while impressive, come via the company and its partner — great case studies, but the companies involved have every reason to present the rosiest version. Also worth noting: even Sierra is now building detectors for AI callers (its Fleming-1 model), because a world full of agents means agents talking to agents — a weird new problem nobody fully understands yet.
Who it's for — who should skip
If you're an enterprise with massive contact volume and complex systems to integrate, Sierra is the category leader — the SoftBank and Virgin Media O2 deployments are the proof points to study. Skip it if you're a small or mid-size business; the cost and implementation effort are built for giants, and tools like Fin or Gorgias fit smaller shops better. And skip any vendor that promises this without the integration work. The agent is only as good as the systems it's wired into.