NUMBERS STORIES So here’s what we’ve got File on 7/10: AI adoption in the contact center (numbers article) (Matt) # numbers that show the importance of data security — July 10 # numbers on how AI changes data infrastructure — July 13 File on 7/13: Genesys state of CX report: briefing for next week 7/14 (Matt) File on 7/14: Creatio briefing today for agentic AI and customer service (for 7/15) (Matt) # stats that show the data literacy gap (Or I can do # stats that show why data literacy is important for AI literacy) — July 16
The Gartner Magic Quadrant (MQ) for Conversational AI Platforms has been published. Like clockwork, as soon as it hits the wire, the market stops to see where the dots landed. Conversational AI has come a long way since the IVR of the 1980s. The rigid menus are gone, and instead of pressing 1 or another digit, we can just naturally converse with a machine via voice or text across one or more channels. It wasn’t long ago that the only thing I would say is “operator,” but these systems today truly do enable functional self-service.
While some of these systems clearly have telephony roots, conversational AI platforms can be obtained from a comms provider, CRM provider, hyperscaler, and many more, even new categories. They can be configured to answer questions from a knowledge base or to perform complex operations with back-end systems. Despite all this innovation, we are still early. The modern generative AI chatbot is only about four years old and the agentic generation is less than a year old. Some experts believe enterprises will attempt to purchase one conversational AI platform for every use case; others expect specialization and integrations will result in a platform per use model.
I was eager to see what Gartner had to say about the current crop of conversational AI platform options. Unfortunately, enterprise buyers looking for clear, actionable insights regarding their platform investment might find this particular report a bit difficult to parse.
I should start by saying I am a fan of the MQ. Gartner analysts are well-informed, and they have excellent access to the vendors and enterprise buyers. The MQ graphic itself is a brilliant, enduring invention. It manages to boil down the sprawling, messy complexities of vendor nuances into a simple, two-axis graphic. It is the perfect conversation starter for the boardroom.
But there is an inherent problem with MQs that we need to address: they don’t show their work.
Defining the Disneyland Problem
To understand why this is a problem, ask yourself: “What is the best ride at Disneyland?” It is an impossible question to answer without first defining your priorities. Are you looking for high-octane thrills, or are you prioritizing nostalgia and originality? Is the length of the line a deciding factor? A Small World and the Haunted Mansion are both spectacular in their own right, but you cannot compare them effectively without shared specific objectives. You can’t just announce Pirates of the Caribbean without explaining the logic behind your selection.
That is the trouble with the MQ. It is a brilliant, simple report, but it lacks the transparency of a “show your work” process. Often, the graphic and the text seem disconnected. MQs are meant to be much more than a simple product review — in fact, they aren’t product reviews at all. They are comprehensive assessments of sales and marketing effectiveness, international reach, business viability as well as some product features. All of these disparate data points end up compressed into a single dot, stripped of the criteria and context that a buyer actually needs to make a decision.
The best MQs offer clues in the “Strengths” and “Cautions” sections, but connecting those clues to the dot’s position can be a leap of faith.
Digging into the 2026 trends
In this Conversational AI MQ, the authors identify three emerging conversational AI trends: agentic AI, multimodality, and self-learning. These are absolutely the right trends. I agree with them wholeheartedly, particularly multimodality — every time I see a demo that integrates an image, sound and text, I’m impressed.
However, the alignment between these trends and the report’s substance is unclear at best. “Multimodal” only appears in the 43-page report under three vendors: Google (Strength), Netomi (Caution) and PolyAI (Strength). “Self-learn” is even scarcer, appearing only under Omilia as a roadmap feature.
Gartner does indicate areas of high or medium weighting within the report. Marketing, for example, received medium weight this year. Yet, “marketing” appears in the report 34 times. Many vendors were called out for strong or weak marketing efforts. For example, Sprinklr received a Caution for its “overall marketing strategy, specifically in the areas of relevancy, content differentiation and channel utilization.” That’s a critique, sure — but does that really help an enterprise buyer decide if the platform will solve their CX challenges?
What stood out
There are, however, moments where the report cuts through the noise with surprising—and useful—feedback:
Google’s Infrastructure Constraints: Google did very well in this report, but there was a surprisingly significant Caution noted: “Gemini Enterprise for CX runs exclusively on Google Cloud infrastructure and is unavailable for isolated, air-gapped deployment, or for being deployed on competitors’ clouds.” That’s a disqualifier for a large segment of the market.
The “Frankenstack” Treatment: SoundHound did surprisingly well in this report, particularly considering it has acquired companies like Amelia and Interactions. Usually, Gartner is notoriously critical toward “Frankenstacks,” i.e. platforms built through rapid acquisition. Here, however, SoundHound received praise for “expanding through strategic acquisitions.”
Strong Strengths: Many of the Strengths and Cautions seem distant from the product, but Cognigy Nice Strengths were all product-specific: higher service levels, vertical and industry expertise, and a high percentage of revenue toward R&D.
Pricing Nuance: While MQs rarely evaluate price/value, they do look at pricing models. Salesforce received a Strength for “differentiated and transparent pricing,” but also a Caution that buyers should ensure they fully understand its pricing and packaging implications. It’s an odd set of comments as they appear to contradict each other.
The Mid-Market Crowd: The report covers 14 providers, but about half are clustered in the center of the graphic. If you are looking for clear separation between the leaders and the pack, this report suggests that the market is still very much in a “show me” phase.
So which vendor is dominating the market?
It is anyone’s game right now. We are particularly early in the deployment of customer-facing agentic AI applications. There are significant differences in provider sizes and how they go to market. There are also some notable differences in languages (and vertical lingo) supported. Providers like Amazon, Salesforce, and SoundHound are still acquiring conversational AI companies (and their attendant talent and technologies). We are going to see significant changes in this report over the next few years, and hopefully more explanation as to what’s driving the dot placements.
On the other hand, the MQ graphic tends to obscure that the technical differences between many of these solutions are often less significant than they appear. MQ graphics don’t have scales, so the significance of the distance between dots is hard to evaluate. I do like the spirit of the MQ as non-product differences, such as go-to-market strategy, pricing, channel support, and integrations, can make or break an enterprise project. It is frustrating the report doesn’t reveal more of what the analyst observed.
The Conversational AI MQ offers some useful information, but do not mistake the dot for the destination. Before you make a decision, look at other sources, evaluate your own internal priorities, and remember: the best ride at Disneyland depends entirely on what you want to experience.
Dave Michels is a contributing editor and Analyst at TalkingPointz.