Personal AI Agents Are About to Expose Brand Loyalty as Mere Friction
Meta's Muse agent has topped 5 million downloads, and OpenAI's Dots works toward consumer goals in the background. Brands built on switching friction face a loyalty reckoning.
By Grace Kim
5 min read
What's News
- Meta's personal AI agent Muse has been downloaded more than 5 million times.
- OpenAI has introduced Dots, an always-on agent that pursues consumer goals in the background.
- The Personal Agent Protocol, developed with Meta, Sierra and others, defines how personal AI agents identify themselves and their permissions to businesses.
- Mastercard and Visa are building infrastructure for secure AI-initiated transactions.
- Success metrics shift from containment to intent-to-outcome as interactions become asynchronous.
Meta's personal AI agent Muse has been downloaded more than 5 million times — and that number is the warning shot for every brand that thinks its customers are loyal.
According to a new industry analysis, most of what businesses call customer loyalty is actually friction. People stay with the same wireless carrier, bank, insurer or subscription because switching takes effort: researching alternatives, comparing complicated offers, negotiating, filling out forms. Personal AI is about to make that work irrelevant.
Consumers will be able to give their AI an objective and the authority to pursue it: "Cut my monthly expenses by 10%, keep my travel within budget or ensure I'm always getting the best deal." When an AI can continuously evaluate whether a customer is reaching the best outcome, the analysis argues, businesses will discover how much of their retention was driven by inertia.
The takeaway is blunt: brands that survive on switching costs, not outcomes, are running out of time.
How fast is this moving?
Fast enough that infrastructure is already being built. OpenAI has introduced Dots, an always-on agent designed to work toward consumers' goals in the background. Google and others are racing toward the same future. And the industry has launched the recently announced Personal Agent Protocol — a joint effort involving Meta, Sierra and other companies to create a common way for personal AI agents to identify who they represent and what permission they have when interacting with businesses.
The author's own company is part of that effort, alongside Meta and Sierra. The fact that the industry is already building shared standards signals the shift: personal AI is moving quickly from something consumers experiment with to something businesses will have to serve.
Now imagine millions of personal AI agents arriving at businesses' front doors, only to encounter infrastructure that was never designed for them. Nearly every part of customer experience today — the journeys, the forms, the authentication systems — was built for a human. That's the collision coming for every brand.
What happens to the customer journey when a human isn't taking it?
Personal AI inserts a new layer between a business and the person it wants to acquire and keep. The consumer may never see the redesigned website, navigate checkout, or consider the retention offer built to keep them.
As personal AI use grows, it becomes the interface to the brand. Consumers tell their AI what they want to accomplish; the AI determines how to pursue the goal. Businesses will need to read that intent, collaborate with the customer's AI, and orchestrate their own people, AI and systems to deliver the outcome.
Metrics change too. As interactions become asynchronous, success shifts from containment to intent-to-outcome: did the customer accomplish their goal, and how much effort did it require?
The brands need new questions, per the analysis:
- Who or what is interacting with us, and who does it represent?
- What is the customer trying to accomplish?
- What has the AI been authorized to know and do?
- When do we engage with the customer versus their AI?
In short: the journey stops being a designed path and becomes a negotiated outcome.
Why identity alone won't cut it
Before businesses can serve personal AI at scale, they have to recognize it. Today, many companies can't reliably distinguish an authorized AI acting for a customer from the customer itself. And identity doesn't establish authority — businesses also need to know what the AI has permission to access, decide and do.
Consumers will set boundaries: negotiate my bill, but don't cancel my service. Rebook my flight, but don't spend more than $500. The analysis frames this as "a digital power of attorney for AI," where identity, authentication and delegated authority become foundational. Permissions need to travel with the agent. Actions need to be traceable, so consumers and businesses know what happened and under whose authority. When an AI hits a boundary or encounters a decision it hasn't been authorized to make, control returns to the human.
No single business can solve delegated authority on its own — which is where common standards like the Personal Agent Protocol become critical. Mastercard and Visa are similarly developing infrastructure for secure AI-initiated transactions. Without a shared trust layer, the analysis warns, personal AI may be capable of acting everywhere but authorized to act almost nowhere.
The one-line summary: trust has to become portable, or agents stall at the door.
Why are we still making AI pretend to be human?
There's a second structural obstacle. Personal AI is entering businesses through doors built for people: it browses websites, fills out forms, navigates chat, even makes phone calls.
The fix, per the analysis, is an agent-native front door — a secure way for authorized personal AI and enterprise AI to communicate directly, establish identity and authority, exchange context and coordinate action. Open standards such as A2A can create that path across vendors and architectures. Instead of making AI navigate an interface designed for people, agents would interact directly around the customer's intent.
The canceled-flight scenario shows the flow: a traveler's personal AI contacts the airline's AI, evaluates alternatives against the customer's preferences, and pulls in an AI-assisted employee when an exception arises. The traveler steps back in only when approval is required. Personal AI to enterprise AI. AI to employee. Employee to customer. And potentially back again. What matters is preserving identity, intent, context and authority as the experience moves between them.
That means the experience no longer belongs to a single channel — or a single participant.
What's the bottom line for brands?
Customer experience has moved beyond human-to-human interaction into a world where people, enterprise AI and personal AI may all participate in achieving one outcome. Interoperability alone won't create a trusted experience — orchestration determines how it actually gets delivered, connecting the right resources around what the customer is trying to accomplish.
And here's the squeeze: when personal AI makes it effortless to compare, negotiate and switch, businesses won't be able to rely on inertia to keep customers. They'll have to earn their loyalty with every outcome.
Source: unite.ai
More from Grace Kim
Show full bio
Market editor covering people, launches and day-to-day practice in entrepreneurship, startups and venture funding at Business Bearings.
2,203 articles