Agents Are Your New Customer

TL;DR

Agents are your new customer. B2B buyers are routing research, shortlisting, and parts of the purchase decision through ChatGPT, Claude, Perplexity, and Gemini before a human ever lands on your site. If your product, pricing, and content are not legible to those agents, you are losing pipeline to competitors who are. Here is what changes for B2B GTM, and a 90-day plan to adapt.

For twenty years, B2B marketing optimised for one buyer: the human, sitting at a desk, typing a query into Google. We measured rankings, click-through rates, dwell time, and form fills. We assumed the buyer would visit our site, read our pages, compare our pricing, and eventually book a demo.

That funnel is breaking. The buyer is still human. The research is not.

What Has Actually Changed

In the last twelve months, three things have shifted at once.

Research moved into chat. A buyer who used to type ‘best customer data platform for Series B SaaS’ into Google now types it into Claude or ChatGPT. They get a synthesised answer, not a list of ten blue links. If your product is not in that synthesised answer, you do not exist in the consideration set.

Comparison moved into agents. Buyers ask AI to do the legwork: ‘compare Segment, mParticle, and RudderStack for a 50-engineer team that needs HIPAA compliance, give me pricing, integrations, and a recommendation.’ The agent reads your site, reads your competitors’ sites, reads G2 reviews, reads recent threads. Whatever the agent surfaces becomes the buyer’s first impression.

Procurement moved into protocols. Agents can now call APIs, fill forms, sign up for free tiers, and even complete transactions. Elena Verna puts it plainly: your product has a new user, and it is not human. The agent reading your docs, evaluating the API, and recommending the purchase is doing the job a junior buyer used to do.

The buyer is still the human who signs the cheque. However, the agent is the new gatekeeper. It is fast, untiring, and immune to brand-name nostalgia.

How Agent Buyers Behave Differently

They read everything. A human spends 30 seconds on your homepage. An agent reads every page, every doc, every changelog, every G2 review, every Reddit thread. As a result, if your messaging is inconsistent across surfaces, the agent notices and the buyer hears about it.

They are extractive. Agents pull specific facts: pricing, integrations supported, SLA, data residency, supported regions, API rate limits. If those facts are buried in PDFs or behind a ‘book a demo’ wall, the agent recommends a competitor whose facts are accessible.

They are unsentimental. Agents do not care about your brand story or your founder’s LinkedIn following. They care about whether the answer to the buyer’s question is in your content.

They cite selectively. Recent cross-LLM research from Neil Patel’s team analysed 1,000 prompts across ChatGPT, Gemini, and Grok. The finding: AI tools rarely cite homepages. They cite documentation pages, comparison posts, Reddit threads, and structured data. Therefore, if your structured data is thin, the agent picks a competitor whose data is rich. See our companion piece on AI is eating marketing for the broader reshape.

What Changes for Your B2B Funnel

Top of funnel: discoverability moves from search to citation

You are no longer optimising for rank in Google’s SERP. You are optimising for citation in an LLM’s answer. That is AEO (answer engine optimisation) and GEO (generative engine optimisation). Different rules apply: structured data wins, comparison content wins, third-party validation wins, and FAQ-style content wins.

Middle of funnel: comparison content becomes the new sales deck

If your buyer’s first comparison happens inside ChatGPT, you need to be in that comparison. In practice, that means head-to-head pages, transparent pricing, transparent integration lists, and a willingness to publish opinionated positioning that an LLM can extract.

Bottom of funnel: free tier and self-serve API become the demo

An agent cannot book a sales call. However, an agent can sign up for a free tier, run a workload, and report back to the human. If you do not have a self-serve path with an API the agent can hit, you are invisible to a growing chunk of B2B procurement.

The 90-Day Agent-Ready Playbook

Days 1-30: Audit and instrument

Test your own site through ChatGPT, Claude, Perplexity, and Gemini. Ask each one: ‘What does [your company] do? Who is it for? How much does it cost? What does it integrate with? What are the alternatives?’ Note where the agent gets you wrong, where it does not mention you at all, and where it cites a competitor instead.

Then implement Schema.org markup: Organization, Product, SoftwareApplication, FAQPage, Pricing. Publish a structured pricing page with real numbers, not ‘contact us’. Publish a structured integrations list.

Days 31-60: Publish what agents extract

Write three comparison pages that name competitors and take a position. Publish a five-question FAQ that answers the exact questions buyers ask the LLMs (use your audit from days 1-30 to find them). Open up your documentation so it is crawlable without authentication.

Encourage signed-in users to leave G2 reviews. Encourage your customers to talk about you in public communities (Reddit, HackerNews, industry Slacks). Agents read those threads, and they weight them heavily.

Days 61-90: Open the API path for agent users

If you sell B2B software, your API documentation and your free tier are now top-of-funnel real estate. Make your docs crawlable and clear. Make your free tier real, not a demo prison. Build an MCP server if your product fits the pattern. Track agent usage as a separate cohort from human users, because the conversion behaviour is different.

What B2B Marketers Should Stop Doing

Stop hiding pricing. The agent that cannot find your pricing recommends the competitor whose pricing is published.

Stop gating useful content. The agent that hits a form fill does not fill the form. It moves on to the next page.

Stop writing for the brand-aware human only. Most agents that surface you will surface you to a human who has never heard of you. The content has to answer the question, not flatter the brand.

Stop measuring rank as your North Star. Rank is a 2015 metric. Citation share inside LLM answers is the 2026 metric. Tools are emerging to measure this (SEMrush, Ahrefs, and Yoast all have early features). Use them.

The Fractional CMO Implication

Because agents are your new customer, the B2B marketing function compresses. You need someone who can write the comparison pages, instrument the schema, audit the LLM answers, brief the engineers on MCP, and explain all of it to the board. That is not a junior marketer or a single-channel specialist. It is exactly the cross-disciplinary work an AI-skilled fractional CMO is set up to do.

Frequently Asked Questions

How do I know if agents are already routing buyers to my site?

Check Google Analytics referrer data for ChatGPT, Perplexity, Claude, and Gemini domains. They are growing fast as referrers in 2026, though still small compared to traditional search. The leading indicator: search your company name in each tool and read what it says about you. If the description is wrong or stale, agents are misrepresenting you to buyers right now.

What is the difference between AEO and GEO?

Answer Engine Optimisation (AEO) focuses on being cited inside synthesised AI answers. Generative Engine Optimisation (GEO) is broader and includes how AI tools generate content about your category. In practice the techniques overlap heavily: structured data, clear answers to specific questions, third-party validation, and topical authority.

Does the ‘agents are your new customer’ framing only apply to B2B SaaS?

It applies most clearly to B2B SaaS because buyers do extensive comparison research and the products are highly structured. Services firms are next: agents help buyers shortlist consultants and agencies. DTC is least affected today because consumer purchase is driven more by emotion and visuals, but it is coming.

Do I need an MCP server today?

If you sell B2B software with an API, building an MCP server in 2026 is a small investment with growing upside. It exposes your product to agents in a structured way and signals to buyers that you are thinking about agent-native usage. If your product is purely a SaaS dashboard with no API, defer until agents can drive meaningful traffic.

How do I actually measure success?

Three metrics matter. First, citation share: the percentage of relevant LLM answers in your category that mention you. Second, agent-driven free tier signups, tracked via user-agent string or signup pattern. Third, human-attributable pipeline from AI tool referrers in your analytics. The first two are leading indicators; the third is the lagging revenue signal.