Meta introduced a personal AI agent called Muse on September 8. It sends email, books trips, fills in forms, negotiates on the user's behalf and makes purchases. That capability list lands directly on the loyalty business — airlines, hotels, co-branded credit cards — an industry that, by the account of the people who run and advise it, earns a meaningful share of its money from customers who do not do the arithmetic. An agent that computes the real value of every card, reward and promotion in real time takes that inattention away.
The question loyalty executives have asked for decades is who owns the customer: the airline whose name is on the program, the bank that issued the co-branded card, the hotel that recognizes the guest at check-in, or the technology platform that sees the transaction. The answer was usually all of them at once. The new entrant is the consumer's own agent, and it is not loyal to any brand. It is loyal to the person's priorities — price, convenience, points, status, time and the quality of the experience.
Jaclyn Wands, vice president of product and AI at Phaedon, argues that AI assistants are the new gatekeepers, because they will calculate the economics of points, know account balances and hold access to the user's loyalty programs.
She demonstrated a personal assistant prototype called Ezer. In her scenario the agent tracked concert presales, checked them against her calendar, compared travel options, checked her loyalty balances, advised which credit card to pay with, and found a relevant store offer. It also caught an error in her own instructions: the venue could be reached by car rather than flown to. Wands says Ezer holds her to-do list, has access to her email and messages, is connected to her American and Marriott accounts, and can book hotels.
The point of the demo was not that an agent can complete every transaction flawlessly today. It was that over time it can consolidate many scattered decisions into a single consumer operating system.
Meta says Muse runs in a separate virtual machine, and that a second agent, Sentinel, reviews actions on the web and asks permission before sensitive steps such as sending an email or completing a purchase. Specific capabilities will change and the technology can still get things wrong, but the direction is not ambiguous: AI is moving from answering questions to coordinating decisions and acting on its own. Wands says the problem is not the idea of AI — plenty of talented people are already using the current version and will fix its shortcomings.
What Muse does is move Wands's prototype from a conference demo into a mass-market question, and the mass market is where loyalty programs have their exposure. Many of these programs benefit from friction and inattention. Customers forget about points, miss promotions, pay with the wrong card, overlook expiry dates, redeem rewards at unfavorable rates. The rules can be complex enough that only enthusiasts consistently extract what the program offers.
An always-on agent turns millions of ordinary members into power users. It can tell a traveler to stay on one co-branded card until they have accumulated points for a companion ticket, redirect spending to a different card to get lounge access, transfer points during a bonus window, or choose a hotel on the combined value of status, room upgrades, location and price. Wands notes that Ezer will be able to compute points economics in real time, taking that work off the user entirely.
Sylvie Uziel, chief executive and co-founder of Blue Bridge Group AI, sees the same pressure. In her view, AI agents will expose the loyalty programs that have relied for years on complexity and consumer inertia. Once a machine instantly calculates the real value of each point, perk and offer, customers have less patience for programs that do not deliver actual benefit.
Wands expects affluent customers to stay enrolled at first rather than walk out. But their points economics becomes transparent and fast enough that they learn to use their programs more effectively, and the ones that fall short of expectations are the first to be abandoned. The largest exposure is to wealthy frequent travelers: more accounts, more perks to optimize, more money at stake, and a higher willingness to experiment with anything that saves time. For that customer the agent is not a search engine but a concierge, financial optimizer and purchasing agent at once.
None of this makes loyalty purely transactional. When rewards become easier to compare, the quality of the experience itself can matter more. Wands cites the Delta and American Express relationship as an example of shared value — the bank gets detailed transaction data, Delta controls a large part of the travel experience and of the member relationship — and says what differentiates now is Delta's service rather than the loyalty program. AI can optimize points math. It cannot make a delayed flight arrive on time, turn an indifferent employee into an attentive one, or make an ordinary hotel stay memorable. A brand that reliably delivers a good experience can stay the preferred choice even when the agent finds something slightly cheaper.
So the programs that work will combine two kinds of value: the measurable kind — rewards, access, upgrades, convenience, savings — and the emotional kind, meaning trust, recognition of the customer, and confidence that the brand will not fail at the moment it matters. Wands's framing is that the technology becomes a baseline requirement and the experience decides.
That leaves loyalty executives with a task that is not building a chatbot. It is making their programs legible and valuable to machines acting on behalf of people, which requires accurate product data, available inventory, transparent rules, useful APIs, clear redemption value and permissioned access to customer data. It also requires restraint. An agent with access to email, messages, calendar, financial accounts and loyalty profiles raises serious questions about privacy, security, user consent and the consequences of a technical error.
Phil Alexander, founder and chief executive of AnswerMyQ, states the economic threat plainly: many loyalty programs quietly make money from confusion, and agents remove the confusion. If a program's value can be calculated in real time, the only durable advantage left is a genuine reason to choose it.
Alexander has the mechanism right, and the industry's prescribed response to it is the part worth watching. Brands are being told to publish clean product data, open useful APIs and expose redemption value so that an agent can evaluate them properly. Publishers were told the same thing about search engines, and the end state was an intermediary that set the terms of access to an audience it did not create. A loyalty program that makes itself perfectly legible to agents has not won the agent's recommendation. It has only made itself comparable.
The more interesting question is who pays the agent. Nothing in this picture explains how Ezer or Muse makes money, and an assistant sitting between a consumer and every purchase they make is the most valuable placement in retail. If the agent is paid by the consumer, loyalty programs are meeting a genuinely rational customer for the first time and Alexander's version holds. If it is paid by the brands, the friction has not been removed — it has moved one level up, into a ranking the customer cannot inspect and does not know is for sale.
There is also the reliability problem, which the products handle by admitting it. Meta's answer to an agent that transacts on your behalf is a second agent watching the first and asking permission before it sends an email or buys something. That is sound design, and it is also a statement about how far the first agent can be trusted alone. The failure mode here is not a mediocre hotel recommendation. It is a booked, paid, non-refundable one.
Which sets up the actual contest. The complexity in these programs was never incidental; by the account of the people quoted here, it is where a good deal of the margin sits, and the agent dissolves it. What replaces it is either a market in which brands compete on service and honest value, or one in which they compete for position inside somebody else's assistant. The second is the same business as before, run one layer further from anywhere the customer can see it.