Four separate research shops landed on the same finding this week, and none of them were talking to each other. IDC says eight in ten B2B technology buyers already use AI agents somewhere in the purchase. Forrester says nine in ten B2B buyers have adopted generative AI across vendor research, RFP drafting, and comparison. TrustRadius says 83% of buyers shortlist three or fewer products, with an average shortlist of 2.7, and 67% buy their first choice. G2 asked 1,076 software buyers what the chatbot actually changed, and 69% said it led them to a different vendor than the one they had planned on.
Read those four numbers in order and you get a sequence, not four separate trends. A buyer delegates the first pass to a model. The model returns two or three names. The buyer picks one of them most of the time. And in seven cases out of ten, that list is not the list the buyer walked in with. By the time your form fill happens, the comparison is over.
The response taking shape is not another tool. Gartner's 2026 digital marketing Hype Cycle frames the CMO problem as a trilemma of flat budgets, growth targets, and answer engines, and it puts the answer in cost governance and brand trust controls rather than capability. Gong is routing Revenue AI through Microsoft Marketplace so it lands inside identity, logging, and conditional access. IDC found 45% of enterprise AI projects failing to deliver, and 16.7% of enterprise AI budgets now going to security. The word that keeps showing up is governance, and it is doing real work: it means deciding what your systems are allowed to say about you, and which of them are allowed to act.
Every item below carries a source link. Where a figure comes from a vendor describing its own product or customers, we label it a claim and attribute it.
The big picture
Three shifts to take into your next pipeline call
- Consistency across your public surfaces is now a revenue control, not a brand chore. If your pricing page, your partner catalog, your review profiles, and your PDFs disagree, a model will resolve the conflict for you and you will not see the resolution. The GSPANN analysis of Forrester's GTM singularity framing calls this a data governance problem, and it is right. Start with the three claims your reps repeat most and check whether every public surface says the same thing.
- Write access is the line that matters, not AI adoption. Reading and summarizing is a light governance conversation. Updating a CRM record, sending an email, enrolling a sequence, or negotiating a quote is not. Forrester expects 20% of B2B sellers to face agent-led quote negotiations before the end of 2026. Ask your team for a system-by-system map of what your AI can change without a human, and watch how long it takes to produce it.
- Nobody has solved AI-visibility attribution, and pretending otherwise is the fastest way to lose a budget line. Demandbase reports ChatGPT-referred visits to B2B brands rose from roughly 645,000 in June 2025 to 2.6 million in June 2026 (their figure, their panel). The traffic is real. The line from that traffic to a closed deal is not drawable with current tooling, and teams that promise one will be asked to show it in Q4.
AI-native GTM
The buyer's first pass now happens somewhere you cannot see
Research worth reading twice
Forrester's "GTM singularity" turns your public claims into a RevOps deliverable
The argument is that every marketing and sales stack built since the early 2000s assumed a human reads the marketing, fills a form, and enters a workflow. When a model sits between buyer and seller, that assumption stops holding. The numbers behind it are the useful part. Forrester's 2026 Buyer Insights research puts generative AI adoption among B2B buyers at nine in ten, across vendor research, RFP drafting, comparison, and validation. A G2 survey of 1,076 B2B software buyers found 69% said an AI chatbot led them to a different vendor than planned, and about one in three ended up buying from a vendor they had not previously heard of. TrustRadius data cited in the same analysis puts 83% of buyers at shortlists of three or fewer, averaging 2.7, with 67% buying their first choice.
The operational read is that pricing hygiene, product information management, and third-party profile maintenance move from marketing housekeeping into the RevOps backlog. The piece also flags Forrester's expectation that 20% of B2B sellers will face agent-led quote negotiations before the end of 2026, which puts deal desk policy and approval matrices on this year's calendar rather than next year's.
Buyer behavior
IDC puts AI agent use among B2B tech buyers at 80%, and enterprise AI project failure at 45%
Two IDC findings from August sit next to each other and complicate each other. Eight in ten B2B technology buyers are already using AI agents as part of the purchasing process, which is a current-state measurement rather than a forecast. A separate IDC report from early August found 45% of enterprise AI projects failing to deliver results, with CIOs asking for clearer return, better security, and defined governance for agentic systems. IDC also reports that 16.7% of enterprise AI budgets are now directed toward security, and projects 1.2 billion agents in operation by 2029.
From the watchlist
Kyle Poyar published the 2026 GTM field guide, and the vocabulary section is the point
Poyar organizes the current playbook into four levers: motions and channels, AI-first execution, signals and data, and pricing. Several numbers in it are worth lifting out. Answer engine optimization is now the single channel where B2B marketers are increasing investment the most, ahead of intent-based outbound and LinkedIn. AI-native products see roughly 75% to 90% of signups arrive on personal email addresses, which is why de-anonymization has become a real product category. There are now more than 400 GTM engineers at US digital-native companies by his hiring-report count. On the plumbing side, he cites a ScaleKit study finding MCP runs 10 to 32 times more expensive than CLI, which is why mature teams are moving back toward API integrations.
The most quotable data point is his own. After adding a "how did you hear about us" field in January, he found 14.3% of new Growth Unhinged subscribers first heard about it through ChatGPT or another model, roughly ten times what his click-based analytics showed.
RevOps
Governance stopped being the boring part
Vendor move
Gong routes Revenue AI through Microsoft Marketplace, which changes who is in the room
On paper this is a distribution update. In practice, transacting through Microsoft Marketplace pulls the purchase into Azure-standard identity, conditional access, data residency, and logging, which means IT and procurement join a conversation that used to be a sales leader signing a seat license. Gong is pairing that with agentic execution language, and separately promotes 400 or more integrations in its Gong Collective ecosystem (Gong's figure). Gong has also claimed, in a November 2024 release on its State of Revenue Growth report, that revenue organizations using AI in 2024 reported 29% higher revenue growth and 11% better go-to-market efficiency than peers. Vendor claim Survey-based, self-selected, and reported by the vendor, so treat the direction as more useful than the number.
Standards
Google's A2A protocol moves to the Agentic AI Foundation, next to MCP
Agent2Agent, Google's protocol for letting independent agents talk to each other, is moving under the Agentic AI Foundation, a vendor-neutral Linux Foundation body and the same home as Anthropic's Model Context Protocol. A2A handles agent-to-agent communication, MCP handles connecting applications to tools and data. The foundation has grown from fewer than 40 members at its December 2025 launch to more than 250.
Adoption data
Salesforce says the average organization now runs 13 agents, up from 5
Salesforce's Agentic Enterprise Index, drawn from production activity at 400 businesses plus a survey of nearly 5,000 people, reports the average number of deployed agents per organization rose from five in early 2025 to 13 by April 2026. It also reports agent creation time down 53%, employee sessions tripled, and seven in ten customer service sessions handled autonomously among organizations in its dataset, with escalation rates holding steady. Vendor claim This is Salesforce measuring activity on Salesforce, so it describes what its own customers are doing rather than the market.
Measurement, AEO, and AI search
The visibility number nobody can connect to revenue
The honest problem
CMOs are buying AI visibility tools and cannot show what they returned
Tracking tools for brand presence in ChatGPT, Google AI Overviews, and other assistants are now standard purchases. The revenue line is still missing. Demandbase data cited in the reporting puts ChatGPT-referred visits to B2B brands at roughly 645,000 in June 2025 rising to 2.6 million in June 2026, a 303% increase. Vendor data But conversions frequently land later through other channels and stay untraceable. John Barham of the performance agency Roast describes every CMO he knows being under pressure from boards and investors to crack this, and says the tooling to draw a clean line from model visibility to sales does not exist. Teams are falling back on statistical approaches, including Google's Causal Impact model, and treating AI search influence as probabilistic rather than attributable.
Volatility
Reddit's citation share in ChatGPT reportedly fell from 4.5% to about 0.5% after a retrieval change
PromptWatch data indicates Reddit appeared in as many as 4.5% of ChatGPT outputs before August 8 and fell to roughly 0.5% afterward. The decline lines up with OpenAI apparently shifting toward more targeted site-scoped searches instead of starting with broad open-web retrieval, though OpenAI has not confirmed the change. Reddit says it remains highly cited in other analyses and that it depends mainly on direct visits and traditional search.
Marketing and paid media
Flat budgets, priced autonomy, and a new ad surface
Analyst framing
Gartner's 2026 digital marketing Hype Cycle puts the hard part in cost control and brand trust
Gartner describes CMOs facing a trilemma of flat budgets, aggressive growth targets, and disruption from what it calls answer engines, and positions autonomous marketing as the operating model forming in response. The framing that matters for buyers is that the evaluation criteria expand from features to controls: metering models for AI features, spend caps when usage spikes, where approvals and audit logs live for AI-generated content, and what happens contractually when a vendor swaps underlying models. Gartner's separate 2026 CMO Spend Survey found marketing budgets effectively flat at 7.8% of company revenue, 15.3% of marketing budgets allocated to AI, and only 30% of organizations with mature AI readiness against 70% of CMOs aiming to lead on AI.
New channel
ChatGPT ads go live in 31 European markets today
OpenAI begins serving ChatGPT ads across 31 European markets on August 24, six months after the US launch. European advertisers buy through major agency groups first, with self-serve access planned later. OpenAI says it has adapted privacy policies for GDPR and that user conversations are not shared with advertisers. The company has shifted mostly to cost-per-click buying and says roughly 20% of ChatGPT queries carry direct commercial intent. Vendor claim Separately, OpenAI's CFO told investors the advertising business is approaching a $1 billion annualized run rate and that enterprise revenue has passed consumer revenue.
Disclosure and spend discipline
IAB updates its AI disclosure framework as labeling laws multiply
The IAB has published a second version of its AI Transparency and Disclosure Framework as labeling requirements take effect across Europe, Asia, New York, and California. It recommends disclosure for consumer-facing synthetic images and video, digital twins, certain synthetic voices, and conversational agents in advertising, while generally exempting routine editing and clearly fantastical imagery. It also warns that over-labeling trains people to ignore labels. The IAB reports 83% of ad executives now use AI in the creative process and 72% favor an industry disclosure standard.
On the spend side, Ramp data reported by PYMNTS shows the top 1% of US businesses spent a median $7,400 per employee on AI in July against $11.95 at the median company, and that even heavy spenders are price sensitive between models.
Put it to work
What to do with this, this week
- Run a three-claim consistency check. Take the three claims your reps make most often, then check the website, the pricing page, the two largest review profiles, and any partner catalog. Every disagreement you find is something a model will resolve on its own. This takes an afternoon and it shows the GTM singularity problem better than any deck.
- Add a "how did you hear about us" field and leave it alone for a quarter. Poyar found 14.3% of subscribers first heard through a model, ten times what clicks showed. Nobody can argue with your own field, and it costs nothing.
- Ask for the write-access map. Which connected systems can be updated without a human, who approved that, and where is the log. If it takes more than a day to answer, that is the finding.
- Build your AEO plan around measurement design. Start from the fact that nobody can draw a clean line from model visibility to revenue, then use the honest stack: self-reported attribution, branded search lift, share of voice, and modeled impact. Setting that expectation is what protects the budget when someone asks for a number.
- Put four governance questions into the next martech renewal. How is AI usage metered and capped, where do approvals and audit logs live, which decisions stay human-gated, and what is the notification process when the vendor swaps models. Gartner's framing gives you the cover to ask.
- Get a per-employee AI spend number. The gap between the top 1% at $7,400 and the median at $11.95 means most companies have no idea where they sit. Knowing the number is the start of treating AI as a budget that competes rather than a line that gets added.
- Do not concentrate an AEO plan on one platform. Reddit went from 4.5% to roughly 0.5% citation share in ChatGPT in about two weeks on an unannounced change. Spread across formats and third-party surfaces, and report citation share as a trend, not a KPI.
Every claim above carries a source link. Figures attributed to vendors, to agencies reporting their own results, or to companies surveying their own customers are their claims, not independently verified facts, and are labeled as such. Several items in this issue (the Digiday AI-visibility reporting, the Reddit citation data, the Salesforce Agentic Enterprise Index, the IAB framework update, and the Ramp spend data) were first published between August 8 and August 16 and are included here because they surfaced in this coverage window through the August 21 MarketingProfs AI Update. Gartner's Hype Cycle for Digital Marketing was published July 10 and Gong's Microsoft Marketplace announcement July 1; both are included because the operator analysis landed inside this window. Coverage window: August 17 to August 24, 2026. Compiled August 24, 2026.