If your company bought Salesforce Sales Cloud in 2023, look it up on G2 today. It is called Agentforce Sales now.
The AI SDRs category sitting next to it didn’t even exist when you signed the contract, and it already lists 400 products.
Sales and GTM is the function AI is reshaping fastest in 2026. Other categories being reshaped are software development, customer support, marketing, security, recruiting, and meeting tools. The biggest movers are the established vendors. They are re-rating themselves, with new names and new category tags.
In equity markets, a re-rating happens when investors decide a company deserves a different multiple. The business may not have changed that quarter. The story about it did.
Software categories work the same way. A category on G2 is a shared memory of what a type of software is supposed to do, who buys it, and which products lead in it. When that memory gets rewritten, every vendor inside it gets re-priced by buyers, whether the product has changed or not.
It shows up first in the taxonomy. G2 adds a category when enough distinct products and buyer demand make the old labels not enough, so the creation rate is a direct read on how fast AI is re-sorting the market.
The 2024 to 2025 window built the scaffolding: Agentic AI, AI Agents, AI Agent Builders, AI SDRs, AI Customer Support Agents, and AI IT Agents all arrived that year. The 2025 to 2026 window built an agent for every department and the plumbing under them: AI Legal Operating System, AI Agents for HR, AI Marketing Agents, Agentic GTM Platforms, AI Interview Agent, AI SRE Tools and Agentic Analytics, alongside MCP Server Infrastructure Platforms, AI Gateways, AI Agent Observability, and Small Language Models.
The fair objection is that a new label proves nothing about buyer behavior. So this piece checks whether buyers are writing new judgments inside these categories, whether vendors are moving their own products into them, and whether independent research explains the pace. A new category tells you a market might be forming. It does not tell you the market has arrived. That is the difference between a real market and marketing real estate.
Tim Sanders, Chief Innovation Officer at G2, described the agent wave in G2's October 2025 AI Agents report as "a shift from measuring individual productivity to predicting organizational velocity."
“Agents are the great leap in the application of AI technology that we've all been waiting for, signaling a shift from measuring individual productivity to predicting organizational velocity.”
Tim Sanders
Chief Innovation Officer at G2
Categories are where that shift becomes visible to a buyer, because the shortlist forms on the category page. Sanders also predicted that more than 35% of enterprise companies would set agent budgets of $5 million or more in 2026. Budgets that size do not flow into categories that did not exist 18 months ago unless buyers believe the map has changed.
Read The Answer Economy: How AI Search Is Rewiring B2B Software Buying to know where the buyer shortlist now forms, before they ever reach your site.
Three signals, all from G2's category and review data as of September 12, 2026, each checked against an independent source below.
Velocity is also the fix for the review-count problem. A category page can show 20,000 reviews, most of which were carried in by older products that were re-tagged into the category. The total tells you very little. The dates are important to confirm recency and frequency.
|
Rank |
Function |
New AI categories, Sep 2024 to Sep 2026 |
Flagship AI category: listed products and proof gap |
Review velocity: share of lifetime reviews written in the past 12 months |
|---|---|---|---|---|
|
1 |
Sales and GTM |
5 |
AI SDRs: 400 products, 74% proof gap |
Salesforce Agentforce 84% (1,051 of 1,253). Outreach 3.4% (123 of 3,590) |
|
2 |
Software development |
8 |
AI Coding Assistants: 369 products, 89% proof gap |
Cursor 94% (297 of 315). GitHub Copilot 60% (238 of 394) |
|
3 |
Customer support |
5 |
AI Customer Support Agents: 897 products, 76% proof gap |
Agentforce 84%. Fin, the renamed Intercom listing, 9% (346 of 3,911) |
|
4 |
Marketing and search |
6 |
Answer Engine Optimization: 623 products, 67% proof gap |
Semrush 32% (about 1,290 of 4,035) |
|
5 |
Security and IT operations |
7 |
AI IT Agents: 146 products, 75% proof gap. AI SOC Agents: 52 products, 52% proof gap |
Splunk Enterprise Security 2.4% (6 of 248) |
|
6 |
HR and recruiting |
3 |
AI Recruiting: 528 products, 79% proof gap |
Paradox 57% (51 of 90). Greenhouse 16% (652 of 4,002) |
Sales tops the index because all three signals fire at once. G2 added five AI categories to the sales stack in two years: AI SDRs, Agentic GTM Platforms, Revenue AI Platforms, AI Sales Roleplay Tools, and AI Proposal Generator Tools. The AI SDRs category alone lists 400 products.
Salesforce Agentforce has 1,253 reviews, and 1,051 of them were written in the past 12 months. That is 84% of its entire review base in just one year. Among the reviewers, 307 identify as administrators, 132 as consultants, and 58 as executive sponsors. And 719 of the 920 who disclosed company size work at mid-market or enterprise firms.
Outreach, the most-reviewed established product tagged as an AI SDR, has 3,590 reviews and added 123 in the same year. The market wrote eight times more fresh judgment about the agent than about the sales engagement leader. That gap captures the story of this category. Outreach's 3,590 reviews make it look like the most proven product on the AI SDRs page. Almost all of those reviews were written about a sales engagement tool before the category existed.
The clearest re-rating signal is a name. Salesforce's Sales Cloud now appears on G2 as Agentforce Sales, with Service Cloud and Marketing Cloud renamed the same way. Salesforce chose the boldest of the three moves a seller can make. It renamed its offering. Outreach chose the cheapest, a new tag on an old listing. The reviews show which one buyers rewarded.
Tommy Söderman, Co-founder of Lito Consulting and a Salesforce Marketing Cloud architect, wrote in September 2026 that "half of an Agentforce project is really a data project, and that should be said more clearly up front."
Kasi Cheaney at Acima, an enterprise financial services company, reported that new admins cut case age by 25 to 30% because they no longer wait on developers.
How reviews have changed
In 2023, sales reviewers judged tools on how well they organized a rep's own outreach: 230 of Outreach's 289 reviews that year were filed under Email Tracking, and only 2 carry the AI SDRs tag today. In 2026, reviewers judge the agent on the work it takes off people's plates, and their complaints moved from dialer glitches to data readiness.
The AI SDRs page is still a waiting room. 298 of the 400 products have no reviews. Twenty real ones from your segment would put you near the front of the line. Sellers must collect reviews about the AI itself, written by the admins and executives who run it every day. A new tag on an old listing will not do it. Outreach is tagged in and still added only 123 reviews last year. You also must be upfront about the setup.
Buyers must read this section in two layers.
First, which products have reviews from this year, and from companies of your size?
Second, which reviews describe the agent, and which came along with an older product?
That second layer is where most shortlists go wrong. Before you sign, you must get three answers from the vendor: Who owns the agent day to day? What data does it need before it works? How will you measure it after 90 days? Then you must budget for the data work.
The shortlist is getting decided even before your reps see the deal. Learn How AI Search Is Changing the Sales Pipeline to know what pipeline teams can still influence.
Software development produced the most new AI categories of any function, eight in two years: AI Coding Assistants, AI Software Testing Tools, AI App Builder, AI SRE Tools, AI Documentation Generators, AI AppSec Assistants, AI Engineering Accelerators, and AI SDK.
The flagship category tells the land rush story in one number: 369 products listed, 328 with no verified reviews yet. Ninety percent of the shelf is a claim. This is what marketing real estate looks like: a category page that is crowded and almost empty at the same time.
The demand signal is the strongest in our research. Cursor has 315 reviews, and 297 of them arrived in the past 12 months, a 94% velocity. GitHub Copilot sits at about 60%, with 238 of its 394 rated reviews written in the same window.
One caution: 301 of Copilot's lifetime reviews come from India, and 206 of Cursor's 288 recent reviewers who disclosed company size work at firms of 50 people or fewer. The judgment is fresh, and it skews small.
Hitarth Anand Rohra, a software engineer at Matrix One, wrote in July 2026 that Cursor "feels like an IDE first and an AI tool second," then shared that it sometimes makes broad changes when a small edit was wanted.
Muhammed Aamer, technical project manager at logistics firm CargoX, wrote that verification against real documentation "remains necessary rather than optional" because the assistant still invents library APIs.
Antonio Tirado Peña, director of IT and strategy at Instituto de Innovación Ciencia y Empresa, recovered the subscription cost in the first few days but flagged usage limits that only show up after the fact. Two years ago, buyers asked whether the tool could autocomplete. Now they ask how much work they can hand over, and what it costs when they push.
How reviews have changed
In 2023, developers reviewed an autocomplete: all 126 of Copilot's reviews that year were filed under AI Code Generation, and reviewers praised suggestions that arrived as they typed. In 2026, they review something closer to a coworker, and ask how much work they can hand over, how to check it, and what it costs when they push.
The bar here is low, and almost nobody has cleared it — 41 reviewed products out of 369. Sellers must get developers at companies with more than 50 people to review them. Do that, and you become one of the most proven tools in the category. You also start talking to a buyer that the current reviews mostly skip. Then you must fix the complaint that shows up in review after review: Tell people what the usage limits are before they hit them.
Developers in these reviews say the tool feels fast, then describe the hour spent checking its output. So you must run a pilot on your own code, with a clock. You must also filter the reviews to your company size before trusting a rating. Most of Cursor's recent reviewers work at very small firms. And you must ask for the cost meter up front.
AI Customer Support Agents is the biggest AI-native category we found, with 897 listed products. 678 of them have no verified reviews yet, a 76% proof gap against 34% in Help Desk, the category from which most of these vendors come. With Contact Center AI Observability, Conversational Interface Agents, AI Voice Assistants, and Enterprise AI Chatbots, the function has five new categories.
Support ranks third rather than first because the velocity picture is split. Agentforce, which is tagged into this category as well, carries the same 84% velocity, and 330 of its past-year reviews were filed under AI Customer Support Agents.
Fin, the AI agent that now fronts Intercom's G2 listing, has 3,911 reviews, but only 346 were written in the past 12 months, because the listing carries years of help desk and live chat history. The renamed product inherited the old category's memory: The vendor wants buyers to see an agent, and the review base still describes a help desk.
Agentforce Service (formerly Salesforce Service Cloud) and ServiceNow AI Agents sit in the same competitive sets. The help desk of 2024 is becoming a category of agents that answer, act, and escalate, and the incumbents are relabeling to stay on the page.
If you are a help desk that renamed or re-tagged, the move bought you a spot on the page and little else. Anyone who reads the dates can see your reviews describe the old product. Sellers must go get fresh reviews about the agent. The best place to find them is the teams onboarding new support staff. That is where the big gains show up.
Buyers must measure the tool on their newest agents. That is where it helps most, so that is where a fair test lives. Pick one queue and give the agent a clear escalation rule. Then track the case age for the new hires on that queue for a quarter. You must also ask the vendor the question reviewers keep raising: When the agent picks the wrong answer or ignores a rule, can you see why?
Marketing's re-rating runs through search. G2 created AI Search Visibility Optimization Tools this year and Answer Engine Optimization the year before, along with AI Marketing Agents, AI Avatar Generators, AI Storyboard Generators, and AI Presentation Tools. Answer Engine Optimization lists 623 products, 417 with no verified reviews yet.
The remarkable thing about the AI Search Visibility category is who is in it. Of the 34 products with a minimum of 40 reviews, at least 18 are established SEO and content tools: Semrush, Ahrefs, Moz Pro, BrightEdge, Conductor, seoClarity, SE Ranking, Similarweb, and Siteimprove among them. The AI-native entrants, tools like Brandlight AI, Scrunch AI, AthenaHQ, and Visby AI, mostly have 40 to 200 reviews each. The incumbents did not wait to be displaced. They re-tagged.
Semrush shows how fast the re-classification runs. It has 4,035 reviews, and about 1,290 were written in the past 12 months, a 32% velocity that is high for a 4,000-review product. Of those recent reviews, 373 were filed under Answer Engine Optimization and 354 under AI Search Visibility Optimization Tools. Semrush did not have to decide whether to re-tag. Its buyers decided for it, by filing their reviews under the AI labels.
Pedro Martinez, who runs SEO for an agency's clients, wrote in August 2026 that he now works with the product "primarily outside of its interface," via chat tools and scheduled workers who call the API.
Pasha Leone, a small-business marketer, uses it to research "keywords + prompts" and track answer-engine traffic.
Jodie Maxwell, senior marketing executive at Xeretec Group, uses it to measure whether SEO and AEO spend is improving visibility. The tool has not changed the category. The buyers have.
How reviews have changed
In 2023, Semrush reviewers talked about keywords, backlinks, and site audits, and 136 of that year's 212 reviews were filed under SEO Tools, with none under Answer Engine Optimization. In 2026, 373 of about 1,290 sit under AEO, and reviewers describe prompts, AI visibility, and running the tool through their own agents.
If you sell SEO or content tools, you are already in the AI search category. Your customers put you there when they filed their reviews. So sellers must price for AEO instead of treating it as a feature. The reviewers have spelled out what they want: prompt tracking in the standard plan. AI-native vendors must accept that the incumbents will always have more reviews. Your edge has to be proof in one segment, with numbers, from marketers doing the new job.
Buyers must change the questions they ask vendors. Rank and clicks measured the old search. Now you must ask how the tool shows whether your brand appears inside the answer. You must ask how it tracks both prompts and keywords. You must decide what visibility means for your team before the demo.
Earning a place inside the AI answer is among the top problems confronting CMOs. Learn How AI Has Redefined the Rules of Brand Discoverability, and what CMOs must do now.
Security and IT ops produced seven AI-native categories in two years: AI SOC Agents, AI IT Agents, AI Security Solutions, AI Security Posture Management, AI Governance Tools, Autonomous Endpoint Management, and Non-Human Identity Management. On the taxonomy signal alone, this function would rank second.
The other two signals pull it down. The categories are small: AI SOC Agents lists 52 products, and AI IT Agents lists 146. And the fresh judgment is thin. Splunk Enterprise Security is tagged into AI SOC Agents, but only 6 of its 248 reviews were written in the past 12 months. Security buyers are slow to publish opinions, and the agent categories here are still mostly claims.
These categories are so small that the first vendors to earn real analyst reviews will define them. Sellers must pitch as a complement to the SIEM and SOAR that the buyer already runs. You must also put your false positive rate in the pitch before anyone asks.
Buyers must treat every agent claim here as unproven until analysts on your own team have run it. You must start small, with alert enrichment and triage. There, a wrong call costs a wasted minute rather than a missed breach. You must measure false positives and time to verdict in your own environment before widening the scope.
AI Recruiting, AI Interview Agent, and AI Agents for HR all arrived in the past two years. AI Recruiting lists 528 products, 415 of which have no verified reviews yet.
Paradox, a conversational hiring assistant, has a 57% velocity, but on a small base of 90 reviews.
What holds recruiting at sixth is the incumbents' behavior. Greenhouse collected 652 reviews in the past 12 months, a healthy 16% of its 4,002 total, and not one of them is filed under an AI category. Its reviews still live in Applicant Tracking Systems, Recruitment Platforms, and Talent Intelligence.
Recruiting incumbents have mostly stayed put. Recruiting is being re-rated at the edges, with screening and interview agents, while the system of record keeps its label. Greenhouse is a case of a seller choosing to hold. It kept its labels, kept collecting reviews, and lost nothing, because recruiting buyers have not moved their search to the AI pages yet. Holding is a real option, but itt only works while the buyers stay put, too.
The AI Recruiting page is wide open because the big systems of record have not shown up. Hiring is the one function where a bad automated decision becomes a legal problem. So buyers move one use case at a time, and the large employers move first. Sellers must start with enterprise talent teams. You must lead with how you document every decision the agent makes. You must collect reviews about the AI feature itself. And if you run an ATS, you can hold your position for now. But you must watch where your own customers file their reviews.
Buyers must keep their system of record and let AI work at the edges: screening, scheduling, and first interviews. That is how the market is moving, and it is the lower-risk path. You must ask how the tool documents each automated decision. You will need to answer that the first time a candidate asks. You must run a bias check on your own applicant pool before you go live, and again every quarter. And you must pilot on high-volume roles first, where the time savings are large, and a human still makes the final call.
The expected story is startups eating incumbents. The data says incumbents are doing much of the reshaping, with names and tags.
Start with the names.
Beyond the three Agentforce renames, the G2 catalog has picked up monday AI Work Platform, GTM Workspace powered by ZoomInfo, UiPath Agentic Automation, and Gemini Enterprise Agent Platform in the past year, and Intercom's listing now leads with Fin. Each rename moves thousands of legacy reviews under an agent banner.
Then the tags. Of the 16 products with 150 or more reviews in AI SDRs, at least 7 predate the category: monday, Outreach, Close, Qualified, Reply, LivePerson, and Conversica. Splunk sits in AI SOC Agents, Evernote in AI Note-Taking, Calendly in AI Meeting Assistants, and Paylocity in AI Recruiting.
Three in four products in G2's AI-native categories have no verified reviews yet. In mature categories, it is one in three to three in five.
The stats above change how you read a category page. A category with thousands of reviews can be borrowed memory, carried in by products that earned those reviews doing something else. The proof gap is the honest floor of the market. The review totals stacked on top are the ceiling companies built by moving in. The truth about a category in 2026 lives in the review dates, and G2 is the one place you can see it, because every listing sits next to a review count, and every review carries a date.
If you sell software and are trying to decide what to do, here is what you must consider:
1. Rename when the new product really is the product now, the way Salesforce did, and accept that your review base will age in public while the new one builds.
2. Re-tag when your buyers are already filing reviews of you under the AI label, the way Semrush's are, because the tag just catches up with them.
3. Hold when your buyers have not moved their search to the AI pages yet, the way Greenhouse has.
If you buy software, treat category placement as a claim and the review dates as the evidence.
Before you add a product to a shortlist in any of these categories, ask three questions.
Then budget for the data project the reviewers keep describing. The tools in the top three functions deliver measured time savings, and the same reviewers say the setup was bigger than the pitch.
If you sell software, the arithmetic is simple. Three in four products in these categories have no verified reviews yet, so 20 real ones from your segment put you ahead of most of the shelf, and category pages are where buyers and answer engines form the shortlist.
Rename with care: A new name can move your review base and strand it under labels buyers no longer search for. Run the three signals on your own category every quarter. A category being re-rated is the one place a challenger can pass a leader in a single year, and the year is not over.
Sales and GTM first, then software development, customer support, marketing and search, security and IT operations, and HR and recruiting. The ranking uses three signals from G2 data: how many AI-native categories G2 has created for each function since September 2024, the share of listed products in the flagship AI category that have no verified reviews yet, and the share of each leading product's reviews written in the past 12 months.
Look at the proof gap first. In mature categories on G2, between a third and three-fifths of listed products have no reviews. In the AI-native categories, it runs from 52% to 90%. The closer a category sits to the mature range, the more buyers have actually voted. Then check review dates on the leading products. A category where most of the fresh reviews are about AI-native products is a market forming. A category where the totals come from re-tagged older products is still mostly a label.
It depends on what your buyers are already doing. Rename when the AI product really is the product now, as Salesforce did with Agentforce, and expect your old review base to age in public. Re-tag when customers are already filing reviews of you under the AI label, as Semrush's are. Hold when your buyers have not moved their search to the AI pages, as in recruiting. In every case, only fresh reviews about the AI capability move you on the page. The tag by itself does not.