Customer Story
How a Tier-1 telecom scaled multi-channel, AI-driven commerce on Akka — 7.5× the orders, on half the hardware.
Verizon replaced a decade-old Oracle commerce platform with Akka and put 7.5× more orders through half the hardware — from 1,600 to 12,000 orders per minute, with response times cut from 6 seconds to 2.4. Total cost of ownership fell by half, sales rose 235%, and Akka absorbed iPhone-launch and holiday peaks with no parallel-site workaround.
Akka ROI Scorecard
less build effort. Overnight builds dropped to seconds; emergency fixes from 24 hours to same-day; features shipped continuously.
the total cost of ownership. Hardware footprint halved, serving the same and growing order volume on a fraction of the infrastructure.
the throughput — 1,600 to 12,000 orders/min — with response time cut from 6s to 2.4s, through peak events.
Business outcome: sales +235%, conversion rate +197%.
Verizon ran online sales on an Oracle commerce platform deployed more than a decade earlier. As demand moved online and new channels emerged — Alexa, Google Home, Facebook chatbots — the platform hit hard limits: new features took an average of two months to ship, software builds ran overnight, standing up a test or production environment took 5–10 days, and an emergency production fix took 24 hours.
Worst of all, it could not hold up under peak load — holiday sales, major iPhone launches. Every year Verizon spent roughly six months rebuilding a stripped-down copy of the storefront — logging disabled, complex transactions removed — that ran for only a few days. The workaround was expensive, brittle, and did not scale.
Verizon's architecture team evaluated the major platforms over two months and standardized on Akka for its scalability, hot reload, operational simplicity, and real-time responsiveness. The deciding factor was efficiency: Akka ran the same commerce workload on a fraction of the infrastructure, with services that restart in seconds and stay responsive under peak load.
This is the Never Fail guarantee: Akka handles clustering, resilience, durable in-memory state, and traffic steering, so the application does not have to.
Scale. Throughput rose from 1,600 to 12,000 orders per minute — 7.5× — in testing, with room to go further. Customer response time dropped from 6 seconds to 2.4. Verizon could now serve its highest-traffic events on the primary platform and retire the annual parallel-site rebuild entirely.
Cost to operate. Total cost of ownership fell by half. The hardware footprint was halved, cutting both machine and human operational overhead for the same — and growing — order volume.
Speed to production. Build effort dropped 40%. Overnight builds became second-scale; day-long emergency fixes shipped same-day; new features moved from a two-month cadence to continuous delivery.
Business impact. Sales increased 235% and conversion improved 197%.
Verizon's growth channels — Alexa, Google Home, and chat-based ordering — are exactly where agentic commerce now lives: conversational surfaces where an AI agent takes the order, personalizes the offer, runs eligibility and fraud checks, and orchestrates fulfillment in real time. On the Akka Agentic AI Platform, those channels become governed AI agents running at the same 12,000-orders-a-minute scale Verizon already proved — with durable state that survives failure, sub-minute recovery, and half the infrastructure.
The platform that scaled Verizon's commerce is the same platform that builds and runs agentic AI today. Enterprises build on the Akka Agentic AI Platform directly, or have a system delivered and operated through Akka Specify. Both provide the production guarantees Verizon has run on for years.
