Sierra crossed $100M ARR 21 months after launch and raised $950M at a $15B valuation in May 2026 — that single trajectory is reshaping how every other AI customer support startup pitches investors. Decagon (now $1.5B valuation, $296M raised) and Cresta ($271M, Series D) are competing on the same outcome: replace tier-one tickets, charge per resolution, and watch contact-center seat counts compress.
The Sierra-Decagon arms race
The two pure-play conversational support agents are running parallel motions with opposite go-to-market shapes. Sierra leans on Bret Taylor's Salesforce roster — Deliveroo, Discord, Ramp, Rivian, SoFi, ADT, Cigna — and prices on resolved tickets so margin tracks deflection. Decagon won internet-native logos (Duolingo, Notion, Webflow) and posted ~$6M ARR by late 2024 against Sierra's $20M, but scaled Series B at a $1.5B valuation. Ada ($191M, $1.2B valuation), Forethought ($92M Series C), and Kore.ai ($250M Series C, unicorn) sit one tier down with longer enterprise cycles and slower per-seat compression.
The deflection metric is the entire pitch. Klarna publicly walked back parts of its AI-only customer service in 2024 and quietly rehired humans; that retreat became the cautionary tale every CX leader cites when negotiating SLAs with Sierra or Decagon today.
Voice is the next frontier
The voice subcategory is where the 2026 cheques are concentrating. Cresta ($271M, $1.6B valuation), Observe.AI ($213M Series C), and Kore.ai embed inside live agent workflows — real-time coaching, automated QA, conversation intelligence on top of Five9, Genesys, and NICE. Avoca AI ($125M Series B, $1B valuation) targets home-services dispatch. PolyAI, Bland AI, and Retell AI are pushing fully autonomous voice agents into restaurants, dental offices, and outbound dialing.
Krisp's $17.5M Series A funds noise cancellation, accent translation, and AI note-taking for the agent side of the call — a quieter moat than the agentic platforms but a genuinely defensible one. Moveworks ($305M Series C, $2.1B valuation) sits adjacent: internal IT and HR helpdesk rather than external CX, but the same deflection thesis.
What 2026 actually decides
Whether outcome-based pricing survives a recession. Sierra and Decagon only earn when AI deflects, so margins ride on accuracy at scale. If reliability dips, Zendesk, Salesforce, and ServiceNow reclaim the surface by bundling adequate AI into existing seats. The vertical plays — EliseAI in property management, Avoca in home services, Moveworks in IT — are the best-positioned to hold ground regardless of which direction horizontal pricing breaks.