Sierra crosses $200M ARR and earns FedRAMP High certification, opening the door to federal government contracts

Sierra has crossed $200M ARR and earned FedRAMP High certification, opening access to federal government agencies for the first time.

Taylor frames the government opportunity around a familiar tension: citizens want better services, but the national debt makes staffing-heavy solutions untenable. Agencies running large call centers — Medicare, Medicaid, the Department of Veterans Affairs, the passport agency — are structurally similar to Sierra's existing insurance and financial services customers like Cigna and Blue Cross Blue Shield. FedRAMP High is the certification that makes the best-of-breed commercial stack available to that market. Sierra is already working with a small number of agencies Taylor says he can't name, and with FINRA as a regulated-industry reference customer.

“We just crossed 200,000,000 in ARR... Today we announced we're a certified FedRAMP high, which really simply put means now federal government agencies have access to Sierra... From concept to 100% of their phone calls [for Nordstrom] was thirty-five days.”

Speed to value

The pitch to government buyers rests partly on implementation speed. Sierra measures pilot-to-production in days, not weeks. Nordstrom went from concept to handling 100% of their phone calls in 35 days, starting at 1% of volume in the first four weeks. Fiserv broke the record more recently, though Taylor says he's not at liberty to disclose the exact number.

Outcomes-based pricing underpins the deployment urgency. Sierra charges only when it successfully resolves a call or completes a sale, which means the company doesn't get paid until the system is live. Taylor argues this removes the usual vendor-customer friction and reorients both sides toward go-live rather than the sales process.

The multilingual case

One of the more concrete arguments for AI in government call centers is language coverage. Staffing a call center to serve Spanish, Mandarin, Tagalog, and other languages across the US population is expensive and logistically difficult. With AI, Taylor says each additional language carries effectively zero marginal cost at scale.

Greeting acceptance and the trust deficit

Sierra tracks what it calls "greeting acceptance" — the share of callers who get past the initial AI greeting and actually engage. The metric is rising, but Taylor acknowledges the industry is still paying down roughly a decade of bad bot experiences. His read is that caller sentiment will flip within three to four years, with people eventually preferring AI agents over hold times the way they now prefer OpenTable over calling a restaurant directly.

Model landscape

Taylor is bullish on the frontier, arguing the gap between the best models — he cites GPT-4.5 and recent Anthropic releases — and open-source alternatives is widening, not closing. But he expects a constellation of purpose-built, specialized models to emerge alongside the frontier, with edge deployment becoming relevant for voice as hardware improves. He speculates, without direct knowledge, that Apple's privacy posture makes on-device inference a natural fit for this use case.

The $200M ARR figure is the commercial proof point Sierra needed before going after federal contracts. FedRAMP High is the regulatory unlock. Whether government agencies move as fast as Nordstrom and Fiserv will depend on how quickly agency leadership commits — Taylor's Fiserv example suggests top-down sponsorship is the variable that matters most.