Genesys Cloud AI Experience tokens explained: allowances, rates and prices

Genesys Cloud CX meters most of its AI in one unit: the AI Experience token. Rather than pricing each feature separately, Genesys measures each one in whatever unit suits it, such as minutes for voice bots, interactions for virtual agents, users for Agent Copilot or evaluations for AI Scoring, and converts that usage into tokens at a published rate. Tokens used across a billing cycle are added together and rounded up once, and usage above the organisation's allowance is billed monthly in arrears, whether the contract itself is monthly or annual.

Not everything consumes tokens. Genesys states that some AI capabilities, such as AI Experience Orchestration and Predictive Engagement, do not, and AI Guides stopped incurring token charges from June 2026. Third-party engines are different again: a voice bot from another vendor is not metered in tokens at all, but carries a per-minute Genesys platform charge on top of the vendor's own bill, which our voice AI cost calculator brings together.

Every Genesys Cloud organisation receives a monthly fair use allocation of tokens, set per organisation rather than per licence, and it is the same whichever of CX 1 to CX 4 your users hold. Organisations on named licences receive less than those on concurrent licences. The allocation renews each month and does not carry over. Genesys Cloud CX 4 then adds an extra monthly allowance for each CX 4 agent, which is how core Agent Copilot features are included on that tier.

In a small organisation the allowance can cover a pilot. In a larger one it is often used up by per-user features before any bot is counted, because Agent Copilot and speech and text analytics consume tokens for every user who has them, every month. That is why we always model the allowance as shared: the question for a bot or virtual agent is not how many tokens it uses, but how many it adds on top of everything else the organisation already consumes.

• Free allocation, named licences: 250 tokens per organisation per month
• Free allocation, concurrent licences: 350 tokens per organisation per month
• Genesys Cloud CX 4: an extra 30 tokens per named agent, or 39 per concurrent agent, each month
• Unused allocation does not carry over to the next month

Each feature converts into tokens at its own published rate. The table shows the current rates from Genesys's token consumption page. Two details matter most in practice. Voice bot usage is measured per call and rounded up to the next fifteen seconds before conversion. And an agentic virtual agent is charged at its rate as soon as an interaction enters a flow that contains an agentic virtual agent block, even if the call never reaches that block, so flow design affects cost as well as experience.

Genesys publishes a month-to-month list price per token in nine currencies, together with the per-minute platform charges for third-party voice bots. The table below shows the pound, dollar and euro figures. These are list prices. Organisations that prepay tokens monthly or annually are invoiced at the rate in their services order, legacy AI Experience bundles continue until renewal, and invoices show an average rate across free and paid tokens, so the figure on your invoice can look lower than the list price even when nothing has changed.

Published by each vendor and checked on 14 October 2026. These are list prices, before any discount, tax or contract terms, and they change; the sources at the end of this page are authoritative. QVCCS does not sell licences.

Take 25,000 calls a month that each spend four minutes with a voice bot: 100,000 minutes, which convert at 17 minutes a token into 5,883 tokens. The same 25,000 calls through a Genesys Virtual Agent use 12,500 tokens, at half a token each, and through an agentic virtual agent 30,000 tokens, at 1.2 each. Minutes drive one model and interactions drive the other, so the comparison changes completely with call length.

The break-even points make this concrete. A Virtual Agent interaction costs the same tokens as about eight and a half minutes of voice bot, and an agentic virtual agent interaction the same as about twenty minutes. For short, structured calls, a scripted bot flow is the cheapest native option by a wide margin; for conversations that need reasoning across systems, the agentic agent's flat charge per interaction can be the better buy. Now add the allowance: twenty Agent Copilot users on named licences consume 800 tokens a month on their own, more than three times a named organisation's free allocation, so every bot token in that organisation is a paid one.

Tokens make AI easy to switch on and easy to lose track of. We treat consumption as a design input, not an afterthought: we estimate it from expected volumes before a feature goes live, decide whether prepaying is worthwhile, and review actual consumption against the estimate in the first billing cycles. QVCCS does not sell licences or tokens; our part is the model, the design choices that drive it and the evidence for your commercial conversation with Genesys or your reseller.

1. Count per-user consumers first: Agent Copilot, speech and text analytics and any CX 4 allowance
2. Estimate each bot or agent from expected interactions and, for voice bots, minutes per call
3. Keep agentic virtual agent blocks out of flows that do not need them
4. Compare native routes with third-party ones, including the Genesys platform charge
5. Decide between prepaying and paying overage, then review the first invoices against the model