Sierra
Outcome pricing took Sierra from $100M to $200M ARR in a year, but the receipt is a black box: no rate card, a ~€130K+ platform floor, and roughly €1.30 per resolved conversation negotiated behind an NDA.
| Annual platform minimum | ~$150,000-750,000+/yr | custom, outcome-based; scales with volume/channels/complexity |
| Implementation/setup | ~$50,000-200,000 | one-time, 4-10 week typical onboarding, up to 7 months for large rollouts |
| Per-resolution fee | ~$1.00-2.50 (reported ~$1.50 typical) | negotiated per contract; unresolved/escalated conversations typically don't trigger a charge |
Fully custom, sales-led, no public rate card. No included helpdesk/ticketing — runs as an AI layer above existing CX infrastructure (Zendesk, Salesforce Service Cloud, etc.).
checked 2026-07-29 · vendor pricing page
Element scores
Strengths
Agent Data Platform gives it a real memory layer most competitors lack, the model-agnostic "constellation" approach avoids single-vendor lock-in, and Sierra Research built τ-bench — the benchmark the rest of the industry now evaluates agents against. Case studies are backed by real speed-to-live numbers: Nordstrom's voice agent in five weeks, Cigna cutting patient-auth time 80% in eight weeks.
Honest dings
Zero pricing transparency — no rate card, no self-serve, a six-figure floor before a single resolution is billed. G2 reviewers flag context loss in long conversations and generic responses; the December 2025 Gap.com incident showed guardrails are only as good as their configuration. Multi-model orchestration adds real latency risk on live voice.
Sources
Every audit lists the research it rests on — transparency and traceability are the product. Tools evolve: each audit is a snapshot of its audit date, and re-audits supersede older versions (kept below for reference).
- sierra.ai/product/agent-sdk — Official Agent SDK docs: goals/guardrails, skills, systems integrations, testing (accessed 2026-07-29)
- sierra.ai/blog/agent-os-2-0 — Agent Data Platform (memory/context layer) and Agent Studio 2.0 announcement (accessed 2026-07-29)
- sierra.ai/blog/constellation-of-models — Official architecture explainer: 15+ models, routing, redundancy/failover (accessed 2026-07-29)
- sierra.ai/blog/enterprise-grade-agents — Official: deterministic guardrails, supervisory agents, input/output filtering (accessed 2026-07-29)
- sierra.ai/blog/building-sierras-mcp-gateway-an-engi… — Official: internal MCP Gateway, 45 services, 89% employee adoption (internal tooling, not customer product) (accessed 2026-07-29)
- sierra.ai/blog/publish-to-chatgpt — Official: customer agents natively MCP-compatible, one-click publish to ChatGPT Apps (accessed 2026-07-29)
- valueaddvc.com/blog/how-does-sierra-ai-make-money-o… — $200M ARR, $15.8B valuation, ~$1.50/resolution pricing model breakdown (accessed 2026-07-23)
- sierra.ai/blog/announcing-our-partnership-with-soft… — Official customer-share stats: 40% of Fortune 50, 1-in-3 top banks, revenue-tier breakdown (accessed 2026-07-14)
- sierra.ai/about — Official: founders (Bret Taylor, Clay Bavor), HQ San Francisco, office list (accessed 2026-07-29)
- litmustools.com/review/sierra — Independent review: 78/100 score, Dec 2025 Gap.com guardrail-misconfiguration incident, latency/consultancy-dependency complaints (accessed 2026-06-27)
- github.com/sierra-research/tau2-bench — Sierra Research's own open-source τ-bench/τ²-bench evaluation framework and leaderboard (accessed 2026-07-29)
- eesel.ai/blog/sierra-reviews — G2-sourced user complaints: pricing opacity, context loss in long conversations, limited self-service editing (accessed 2025-07-28)