Conversation Intelligence Solutions For Contact Centers

Two Leaders sit atop this Wave, and they are not selling the same product.

The Forrester Wave: Conversation Intelligence Solutions For Contact Centers, Q2 2025, is published, and its headline result is easier to misunderstand than to get right. Cresta took the highest score in the current offering category. CallMiner took the highest score in strategy. One sells what happens during the call. The other sells what the call becomes afterward.

That is two products wearing one category name, and the distinction explains more about this market than any feature list will.

The ten-provider field of The Forrester Wave: Conversation Intelligence Solutions For Contact Centers, Q2 2025

The evaluation scored ten providers against twenty nine criteria across current offering, strategy, and customer feedback. The placements made public so far run two Leaders and one Strong Performer; the full roster sits behind Forrester's paywall.

Cresta is Forrester's "force to be reckoned with." The report's own line carries the weight: enterprises succeeding with real-time guidance are likely leveraging Cresta, and few vendors come close to a competitive feature set there. Sixteen of its criteria scores hit the maximum, spanning insight discovery, real-time guidance, outcome analysis, genAI safety and controls, genAI summarization, revenue generation, redaction and masking, coaching efficiency, and innovation. The references run United Airlines, CarMax, and Cox Communications. Forrester's fit verdict is direct: enterprises serious about leveraging genAI's full potential in their contact center.

CallMiner won the strategy category, with maximum scores in vision, partner ecosystem, and adoption, plus the highest marks in the current offering criteria that matter to its thesis: insights exploration, topic and behavior classification, signal extraction, coaching efficiency, and a natural language interface. Forrester's reading of the vendor is the most useful sentence in the report: its greatest strength lies in mastering contact center workflows and translating that expertise into a flexible toolkit. The natural language interface, AI Assist, is called the cornerstone of a new agentic AI framework, and the fit verdict points at enterprises with sophisticated operations ready to evolve from collecting insights to driving impact through automated actions.

Invoca placed as a Strong Performer, and its distinction is specific enough to matter: Forrester calls it the only vendor in this market that supports the full buyer journey. Eight maximum scores run through vision, roadmap, revenue generation, insights discovery, pre-contact insights, marketing optimization, and signal extraction. It is the vendor that connects the marketing budget to the call, and it is the reason the category's borders keep drifting toward revenue.

The Q2 2025 edition follows a Q2 2024 edition that also named Cresta a Leader. The category is young, and its scorecard has already moved.

The call as it happens, the call as a data set

Hold Cresta and CallMiner side by side and the category's internal split becomes visible.

Cresta's criteria cluster around the live moment: real-time guidance to the agent, genAI safety and controls around what the model is allowed to say, revenue generation scored as a feature of the conversation itself. The product intervenes. It whispers the next step, checks itself against policy, and measures whether the intervention changed the outcome. The reference detail is the tell: enterprises succeeding with real-time guidance are likely leveraging Cresta, because that is the entire company.

CallMiner's criteria cluster around the aftermath: insight exploration, topic and behavior classification, signal extraction from millions of recordings, coaching that starts after the call ends. The product mines. Then its agentic layer takes the mined insights and turns them into automated actions. The pivot sentence in its citation is the promise: from collecting insights to driving impact through automated actions.

Neither position is the old category. The old category sold a recording with a search box. The new one sells interventions at both ends of the call: during it, and after it, at scale.

Why the scorecard cannot tell you which one to buy

Here is the honest limitation, and it is structural rather than tactical.

A Wave with two different products at the top is a scorecard of a market in motion, not a ranking of substitutes. Cresta and CallMiner both scored as Leaders because the criteria had to cover both kinds of vendor. That is fair to the vendors and hard on the buyer. The report cannot tell you whether your operation needs the intervention or the mine, because the criteria were written to score both.

The genAI safety criterion makes the stakes concrete. Forrester scored Cresta maximum on genAI safety and controls, which means the report formally rewards vendors for building guardrails around real-time AI in customer conversations. That is the right thing to reward. It is also an admission that the risk being guarded against exists: a model advising an agent in real time can say the wrong thing at full speed, and no scorecard can price the liability that lands on the buyer when it does. The criterion exists because the risk is real. The score does not transfer the risk.

There is a larger shadow over the category itself. Forrester announced in December 2025 that its separate conversation intelligence coverage would fold into the new Customer Feedback Management And Analytics Solutions market, which makes the Q2 2025 Wave likely the last standalone edition under this name. Buyers reading this scorecard in 2026 are reading a category in its twilight, still scored, still published, and already being absorbed.

The name that is not this one

One distinction is worth pinning down, because the names are almost identical.

This Wave, for contact centers, is separate from The Forrester Wave: Conversation Intelligence For Customer Service, Q3 2023. That evaluation covered eleven vendors against thirty two criteria, and it ran on different entry rules: vendors had to sell a standalone solution independent of interaction channels and workforce optimization suites. Cresta placed as a Strong Performer there, and a Leader here. Same conversations, two scorecards, different doors.

For a buyer the practical consequence is simple. Know which table you are buying at. The contact center Wave scores vendors against operations inside the center. The customer service Wave scored a different field under different inclusion rules. If a vendor leads with one scorecard in its marketing, ask for its placement in the other.

Three questions that split the field

Do you need the intervention or the mine? If your agents are live on calls all day and the money is lost inside the conversation, the Cresta profile is the one that matches. If the value sits in hundreds of thousands of recordings that nobody has the time to read, the CallMiner profile matches. The tier does not answer this. Your call volume and your failure points do.

What happens when the model whispers wrong? Real-time guidance is the category's sharpest feature and its sharpest risk. Ask any vendor, Leader included, for its incident history, its guardrail architecture, and the customer who will take the call about it. GenAI safety is a scored criterion now. Your liability is not.

Does the vendor's scorecard survive the category's retirement? The Q2 2025 Wave is likely the last standalone edition. Ask how the vendor plans to be evaluated inside the merged Customer Feedback Management And Analytics Solutions market, and what its placement here will be worth then. A Leader score in a retiring category is an asset with an expiration date.

The category has stopped selling recordings. It sells what happens during the call and what the call becomes afterward. The Q2 2025 Wave is the most precise scorecard of that split ever published. Read it as a field guide to two different products, not as a ranking of one.

Forrester does not endorse any vendor named here, and tier placement should not be read as a recommendation to buy.