Founder Intel Tuesday Intelligence Brief
115 articles across 50 sources scanned this week covering AI model releases, buyer evaluation behavior, B2B productivity benchmarks, infrastructure economics, and cybersecurity. One development rose above the rest. Edelman’s new B2B research found that 84% of decision-makers used AI to summarize thought leadership during vendor evaluation, while 59% of final decision-makers said poor or absent thought leadership contributed to removing a vendor from consideration. AI is not replacing the buying process, but it is becoming another way buyers interpret what vendors know before making a decision. Here is what the market is telling you.
84 Percent of B2B Buyers Are Using AI to Summarize Vendor Expertise. What Will It Say About Yours?
Tuesday, November 10, 2026
01 Lead signal — This week's market signal
Market Signal
Edelman’s 2026 B2B Thought Leadership Impact Report, released November 9 and based on more than 1,100 executives across seven countries, found that 84% of B2B decision-makers used AI to summarize thought leadership during vendor evaluation. The same research found that 59% of final decision-makers said poor or absent thought leadership contributed to removing a vendor from consideration, while 52% said thought leadership helped their buying group decide which vendor to select. That does not mean AI is independently choosing vendors or replacing the buying committee. But it does mean a large share of buyers can now compress a firm’s published expertise into a summary before deciding what deserves deeper attention. For firms that sell expertise, that changes the question. It is no longer enough to have strong thinking somewhere inside the business. The thinking has to exist in a form that a buyer, and increasingly the AI helping that buyer, can actually find, understand, and summarize.
 
Thesis
You have built your business on expertise and relationships. You win when the right person makes an introduction, when your track record speaks for itself, and when your reputation gives a buyer enough confidence to take the meeting. None of that stopped mattering. What changed is that buyers now have another way to evaluate what you know. They can take your articles, case studies, interviews, research, website copy, and public commentary and ask an AI to summarize your position on the problem they are trying to solve. That creates an uncomfortable distinction for expertise-led firms: having expertise and having evidence of expertise are not the same thing. If your strongest thinking lives mostly in meetings, proposals, private conversations, or the heads of your senior people, an AI cannot summarize it very well. Neither can a buyer who has not met you yet. The goal is not to publish content for robots. It is to make the useful parts of your expertise visible enough that both humans and machines can understand what you believe, how you approach the problem, and what evidence supports it. AI did not create that requirement. It is making the absence of it easier for buyers to notice.
Do This Today
Open ChatGPT, Perplexity, and one other AI tool you already use. Give each one a problem your ideal client would realistically research, then ask: “What does [your company] believe about this problem, and what evidence supports its approach?” Do not grade the answer on whether your company appears first. Grade it on whether the system can explain what you believe about the problem, how your approach is different, what evidence or examples support it, and where that information came from. If the answer is vague, generic, or mostly inferred from your homepage, that is useful information. You have identified an expertise gap that exists publicly even if the expertise itself exists inside your firm.
Do This Week
Pick one buyer problem that matters to your firm and run the same evaluation across ChatGPT, Perplexity, and Google AI Mode. Write down what each system says about your firm, which competitors appear, what claims keep resurfacing, and which sources the systems rely on. Then choose one missing piece of evidence from your own business: a methodology you have never documented, a client result you have only talked about privately, a strong point of view buried inside a proposal, or a recurring question your senior team answers in every sales conversation. Publish that evidence in plain language and tie it directly to the buyer problem. Then run the same test again later. The goal is not to “game AI.” It is to make sure the expertise buyers are trying to evaluate actually exists somewhere they can inspect.
02 Secondary patterns — Three other themes that moved this week
Pattern 01 · AI MODEL COSTS ARE MOVING FAST
One production AI system cut costs 72% after changing its model mix.
A practitioner case published this week reported that SaaStr Connect reduced its AI costs by 72% after moving to Haiku 5.5 and a cheaper mix of models, while also reporting faster output and improved quality. Separate research on multi-step agent workflows found a similar result: AgentRouter reduced inference costs by 72% relative to using frontier models for every step while retaining 97.3% of the baseline quality. The takeaway is not that every AI workflow can cut costs by 72%. It is that model choice and routing can materially change the economics of an AI-enabled product or service, and those economics are moving quickly enough that a model decision made at launch may deserve another look a few months later.
Watch for firms to start re-benchmarking production workflows across model tiers instead of automatically routing every task to the most capable model. The more repeatable the workflow, the easier it becomes to measure where cheaper models are good enough and where frontier capability still earns its cost.
Pattern 02 · LARGE FIRMS ARE BUYING SCALE AND AI CAPABILITY TOGETHER
Grant Thornton’s $5B CBIZ acquisition adds scale while accelerating a $1B AI investment.
Grant Thornton completed its $5 billion acquisition of CBIZ this month, creating the fifth-largest U.S. provider of professional, tax, and advisory services. The firm said the combination expands its industry, technology, and multinational capabilities while providing greater scale for a previously announced $1 billion investment in AI and advanced technologies. That does not mean AI caused the acquisition. But it does show how larger professional-services firms are combining traditional expertise, broader delivery capacity, technology, and AI investment inside the same platform rather than treating those capabilities as separate strategies.
Watch for more professional-services acquisitions where the rationale explicitly combines domain expertise with software, data, or AI-enabled delivery. The stronger signal will be whether those capabilities become integrated into the client offering rather than remaining separate technology investments.
Pattern 03 · AI-FORWARD COMPANIES ARE SETTING NEW PRODUCTIVITY BENCHMARKS
ICONIQ’s $100M+ Pacesetters are generating a median $655K per employee while growing 115%.
ICONIQ’s 2026 Pacesetter Index tracks AI-forward companies with revenue growth in at least the top quartile for their size. Among Pacesetters above $100 million in annual recurring revenue, median growth is 115% and median revenue per employee is approximately $655,000. Those numbers should not be treated as a normal target for every B2B company. The cohort is deliberately made up of unusually fast-growing businesses, and many are AI-native or AI-driven. But it provides a useful picture of what operating leverage can look like at the leading edge as AI becomes embedded across products and internal work.
Watch whether revenue per employee continues rising across the broader market rather than only among Pacesetters. That would be stronger evidence that AI-enabled productivity is changing ordinary staffing economics rather than simply describing a small group of exceptional companies.
03 Tools — Worth knowing this week
Writesonic AEO Suite
Tracks how often your firm appears in AI-generated answers across major platforms, which pages are being cited, how your visibility compares with competitors, and whether citations are gaining or losing momentum over time. It is useful for founders and principals who want a standing view of how their firm is represented across AI search rather than manually repeating the same prompts every week.
Founders and principals at professional services or B2B firms who want to know whether their expertise is surfacing when buyers use AI to evaluate vendors before making contact.
This week’s Edelman data makes that monitoring more relevant: if buyers are using AI to summarize vendor thought leadership, it is useful to know what those systems are actually finding and which sources they are relying on. Writesonic does not tell you why a buyer chose one firm over another, but it can show where your brand is appearing, where competitors are appearing, and which content is earning citations.
Perplexity
An AI-powered research tool that answers questions using cited web sources, making it useful for seeing which companies, articles, and evidence surface around the problems your buyers care about. Founders can use it alongside ChatGPT and Google AI Mode to test how their firm and competitors are represented when a buyer researches a category or problem.
Any founder who wants to see their competitive landscape from the buyer's perspective and identify which firms are appearing in AI-generated results for the problems they solve.
With more B2B decision-makers using AI to summarize vendor expertise, Perplexity is particularly useful because its citations make the evidence trail visible. Run the same buyer question across several AI systems, then compare which firms appear, what each system says about them, and which sources support those answers.
04 Analysis — The strategic read

The important shift this week is not that AI has replaced Google, referrals, or relationships in B2B buying. It has not. The shift is that buyers now have another way to interrogate what a vendor knows before making a decision.

A buyer can take a category, a problem, or a set of firms and ask an AI to compress the available evidence: what each company believes, what it has published, what examples support its position, and how those approaches differ.

That raises the value of something many expertise-led firms have historically been able to leave implicit.

Evidence of expertise.

Your senior people may understand the client’s problem better than anyone in the market. But if that thinking lives mainly in meetings, proposals, private conversations, and experience that has never been documented, there is very little for a buyer or an AI system to inspect before the relationship begins.

This does not mean producing more generic content. In fact, AI makes generic information easier to reproduce. The more useful response is to make the parts of your expertise that are difficult to commoditize more visible: original data, specific methodologies, real outcomes, nuanced positions, and explanations of why you approach a problem differently.

The strategic question is becoming less about whether your firm “does content” and more about whether the market can see enough evidence to understand what you actually know.

05 Forward look — On our radar next week
Claude Fable 5.2 can work for hours and, in some cases, days on technical tasks without continuous supervision. Watch for buyers to begin testing whether longer-running agent workflows can absorb portions of software, research, analysis, and other project work that previously required continuous human execution. The meaningful evidence will be successful production use on real client work, not benchmark demonstrations alone.
Researchers documented hundreds of publicly accessible deployments of ARTEX, an autonomous penetration-testing framework that has also been linked to real-world attacks. Check Point separately observed more than 1,600 live publicly accessible instances and found examples where agents operated with little human interaction after receiving an initial objective. Watch for autonomous offensive tools to push agent security from a model-safety question into an ordinary cybersecurity concern involving credentials, exposed systems, monitoring, and response speed.
The falling cost of AI inference is not guaranteed to move in a straight line. U.S. AI infrastructure remains heavily dependent on imported semiconductors and other high-tech equipment, while data-center power and grid access remain constrained in several markets. Watch whether model-efficiency gains continue outrunning increases in compute, power, and infrastructure costs. If that relationship changes, some of the delivery economics currently making AI cheaper to deploy could change with it.
Founder Intel is funded and managed by GS Market Media.
founderintel.io  ·  Unsubscribe  ·  View in browser