A Colorado plumber received a call last month from someone asking for a quote on three bidets. The caller disclosed they were transcribing the conversation. The plumber's response, as reported by New York Magazine, was unprintable but perfectly understandable: he'd just been cold-called by an AI agent working on behalf of a customer he'd never meet.
Welcome to the new customer relationship. It has three parties now.
The Middleman You Didn't Hire
Meta's Muse launched on September 8, 2026, and crossed 5 million downloads in 22 days. That's faster than ChatGPT reached the same milestone. OpenAI followed with Dots, an always-on assistant that runs workflows across connected apps. Google has Gemini 4 Argon in limited preview. Anthropic offers desktop control through Claude's Cowork app.
These aren't chatbots. They're agents. The difference matters. A chatbot answers questions. An agent books your flight, negotiates your insurance premium, disputes your credit card charge, and completes your purchase while you're making dinner. The customer sets conditions; software handles the follow-through.
As MarTech.org put it: "Customer preferences, purchase decisions, and even checkout are moving beyond brands' control."
For marketers who've spent careers building funnels, optimizing touchpoints, and crafting brand experiences, this is the equivalent of discovering your carefully designed store now has a bouncer at the door who decides whether customers even see your merchandise.
The Amazon Standoff
On September 20, 2026, Amazon started blocking Muse from shopping on its site. The popup message was blunt: "Continued access by an unauthorized AI agent violates Amazon's Conditions of Use."
Two days later, PayPal announced customers could shop and check out through Muse at PayPal merchants worldwide. Shopify enabled Shop Pay checkout for Muse across all its stores the day before.
Three major players. Two opposite decisions. Within 72 hours.
Amazon's concern isn't irrational. The company generated over $68 billion in ad revenue last year, a business that depends on people browsing pages and seeing sponsored products. When an agent does the shopping, nobody sees the ads. The carefully constructed discovery experience, the "customers who bought this also bought" recommendations, the sponsored placements: all bypassed by software that just wants the best price on a stroller.
But here's the tension: Amazon is blocking a tool that represents what its own customers want. The person behind Muse is a real customer with real money. The agent is just their hands.
The Friction Advantage Is Dying
One newsletter put it uncomfortably well: "Some of the revenue your business calls predictable depends on customers having better things to do."
Think about that. How much of your retention comes from customers who haven't checked whether your renewal price is competitive? Who've put off switching because it means filling out forms and taking sales calls? Who haven't disputed that charge because recovering it would consume an evening?
Personal agents make that distinction expensive.
A customer can now tell their agent: find out why my premium increased, compare what I'm paying against competing offers, negotiate a better rate, and if my insurer won't match it, handle the switch. The agent does the work. The customer does the leaving.
Perch Insights reports that Muse passed three million weekly users within a month. These aren't early adopters playing with a toy. They're people delegating the tedious parts of being a customer to software that doesn't get tired, doesn't forget, and doesn't accept "please hold" as an answer.
When Bots Talk to Bots
Here's where it gets genuinely strange. Most support centers have been deploying their own AI agents to handle customer conversations. Now customers are bringing their own agents to do the talking.
The Collation describes the shift: "Almost overnight, we've shifted from how a business deploys CX agents to the more pressing question of the hour: How do a business's CX agents serve customers' personal agents?"
Your chatbot is now talking to their chatbot. The human is somewhere else, probably making dinner.
This isn't science fiction. Gartner found in August 2026 that 58% of people who used AI for a service task let it complete the task for them. The same research found customers were roughly three times more likely to use a third-party AI service than a company's own chatbot.
Your carefully designed conversational experience? It's being experienced by software that doesn't care about your brand voice.
The Protocol Wars
The industry is scrambling to figure out the rules. TechCrunch reports that Meta, Walmart, Stripe, Sierra, Genesys, and others have begun working on an open standard for how AI agents communicate with businesses. The goal: separate the good bots acting on behalf of customers from the bad ones.
But the blocking is chaotic. Amazon's block is deliberate policy. Walmart's blocks are often accidental, their human-verification buttons throwing out agents even though Walmart is officially a Muse partner. Users can't tell the difference. They just see a failed task and blame everyone involved.
One analysis identified three types of blocking:
- Deliberate policy decisions
- Accidental anti-bot measures catching legitimate agents
- Infrastructure that simply wasn't built for this
Only the first is actually a choice.
What Marketers Can Still Control
The temptation is to see this as purely defensive. Block the agents. Protect the funnel. Preserve the customer relationship as it was.
That's a losing strategy. The customer relationship as it was is already gone for millions of consumers. The question is what you do with the relationship as it's becoming.
First, recognize that agent-friendliness is becoming a competitive advantage. If your competitor's site works with Muse and yours doesn't, you've just made the agent's job easier: recommend the competitor. Shopify and PayPal aren't enabling agents out of charity. They're positioning for a world where agent-mediated commerce is normal.
Second, your data strategy matters more than ever. When an agent compares products, it's parsing structured data, following entity relationships, synthesizing across sources. As one analysis noted, if your best answers aren't clearly stated and machine-readable, you're invisible at the exact moment a buyer is forming their options.
Third, the brand experience needs to work at the API level, not just the human level. Your carefully crafted website might be bypassed entirely. What remains is your product data, your pricing, your policies, your reputation as synthesized by an agent that's read every review you've ever received.
The Real Question
Marketing has always been about influence. We influence awareness, consideration, preference, purchase. We've built entire disciplines around understanding and shaping the customer journey.
Now there's a new actor in that journey, one that doesn't respond to emotional appeals, doesn't notice your beautiful photography, and evaluates your brand based on structured data and aggregated sentiment.
The brands that win the next two years won't be the ones who successfully block agents at the door. They'll be the ones who figure out how to be the answer when an agent asks, "What's the best option for my human?"
Data tells you the what. Brand tells you the why. But when the customer's agent is doing the asking, you'd better make sure both are machine-readable.
