Meta’s new AI agent, Muse, is gaining popularity for its ability to act as a personal concierge for consumers: It can negotiate cable bills, book flights, cancel subscriptions, shop and pay for goods and do many other tasks. The field is getting crowded: OpenAI has Dots, Amazon has rolled out Alexa+, Google has Gemini Spark, and more.
This trend will force businesses to rethink customer service systems designed for human interactions. It could change how companies compete for customer loyalty as AI makes it easier for consumers to compare prices and switch providers.
Mike Szilagyi, head of product at Genesys, a major customer experience technology company that handles over 33 billion interactions annually, said businesses should prepare for the transition.
Genesys is working with Meta, Sierra and other companies on the Personal Agent Protocol, an emerging open standard designed to help personal AI agents securely identify themselves, obtain authorization and interact with businesses.
In an email Q&A with The AI Innovator, Szilagyi discusses this changing landscape.
The AI Innovator: How will personal AI agents change the way consumers interact with companies, and what will that mean for traditional customer service?
Mike Szilagyi: Personal AI will have a significant impact on the customer journey, changing how consumers engage with a brand, how businesses serve them, and how success is measured. It challenges a fundamental assumption of customer experience today that the entity interacting with a business is always human.
In the future, the best customer experiences may be the ones a human customer never had to join. Consumers may be able to give an AI a goal – such as cancel my subscription, rebook my flight, dispute this charge, find me a better plan – and let it do much of the work on their behalf. The customer remains in control of the outcome but may not take part in the interaction at all.
Businesses may increasingly serve either their customers directly or through their AI agents, or both, within the same customer journey. This will require them to preserve context as they orchestrate interactions across personal AI, enterprise AI, employees and customers to drive an outcome.
As a result, measuring the quality of customer service shifts from a focus on interactions with the customer to the outcome, making traditional metrics such as average handle time less valuable. What matters is whether the customer’s goal was achieved and how little effort they had to exert. This will also shift how customer loyalty is won and maintained.
Most companies built their customer experience systems for humans, not AI agents. What technology or infrastructure will businesses need to change to support agentic interactions?
Today, organizations are essentially making AI pretend to be human. Personal agents have to browse websites, fill out forms, navigate chat or even make phone calls because those are the front doors businesses built for people. Over time, enterprises will need an agent-native front door that provides a secure, direct way for authorized personal AI and enterprise AI to establish identity and authority, exchange relevant context and coordinate action.
That will require open, interoperable standards alongside orchestration. Technologies such as A2A can help agents communicate across vendors and architectures, while MCP can connect AI to the enterprise systems, tools and data needed to take action.
Emerging efforts such as the Personal Agent Protocol could provide another piece of that infrastructure by creating a common framework for personal agents to discover business services, authenticate themselves and obtain authorization. Businesses will then need an orchestration layer that preserves intent, context, permissions and governance as the customer experience moves through personal AI, enterprise AI, employees and the customer.
Personal agents could potentially carry enormous amounts of information about their users. What new security, privacy or identity risks do they create for businesses?
The challenge isn’t simply identifying an AI agent; it’s establishing who it represents and what that person has actually authorized it to do. There’s a big difference between telling an AI to find a flight and giving it permission to spend $800 and book one. Businesses will need trusted mechanisms for identity and delegated authority, with clear boundaries around what an agent can access, decide and do and when control needs to return to the human.
A personal AI may know far more about someone than any individual company should. Receiving data the business doesn’t need or shouldn’t hold creates risk for the enterprise and the consumer. Only the identity, profile and context necessary to complete a task should be shared with the business. Actions also must be traceable so both consumers and businesses can understand what happened, why and under whose authority.
If personal AI agents increasingly choose which companies to interact with, how does that change the way brands compete for customer loyalty?
Personal AI could fundamentally change the economics of customer loyalty because it removes much of the friction that makes them stay with a provider if it’s too troublesome to switch. Some of what businesses think of as loyalty may actually be inertia, and AI weakens inertia as a retention strategy.
The agent can continuously compare hundreds of options, negotiate a better deal or switch services without getting tired or frustrated. That means brands will have to compete much more directly on value, experience and outcomes.
But it also raises an important question: Whose interests is the personal agent actually serving? If agents favor certain providers based on commercial relationships, embedded preferences or how they were designed, brands may find themselves competing not only for the customer’s loyalty, but for the agent’s recommendation. Businesses will have to give both consumers and their AI a compelling reason to keep choosing them.
What is the biggest hurdle you’re hearing from companies about why they’re still not prepared to cater to personal AI agents?
The biggest hurdle is the mismatch in speed: consumer AI is evolving in months, while enterprises have spent years building customer journeys, identity systems, workflows and governance models around people. Now an AI agent may arrive on the customer’s behalf, and enterprises need to know who it represents, what it’s authorized to do and how to engage with it securely.
The first instinct for some businesses may be to block these agents because their systems weren’t designed for them. I think that’s the wrong direction. Businesses should be preparing to embrace personal AI and enable their own enterprise AI agents to interact with it directly.
Forcing one AI agent to navigate a website, fill out a form or make a phone call to another AI is inefficient. The better path is direct, secure agent-to-agent interaction that allows personal and enterprise AI to exchange the right context and work toward an outcome.
That requires infrastructure the industry is still developing. The Personal Agent Protocol is an important step because it can give enterprises a consistent way to identify personal AI agents, authenticate the people they represent and understand their delegated authority. That kind of common trust layer will be necessary if businesses are going to support these interactions at scale.
There are still pieces that need to come together, from identity and security to interoperability, data and governance. But one of the most important steps businesses can take now is building the orchestration layer that can manage experiences across all of these participants.
The future of customer experience will move fluidly between AI-to-AI, AI-to-human and human-to-human interactions, with context, permissions and intent carrying across each handoff. Personal AI is helping usher in that future. Businesses that build for orchestration now will be better prepared to serve consumers however they choose to engage, rather than having to redesign their customer experience every time a new type of interaction emerges.
Get the latest insights about enterprise AI.
Subscribe to our newsletter. Thank you.




Be First to Comment