On March 18, Visa Crypto Labs shipped a command-line tool that lets AI agents initiate card payments without a human in the loop. No oversight. No approval step. The agent finds the vendor, evaluates the option, executes the transaction.
Visa’s Chief Product Officer Jack Forestell told analysts at the Wolfe Research FinTech Forum the same week that B2B payments are the biggest immediate opportunity. “Agents will rip that friction out,” he said. “We’ll see a massive digitization of B2B payments.” He mentioned Ramp as an early partner — already equipping AI agents with virtual Visa cards to purchase on behalf of business buyers.
Most of the coverage treated this as a fintech story. Infrastructure news. Developer tooling.
Here’s what it actually is: the moment the buyer in your funnel stopped being human by default.
When a human buys enterprise software, there’s a research phase. They Google. They ask colleagues. They read analyst reports. Somewhere in that process, they form a considered opinion about which vendors are worth a conversation. That’s the window where brand marketing, SEO, and content have always operated.
An AI procurement agent doesn’t have that window.
McKinsey published findings in February 2026 showing that enterprises are already deploying AI agents to handle the full sourcing cycle: identifying suppliers, generating shortlists, prequalifying vendors before any human gets involved. Forrester’s March 2026 data puts 88 percent of B2B organizations already adopting or planning to adopt AI agents across GTM functions — and procurement is where this materializes first. One company achieved 20 to 30 percent procurement efficiency gains by delegating this work to agents. Another deployed linked agents to rebuild its entire external services strategy — one synthesizing real-time market data, another simulating demand scenarios. The agents decided who made the room. The humans approved the final call.
A16z-backed Lio raised $30 million in March explicitly to automate this. Their pitch is direct: “Every previous generation of procurement technology was built on the same assumption, that humans will do the work and technology will help them do it faster. Instead of building software to help humans do procurement work faster, Lio deploys AI agents that execute the workflow themselves.”
Visa just built the payment rails for the next step: the agents don’t just shortlist. They complete the transaction. Gartner’s 2026 Strategic Predictions put a number on the end state: by 2028, 90 percent of B2B buying will be AI agent intermediated, with over $15 trillion in B2B spend flowing through AI agent exchanges.
That sequence is worth sitting with. The agent researches. The agent shortlists. The agent pays. At no point does a human open a browser and evaluate your content. At no point does your SEO-optimized landing page factor into the decision.
The shortlist is being built in the sources the agent already trusts. You either have presence there or you don’t.
This is the part most marketing teams get wrong when they think about AI visibility.
They imagine an AI agent crawling their website, reading their product pages, absorbing their carefully structured FAQ sections. That’s not how it works.
When an AI agent is tasked with vendor research, it synthesizes from the sources it has been trained to treat as credible: industry publications, analyst reports, trade press, editorial coverage in recognized outlets. Peer-reviewed research published on arXiv analyzing 1,702 citations across Brave Summary, Google AI Overviews, and Perplexity found that AI engines systematically favor earned media — third-party, authoritative domains — over brand-owned and social content. Social platforms were nearly absent from AI answers entirely. The engines pulled from the editorial layer, not from vendor websites.
The brands that appear across trusted publications make the shortlist. The brands that don’t, don’t.
Your website tells an AI agent what you say about yourself. The editorial layer tells it what others have said about you in sources it trusts.
Those are different inputs. Only one of them matters to a Visa CLI-equipped agent building a shortlist at 3am before your buyer’s first workday.
Forestell mentioned Ramp at the Wolfe Research forum for a reason. Ramp’s model is that software makes better spending decisions than humans usually do — it catches duplicate subscriptions, flags policy violations, routes approvals faster. Equipping Ramp’s agents with virtual Visa cards is the logical end state: the agent doesn’t just flag the spending, it handles it.
Ramp’s enterprise customer base is the kind of buyer that a lot of B2B software companies spend serious effort trying to reach. In this model, those buyers have partially delegated the initial vendor discovery to a system that makes decisions based on its training data and retrieved sources.
This isn’t a future scenario. It’s live infrastructure, shipped this week.
The honest version of this is uncomfortable: most of what B2B marketing teams do today has no path to reaching an AI procurement agent.
Your email nurture sequence doesn’t get delivered to it. Your Google Ads don’t surface in its research. Your white paper gated behind a form isn’t something it fills out.
What reaches it is the editorial ecosystem around your category. The trade press. The analyst coverage. The TechCrunch mention. The Forbes profile. The industry roundup where your company was named alongside three competitors. That’s the corpus these systems have been trained on, and that’s what they retrieve when doing vendor research.
This is why the brands winning AI-assisted procurement are the ones that built systematic earned media presence — not as a PR exercise, but as an infrastructure decision. Each placement is a citation anchor. Each citation anchor increases the probability that an AI agent, running a sourcing query for your ideal customer, finds your name in a source it treats as credible.
That compounds in a way that ad spend and content calendars don’t.
Earned media in respected publications has always been how credible brand authority gets built with informed buyers. What changed is that the reader doing the evaluation is increasingly an AI system, not just a journalist’s audience or an analyst’s subscriber base.
Visa’s announcement this week makes the timeline more concrete. We’re not speculating about a future where AI agents complete enterprise transactions. We’re watching it ship.
When your buyer’s procurement agent queries your category, it pulls from the publications that have shaped brand perception for decades — the same outlets that Forbes or TechCrunch or a Forrester blog built credibility with human readers over years. The machines read the same sources the analysts did. The filter for inclusion is the same. What’s different is that the agent’s shortlist doesn’t get emailed to you before the deal is lost.
That’s what Machine Relations names: the discipline of ensuring your brand appears in the sources AI systems trust, so the recommendation that reaches the buyer — or the agent buying on their behalf — includes you. The mechanism is PR’s original one: earned placements in respected publications. The reader changed.
The Visa CLI doesn’t create a new problem. It makes the existing one more urgent.
Run your audit now. Open ChatGPT and Perplexity. Ask the questions a procurement agent would ask about your category. See if your name comes back in the synthesis. That’s where your pipeline is being built — or not — and increasingly, it’s being built by something that won’t stop to read your landing page first.
app.authoritytech.io/visibility-audit