Why AI is reshaping UCaaS and CCaaS pricing for Australian buyers

an Australian organisation we spoke with recently had budgeted for a contact centre platform the way it always had. Seats, a service tier, a few modules for analytics and workforce management. The numbers scaled with headcount, so finance could forecast them a year out. Then the shortlisted vendors started quoting AI features, and the tidy spreadsheet stopped working. One bundled AI summaries at no extra charge. Another sold an AI agent licence per seat. A third metered consumption by token. The same capability carried three price tags, measured three different ways, and nobody internally could say what the platform would actually cost once people began using it.

That situation is becoming common, and it has a single cause: AI changes the economics of these platforms in a way older features never did.

Why AI breaks the old pricing model

For most of the past decade, pricing here was predictable. You bought seats, picked a tier, and added modules as needed. Costs tracked headcount, so forecasting was simple. AI does not behave like another module. Every time someone uses it, the vendor pays. A meeting summary, a reply suggested to an agent, a customer conversation handled by a virtual agent, an automated analysis of a call: each triggers an inference cost on the vendor’s side, and that cost is variable. It rises with usage rather than sitting fixed against a seat count.

This puts vendors in an awkward spot. They want AI adoption to climb, because heavy use makes a platform harder to replace. They also cannot let the busiest users consume unlimited model capacity with no way to recover the spend. The answer most have reached for is a hybrid: a base subscription, plus consumption charges on the capabilities that cost real money to run.

Vendors draw that line in different places. Passive, low-cost features such as meeting notes, transcription and writing assistance often stay bundled, and several UCaaS providers treat bundled AI as a selling point. Features that take action, generate content or run a workflow tend to be priced separately. In the contact centre market this is sharper. Some platforms meter by token, some charge for AI agent licences, and usage-based platforms simply extend existing metering to cover AI work. Credit pools are appearing too, where you buy a block of capacity and different actions draw down different amounts. Summarising a recorded meeting is cheap. An AI agent that drafts a follow-up email, updates a CRM record and books the next call costs the vendor every run, and pricing reflects that.

The practical effect is fragmentation. Across Australian organisations, organisations cannot run a like-for-like comparison because the unit of measurement keeps changing. One vendor charges per interaction, another per token, a third bundles some things and meters others, a fourth quotes on agent count plus expected AI usage. Forecasting suffers for it.

How to handle AI pricing before you sign

An organisation choosing a UCaaS or CCaaS platform now needs to treat AI pricing differently from traditional software, and the work belongs early. Start by separating what is bundled from what is metered. Do not assume every AI feature sits inside the base subscription. Ask which features are consumption-based and how usage is counted. If the vendor uses credits or tokens, ask for worked examples against your own expected usage. If the vendor quotes a flat rate, ask what usage that rate assumes and what happens when you go past it.

Then model your own consumption before you commit. For agent assist, work out how many interactions a day that is and the marginal cost. For virtual agents, estimate deflection rates and the cost per deflected contact. For summaries and transcription, estimate how many meetings will request them and whether that sits inside the bundle. The gap between light and heavy AI use can move your total cost of ownership materially.

The third question is internal, and most organisations have not answered it. Do you charge AI consumption back to the departments that use it, or absorb it centrally? If you charge it back, can your systems track usage by department accurately? If you absorb it, do you have approval for variable spend and a way to pull usage back if costs run ahead of budget?

One pattern shows up repeatedly. AI pricing gets ignored until late in the selection, because vendors lead with capabilities and defer the commercials. By then the organisation has effectively chosen on other grounds and has little appetite to walk away over a pricing structure.

The question is not whether consumption-based AI pricing is fair. It is whether it fits your cost structure and whether you can forecast it with confidence. Settle that before you sign, not after, because once the contract is in place the meter does not stop.