Conversational AI Platforms (Gartner)

Gartner published the Magic Quadrant for Conversational AI Platforms on 7 July 2026, credited to Gabriele Rigon, Justin Tung, Arup Roy, Adrian Lee and Uma Challa, with a companion Critical Capabilities report carrying the same date and the same five names in a different order. Fourteen vendors were evaluated and sorted into four Leaders, two Challengers, three Visionaries and five Niche Players.

The four Leaders are Google, Salesforce, Kore.ai and SoundHound AI. That rung is the smallest this market has had, and it is the first edition in which the top of the quadrant and the top of Forrester's equivalent list disagree about a vendor by three months and one quadrant.

The Magic Quadrant for Conversational AI Platforms scored fourteen vendors and kept four at the top

The field is the smallest part of the story. Fourteen vendors into four quadrants means five of them were Niche Players, and Gartner's own framing of the market has changed to match: the firm describes conversational AI as entering early maturity, with the grounds for differentiation moving away from raw model capability and toward operational stability, pricing clarity, governance and acquisition risk.

That is a criteria shift with consequences for buyers. A market scored on innovation rewards the vendor with the most interesting demo. A market scored on operational stability rewards the vendor whose platform survives a bad week, whose pricing can be explained to a finance team, and whose governance controls can be shown to an auditor. The same fourteen platforms get ranked differently under the two rubrics.

The companion Critical Capabilities report is where that shows up in detail. It scores the same vendors against four named use cases, and the aggregate quadrant position does not predict the use case result.

Gartner says the market is entering early maturity, and the criteria moved to governance and pricing

Two things moved at once. Gartner moved the criteria, and the vendors moved to meet them. The clearest evidence is what the quadrant rewarded this year against what it rewarded last year.

Salesforce entered and went straight to a Leader placement, on the strength of Agentforce, a catalog of more than three hundred prebuilt agent templates, transparent tiered pricing, and vertical packaging such as Agentforce Health. The pricing point is the tell. In a market Gartner now describes as maturing, a vendor that publishes how the meter runs is answering a question buyers are actually asking, and one that leaves it to a sales conversation is not.

The caution attached to Salesforce is equally specific: it relies on third parties for speech recognition and speech synthesis. In a category named for conversation, the layer that hears and speaks is the layer the vendor does not own. That is the shape of the risk across this market. The reasoning layer has commoditised quickly and the voice layer has not.

Gartner also notes an appetite problem at the top of the stack. Forrester's own reading of the same market found what it called zero appetite among buyers for fully autonomous agentic applications, and most of the leading platforms have responded by supporting modular autonomy, where a buyer can dial how much the system decides on its own. That is a product decision driven by a procurement reality rather than a technical one.

Google holds furthest on vision and highest on execution for the second year running

Google is the only vendor in this report leading both axes. Its own account of the placement says it was positioned furthest on Vision and highest on Execution, a Leader for the second consecutive year, and ranked first in three of the four Critical Capabilities use cases.

The stated reason is worth reading carefully, because it is not a reason a software vendor can copy. Gartner credits advances in Google DeepMind's models, the multimodality of those models, and proprietary chip infrastructure as a scalability advantage. Two of those three are inputs a competitor has to buy rather than build.

That is the structural asymmetry in this market. A conversational AI platform is a product assembled from a model, a speech stack, an orchestration layer and an integration surface. Google owns the model and the silicon and buys almost nothing. Most of the field buys all four and competes on how well it assembles them. The firm leading on both axes is the one whose supply chain is shortest.

For a buyer the question is what happens to that advantage. Vertical integration is a durable position, which makes Google the safe long-term bet, and it is also the position most likely to be priced accordingly once the rest of the field has consolidated.

Salesforce debuted directly into the Leaders quadrant, which almost never happens

A first appearance landing in the Leaders quadrant is rare in any Gartner market and worth pausing on. It usually means the vendor arrived with a product built somewhere else and a customer base already in place, which is exactly what happened here.

Salesforce brought an install base from CRM, a data layer from Data Cloud, and a product line it had been building for two years before submitting itself for evaluation. The template catalog and the pricing model were assembled for this report as much as for customers. That is not a criticism; it is how a large vendor converts an adjacent position into a category placement in a single cycle.

The strategic consequence for the rest of the field is that Salesforce did not have to win on conversational quality to reach the Leaders quadrant. It had to be credible on conversational quality and overwhelming everywhere else. A specialist competing on the quality of its dialogue engine is now competing against a vendor whose argument is that the dialogue engine should not be a separate purchase at all.

SoundHound AI moved up from Visionary while NiCE Cognigy moved down from Leader

Gartner called the SoundHound promotion the most notable positional change in the report, and it is the cleanest single-vendor story in the 2026 cycle. SoundHound moved from Visionary to Leader on the strength of its acquisitions of Amelia and Interactions, which it folded into Amelia 7, plus a proprietary speech recognition and synthesis stack it tunes itself and real-time voice-to-voice models.

The caution attached is the mirror of the achievement: integrating that much acquired technology carries execution risk, and a platform assembled through acquisition has to prove it behaves like one product. SoundHound is now on the rung precisely because it owns the layer most of the field rents.

Moving the other way, NiCE Cognigy went from Leader to Visionary. The vendor appears in this report as NiCE Cognigy and appeared in the previous one as Cognigy, following the acquisition that produced the compound name. Consolidation is the mechanism here rather than product regression, and it is the pattern Gartner's early-maturity framing predicts.

The wider movement is larger than either name. Boost.ai went from Leader to Challenger in twelve months. IBM moved from Challenger to Visionary. Druid AI and Yellow.ai both fell from Challenger to Niche Player. Netomi entered as a Challenger. LivePerson was removed from the report entirely. Nine of the fourteen vendors changed position or status in one cycle, in a market Gartner describes as maturing.

Forrester drew this market twice, splitting it by audience rather than by capability

Forrester's coverage of the same technology is not one report and not one boundary. Its Wave for conversational AI platforms for customer service, published 16 April 2026 and led by Max Ball, scored fourteen vendors drawn from a field of more than six hundred and fifty. Its Wave for conversational AI platforms for employee services, published 22 July 2026, scored thirteen providers, with the previous edition covering fifteen against twenty-eight criteria and a Q1 2026 Landscape mapping thirty vendors without scoring them.

Gartner drew one category around a capability. Forrester drew two around audiences, an employee-facing platform and a customer-facing one, and evaluated them separately even though the underlying technology is largely the same. Neither firm is wrong. They are answering different questions about where the budget sits.

The scoping difference has a practical effect. A buyer reading Gartner gets a single ranking of platforms that may be deployed for either purpose, and then has to work out which of the four Leaders actually fits their use case. A buyer reading Forrester gets a list already filtered by audience, at the cost of having to read two documents to see the whole field.

Kore.ai is the only vendor on both firms' Leader rungs

Set the two firms' 2026 results side by side and the overlap is one name. Kore.ai is a Leader in Gartner's quadrant and a Leader in Forrester's customer-service Wave, where its own account says it scored highest in the current offering category with top marks in guardrails and data security, agentic framework, resource orchestration, voice and telephony support, scalability and reliability, and development tooling.

The more interesting comparison is the vendor the two firms place differently. NiCE Cognigy's own account of the Forrester Wave says it recorded the highest combined score of any vendor evaluated and the top Strength of Strategy score, at 4.5 out of 5, with the only above-average customer feedback scores in the report. Three months later Gartner placed the same vendor as a Visionary. Both can be true. Forrester was scoring strategy and current offering against customer-service criteria with customer feedback weighted in; Gartner was scoring a broader category against execution and vision.

That gap is the most useful thing in either document for a buyer, because it identifies the vendor whose ranking depends most on which question is being asked. Kore.ai is the safest pick under both rubrics. NiCE Cognigy is the strongest pick under one and a question mark under the other.

There is a third reading worth holding alongside both. Forrester found that newer AI-native entrants fell below par on legacy system integration, agent escalation handling, reporting, governance and mature development tooling, and that the differentiators separating the leaders were enterprise readiness rather than sophistication of the models. The market named for artificial intelligence is being decided on the parts that are not artificially intelligent at all.

What to ask before you buy a conversational AI platform

Ask which use case is being scored. The Critical Capabilities companion report names four and no vendor led all of them, so a quadrant position is not a statement about the deployment you have in mind.

Ask who owns the speech layer. The leader with the strongest model position buys the least and the leader promoted this year owns the most, and that difference shows up in what each can promise about latency, accents and cost per minute.

Ask for the governance controls as artefacts rather than as features. The criteria moved toward operational stability and auditability this year, which means a vendor's position in last year's report is not evidence about this year's question.

Ask what the pricing does at volume. Transparent tiered pricing is now a scored advantage rather than a courtesy, and a platform whose meter is explainable is one whose bill can be forecast.

Ask how much of the platform arrived by acquisition. Two of this report's most significant movements in either direction were driven by purchases, and an assembled platform has to be assessed on whether it behaves like one product.

Gartner does not endorse any vendor, product or service depicted in its research publications, and does not advise technology users to select only those vendors with the highest ratings or other designation. Gartner research publications consist of the opinions of Gartner's research organization and should not be construed as statements of fact. Gartner disclaims all warranties, expressed or implied, with respect to this research, including any warranties of merchantability or fitness for a particular purpose.

The customer-service half of Forrester's split lives next door to another page on this site. Contact-Center-As-A-Service Platforms covers the platform the conversational layer is usually bolted onto, where the same analyst, Max Ball, has tracked a market whose vendors are all selling automation that reduces the billable seats they charge for. A conversational platform bought on its own has to justify itself against that contradiction, and one bought as part of a contact-centre suite inherits it.

The layer that decides how well any of this works is scored elsewhere. Generative AI Model Providers covers the firms supplying the models these platforms reason with, and the asymmetry in this quadrant comes straight out of that market. The vendor holding both axes owns its models and its silicon, while the caution attached to the strongest new entrant is that it rents the speech layer. A conversational platform is a product assembled from parts it mostly does not make.