What Genesys released
On 17 November 2026, Genesys made Standalone AI Agents generally available. The product lets an organisation run the Genesys Cloud Agentic Virtual Agent inside its existing customer service setup without moving the contact centre to Genesys Cloud. Until now, Genesys's agentic AI was mostly something customers got after a platform migration. Under the new offer, the virtual agent can sit beside a contact centre from another vendor, or an on-premises system, and use the phone numbers and carrier the business already has.
Genesys says deployment takes a few days. The agents follow business rules the customer sets, and they hand a conversation to a human with its context attached so the caller doesn't have to repeat themselves. Futurum's analysis says Genesys is aiming first at high-volume requests such as account changes, payments, claims and appointment scheduling.
Alongside the standalone product, Genesys launched Genesys Arena, a build space with prebuilt Agent Kits for more than 50 industry scenarios. Each kit comes with knowledge, integrations, skills and guardrails, and Genesys claims a working agent can be built in as little as 10 minutes. The Agent Kits are due for general availability later in November. Voice persona tools, which let builders set the tone and delivery of a voice agent, are coming but have no release date.
How the pricing works
Genesys has renamed its AI tokens to Experience Units. Each customer interaction is now counted once and put into a pricing tier based on how much tool use the agent needed to finish the job. Futurum sets out the tiers as simple at one to three tool calls, medium at four to six, and complex at seven or more. Genesys hasn't published what an Experience Unit costs, and CMSWire confirms the per-tier rates weren't disclosed.
Linking price to task effort makes a bill easier to trace back to what the agent actually did. The model still charges for consumption, though. Futurum's Keith Kirkpatrick points out that extra tool calls caused by weak integrations or retries would raise the cost of the same result. If a customer's booking system times out and the agent calls the booking API three times, the interaction can move up a tier even though the customer got the same outcome.
Why the release matters beyond Genesys customers
Most New Zealand contact centres don't run on Genesys. Plenty sit on Microsoft, Cisco, Avaya, Amazon Connect or a local telco platform, often tied to multi-year contracts. Until now, adding a capable agentic voice or chat agent usually meant either waiting for the incumbent vendor to ship one or starting a full platform replacement. Futurum reads the Genesys move as a way to sell to customers on rival or on-premises platforms without waiting for contract renewals.
For a business owner, the appeal is clear enough: an AI agent can take on one well-defined job, such as rescheduling appointments, while the rest of the contact centre stays as it is. The business case can then rest on one measurable task, without a migration project attached. Genesys hasn't hidden that it still wants those customers on Genesys Cloud over time, and Futurum notes the longer-term aim as well. A standalone agent is a reasonable way to test the technology, and buyers should treat the pilot as a possible first step in a platform sales process.
The evidence Genesys offers needs reading with care. Its cited customer results, including Electrolux Group and Riachuelo, come from full Genesys environments. None of the published results show how a standalone agent performs on another vendor's contact centre. The nearest local reference is Charles Sturt University in Australia, which already uses the Agentic Virtual Agent. We haven't seen a New Zealand deployment announced, and Genesys hasn't said which third-party contact centre platforms are formally supported or how the integration with each one works.
Where standalone deployments will go wrong
The handover from the AI agent to a person is the weakest point. Futurum says the human handoff is the likeliest failure point in standalone deployments, especially when the context has to land in another vendor's agent desktop. If the summary, the customer's identity check and the half-finished transaction don't arrive with the call, staff will ask the customer to start again. Customers will then remember the AI as the thing that wasted their time.
Data handling is the second gap. A Genesys agent running beside a non-Genesys contact centre will process call audio, transcripts and customer records in Genesys's cloud. Under the Privacy Act 2020, a New Zealand business that sends personal information offshore stays accountable for how the receiving provider protects it. The announcement coverage doesn't say where Genesys processes standalone agent traffic for NZ customers, so buyers will need that confirmed in writing.
The third gap is measurement. Vendors like to report containment, the share of contacts that never reach a person. Containment alone rewards an agent that traps customers. Futurum recommends tracking resolution, escalation and failure reasons, and we'd add repeat contact within seven days as a check on whether a "resolved" interaction really stuck.
What NZ businesses should do next
Organisations thinking about Genesys's standalone agents, or any vendor that follows with a similar offer, can prepare before a sales conversation starts:
- Pull three months of contact data and find the two or three request types with the highest volume and the simplest resolution. Appointment changes and payment queries are usually better first candidates than complaints.
- Count how many systems an agent would have to touch for each request type. Each lookup or update is a tool call, and the count gives an early guide to which Genesys pricing tier the request would fall into.
- Ask Genesys for published Experience Unit rates, any spending cap or committed-use option, and a worked cost estimate based on your own volumes before you sign anything.
- Get written confirmation of the hosting region for audio, transcripts and logs, and which subprocessors handle New Zealand traffic.
- Test the handoff into your current agent desktop during the pilot, with real staff, and agree on what context has to arrive before the pilot can count as a success.
- Check your existing contact centre contract for clauses on third-party AI access to routing and call recording, since incumbent vendors may restrict such access.
- Decide who owns the agent's rules and escalation thresholds once the pilot finishes, and record those decisions in your change process.
Customer-facing agents need the same permission design and audit trail as internal ones. We cover how we approach both on our AI agents page, and our AI consultancy work deals with the governance and vendor questions that come before a contact centre pilot.
Sources
Want to know what this means for your organisation? Our team in Auckland can show you your top three opportunities in a free 30 minute assessment.
Get my free assessment How we help