Every established software category was once a phrase nobody agreed on. "Customer relationship management" was contested ground before it was Salesforce. "Expense management" was a feature of accounting software before it was Concur. The moment a category gets a name, a race begins to define what the name means, and in 2026 that race is increasingly refereed by AI.
QuadrantX added more than fifty categories this year, many of them B2B software segments that did not have a settled definition twelve months ago. Because each category is scored independently by five AI models and regenerated on a rolling schedule, we can watch a category's narrative form in something close to real time. This article looks at what the newest B2B software categories reveal about how AI models handle markets that are still being built.
Three Kinds of Emerging Category
Not every new category behaves the same way. The categories we added this year sort into three patterns, and each has different implications for the vendors in it.
1. Instant consensus: the category was new to us, not to AI
Some categories arrived with their narrative already settled. In financial planning and analysis software, all five models named Anaplan first from the outset, with Workday Adaptive Planning, SAP Analytics Cloud, and Oracle EPM behind. In unified endpoint management, Microsoft Intune was unanimous. In voice of the customer platforms, Qualtrics was unanimous and its acquired brands, Clarabridge and Delighted, filled the leaders quadrant beside it. In sports data providers, Sportradar took the top spot for every model with Genius Sports and Stats Perform close behind.
These are mature markets that analysts and buyers have described consistently for years. AI models had already absorbed the story. For a vendor in one of these categories, the emerging status is an illusion: the shortlist is as sticky as it is in CRM.
2. Contested definition: the models disagree about what the category is
The most revealing categories are the ones where models cannot agree on the question, never mind the answer.
In AI visibility and observability, the five models named five different leaders. DeepSeek and Gemini named Arize AI, an ML observability platform. GPT-5 named Datadog. Sonar Pro named Braintrust, an evaluation tool. Claude named Profound, a generative-engine visibility platform that has nothing to do with model monitoring. Two entirely different industries, model observability and AI search visibility, share a phrase, and each model chose a side. Across more than a hundred editions this year, ten different vendors have held the top spot, and 65% of vendors in the category were mentioned by only one model.
Crisis communications solutions split the same way, between mass-notification software (Everbridge, OnSolve) and public relations firms (Edelman). Secure communications split between network vendors (Cisco), collaboration suites (Microsoft), and defence-grade encryption (Thales). Digital advertising software produced five different leaders, four of which were different descriptions of Google.
Category Ambiguity — A state in which AI models interpret the same category label as different markets, and therefore recommend vendors from different competitive sets. In an ambiguous category, a vendor's AI visibility depends as much on which interpretation each model adopts as on the vendor's own strength.
3. Contested leadership: the models agree on the category but not the winner
The third pattern is the classic emerging market: everyone knows what the category is, and nobody knows who wins it.
Generative AI for legal professionals, added in late May, is the textbook case. Two models named Harvey AI, the venture-backed newcomer. Three named Thomson Reuters and its CoCounsel product, the incumbent. Lexis+ AI held the highest overall Narrative Dominance. The lead has changed hands in well over half of its editions, and five different vendors have held it. This is a narrative being written in public, edition by edition.
Real-time revenue execution platforms shows a similar fight: Salesforce Revenue Cloud for two models, Clari for Gemini, Cresta for Sonar Pro, and a bare "Salesforce" for DeepSeek. Three-quarters of the vendors in the category were named by a single model, the highest fragmentation of any category we track. Enterprise-ready SaaS building blocks, a developer-infrastructure category, split between WorkOS, Okta, and Auth0, with Stripe unanimously in the leaders quadrant beside them. Customer identity and access management had Okta on top for four models and Microsoft Entra for Sonar Pro, but with fifteen distinct vendors across the five top-five lists.
In an emerging category, the first vendor to be described the same way by everyone becomes the vendor AI describes to everyone.
What the Data Says About How Categories Settle
Because we regenerate every category repeatedly, we can compare how new categories behave against how mature ones behave. The contrast is sharp.
- Across the categories added since February, all five models agreed on the leader in 38% of cases, the same rate as the portfolio as a whole. But nearly a quarter had no majority leader at all, against 15% overall.
- In the new categories, one vendor in five appeared in every model's top five. The rest of the shortlist was contested.
- Leadership stabilises quickly once it forms. In the categories where models agreed from the first edition, the leader has held its position in more than 90% of subsequent editions. In the contested categories, it has changed hands in as many as half of them.
The practical implication is that the window in which an emerging category can be shaped is short. A category that reaches consensus stays there; the incumbents in established quadrants did not get there by accident, and they are not easily moved.
Who Is Winning the New Categories
Three kinds of vendor are winning the emerging categories in our data.
The category-definers. Affinity leads relationship intelligence platforms for four of five models, with a Narrative Dominance of 100 and a Sentiment of 95, because it has spent years explaining what relationship intelligence is. Airtable leads small business custom database solutions for three models for the same reason. These vendors did not win the category; they wrote it.
The incumbents with a new name. Thomson Reuters in legal AI, Salesforce in revenue execution, Qualtrics in voice of the customer, Microsoft in endpoint management. When a mature vendor extends into a new category, AI models tend to follow, particularly the training-data models whose picture of the market lags the present by a year or more.
The retrieval favourites. Cresta in revenue execution, Braintrust in AI observability, Microsoft Entra in customer identity: the picks that only Perplexity's web-retrieval model made. These vendors have recent, specific content that the live web rewards immediately. Whether the training-data models follow is the question that will decide the category, and in our data the answer takes months to arrive.
If you are a vendor in an emerging category, find out which of the three patterns you are in. If the category is already settled, compete for shortlist position, not the lead. If the definition is contested, publish content that anchors the category to the reading where you win, and say so consistently across your site, directories, and press. If leadership is contested, move fast: consensus forms within months and, once formed, does not move.
The Categories to Watch
Among the categories we added in 2026, three are worth watching closely over the coming quarters. Generative AI for legal professionals will resolve the Harvey-versus-incumbent question. AI visibility will either split into two categories in the models' understanding or be captured by one interpretation. Real-time revenue execution will show whether a platform vendor can own a category that specialists defined. In each case, the AI models' answer will arrive before the analyst firms' does, and the vendors that are watching will know first.
You can explore every category we track, including these, on the QuadrantX Explore page. For how models diverge on newer entrants generally, see our analysis of the expanding model ecosystem.