Genesys AI agents work without cloud migration

Genesys has launched AI agents that companies can run inside the contact centre they already have, removing what has been one of the biggest barriers to adopting agentic AI: a full platform migration first.
The company announced Genesys Standalone AI Agents on 17 November 2026 at Genesys Xperience in New York. The offering is generally available now. It lets organisations deploy the Genesys Cloud Agentic Virtual Agent alongside their existing contact centre, keeping their current phone numbers, carrier, human teams and workflows in place.
Start with one customer need
Genesys says the starting point changes from an infrastructure decision to a customer decision: which need should AI solve first. That could be changing an account, servicing a payment, progressing a claim, scheduling an appointment, or handling another high-volume request where the customer expects an outcome rather than just an answer.
The agents are powered by Large Action Models, AI models designed not only to understand what a customer is asking but to reason through the work required and act, using approved tools across enterprise systems. The action happens inside controls the business sets, with approved tools, permissions, policies and guardrails, and an interaction can be handed to a human employee with the context intact when judgement is needed.
Arena offers a working agent in 10 minutes
Alongside the standalone agents, Genesys introduced Genesys Arena, a build space with prebuilt Agent Kits covering more than 50 industry use cases, plus working prototypes, voice personas and workflows. Genesys says teams can create a working AI agent in as little as 10 minutes, customise it in natural language, preview the experience and deploy it. Agent Kits will be generally available through Arena later in November.
Pricing by effort, not by message
Genesys also introduced effort-based pricing, charging according to the work each interaction requires rather than treating every conversation the same. Interactions are classified by the number of tool calls they need: simple means one to three tool calls, medium means four to six, and complex means seven or more. Consumption is measured in Genesys Cloud Experience Units, the new name for what were previously AI Tokens.
The company pointed to results from its agentic AI deployments in broader Genesys environments as evidence of what the underlying technology can do. Electrolux Group has more than 65% of service demand supported by agentic AI, with 100% containment in select journeys. Riachuelo reported 84% self-service resolution, against 30% with its previous traditional chatbot, alongside a 60% increase in customer satisfaction. Trademark Global reported a 111% increase in referral-code usage, 42% containment and a 91% success rate for warranty claims.
Genesys says the aim is governed autonomy: enough independence for AI to resolve more customer needs, while the business stays in control of what it can do.