UCaaS and CCaaS Vendors Navigate Rising AI Inferencing Costs

UCaaS and CCaaS Vendors Navigate Rising AI Inferencing Costs

Enterprise communications vendors are rethinking their pricing strategies as the cost of powering AI features becomes harder to predict. The shift affects unified communications as a service (UCaaS) and contact center as a service (CCaaS) providers that rely on large language models for capabilities like virtual receptionists, real-time transcription, and agent assistance.

The pressure stems from volatility in the foundation model market. Uber reportedly exhausted its planned AI budget just months into 2026, while Anthropic changed its pricing model in April and OpenAI is reportedly considering significant price cuts to compete. Because UCaaS and CCaaS vendors use these models for inferencing, rising costs flow directly to their bottom lines.

"Vendors want to promote AI adoption because AI makes their platforms more valuable," wrote Kevin Kieller, co-founder and lead analyst with EnableUC. "At the same time, vendors cannot afford to let heavy AI users consume unlimited model capacity without some mechanism to recover those costs."

Hybrid Models Replace Simple Bundles

The result is a move away from flat per-seat pricing toward hybrid models that combine base subscriptions with consumption-based charges. Zoom introduced a hybrid seat-and-usage model when it launched ZoomMate, while 8x8 has gradually incorporated usage-based pricing over recent years.

RingCentral outlined its approach during a presentation at the Mizuho Technology Conference 2026, first reported by No Jitter. Devang Shah, RingCentral's Senior Vice President of Growth, said the company offers both pricing structures depending on customer preference. Its AI Receptionist uses usage-based pricing, while ACE, RingEX, and RingCX are priced per seat.

"Many customers want simplicity," Shah said. "They want to know what they're going to pay at the end of the month, and some customers want usage-based pricing. And so, we are offering both type of products or type of pricing mechanisms in our products."

Cost Mitigation Strategies

Vendors are deploying technical approaches to reduce inferencing expenses. Small language models, which are narrower in scope and can run on commodity GPUs, help avoid expensive cloud API calls. Beth Schultz, VP of Research & Principal Analyst at Metrigy, noted that SLMs "step in where speed and focus matter most — plus drop the cost of per-token spend."

Some platforms use federated AI architectures that route requests to the most cost-efficient model capable of handling a given task. Zoom's approach allows its platform to select among multiple models based on accuracy and cost trade-offs.

Hidden Costs Beyond Tokens

While token costs dominate current discussions, an EY report published in June 2026 argued that organizations overlook significant expenses. The real economics of agentic workflows include infrastructure, operations, personnel, risk management, and engineering workarounds for AI limitations.

EY cited an example where a customer service AI assistant built two years ago cost $0.04 per chat but now involves tool retrieval, planning, and subagents that push the cost to $1.20 per orchestration. The report identified governance burden, organizational change costs, failure recovery expenses, and potential AI taxes as budget items frequently omitted from planning.

Why it matters

The pricing uncertainty around AI features creates strategic risk for enterprises evaluating UCaaS and CCaaS platforms. Organizations building business cases on current AI economics while reducing human expertise may find themselves exposed if model pricing shifts or usage scales faster than anticipated. The transition from predictable per-seat costs to hybrid consumption models also complicates budget planning and vendor comparison.

RingCentral's Vlad Shmunis noted during the Mizuho event that the company is using its own AI products internally, resulting in productivity gains that have reduced hiring needs. "People are able to deliver a lot more and…people who are there at the company are being more productive and more efficient," he said.

These details were first reported by No Jitter.