Beyond Banks: Mapping the Winners and Losers of the AI Agent Economy

Meta's viral AI agent Muse put a spotlight on one specific worry for banks — that AI could finally break the customer inertia banks have quietly profited from for years. But zoom out, and the same underlying shift — software that acts on a person's behalf instead of just answering questions — touches a much wider set of industries than banking alone. Here's a sector-by-sector look at where autonomous AI agents could create winners and losers, and why.

Edge cloud infrastructure: Cloudflare, Fastly

Autonomous agents don't browse the way people do — they continuously poll APIs, inspect endpoints, and execute programmatic tasks, which drives a multi-fold increase in edge routing requests and API calls compared with a human clicking through a website. That creates two distinct opportunities for edge platforms. First, web properties need a reliable way to tell a malicious scraper apart from a legitimate consumer agent like Muse, which is pushing demand for cryptographic verification, rate limiting, and specialized web application firewalls. Second, decentralized execution environments — Cloudflare Workers, Fastly Compute — become more valuable as they cut the latency involved in multi-step automated transactions. Both dynamics point toward accelerating enterprise contract values and expansion in API gateway and edge compute monetization for these platforms.

Digital fraud and risk intelligence: Riskified

Traditional fraud signals — keystroke patterns, mouse jitter, screen dwell time — simply don't exist when software, not a human, is executing the purchase. That's a genuine problem for legacy fraud-detection models, but also a reason risk-intelligence platforms are moving quickly: Riskified has already announced an expansion of its AI-agent-intelligence tools aimed specifically at securing native merchant AI shopping assistants. The upside case is a broad expansion of the addressable market for machine-to-machine identity and transaction-guarantee platforms, as bad actors also start deploying adversarial agents to exploit pricing errors, exhaust inventory, or cycle compromised credentials at scale. The risk cuts both ways, though: mispricing risk on automated agent flow could just as easily trigger a spike in guarantee payouts for these platforms rather than a windfall.

Telecommunications: T-Mobile, Verizon

Wireless carriers have long benefited from a version of the same inertia banks rely on — subscribers who stay on outdated rate tiers rather than chasing better promotions. An AI agent capable of monitoring usage and executing an eSIM carrier swap on a customer's behalf increases that competitive friction directly. Automated billing audits could also strip out some of carriers' highest-margin add-ons, like device insurance or premium technical support. The bigger structural risk is what's sometimes called the "dumb pipe" problem: if the economic value of financial execution and personal assistance increasingly accrues to whichever platform runs the AI agent, carriers are left absorbing rising data throughput without the pricing power to match it — a dynamic that would show up as slower ARPU growth and higher retention costs.

Banking and wealth management: a quick recap

We've covered the banking side of this story in more detail separately, but the short version: idle checking, savings, and brokerage balances have long been cheap funding for banks. Agents that programmatically sweep that cash into higher-yielding accounts, Treasuries, or money market funds threaten to force banks toward costlier wholesale funding — FHLB advances, CDs — to replace it, compressing net interest margins. In-house automated cash products, like JPMorgan's planned Smart Cash, are one defense; whether they're enough is still an open question.

Online travel and transaction brokers: Booking, Expedia, Arthur J. Gallagher

Travel and insurance intermediaries have historically captured 10% to 25% take-rates by owning the top of the search funnel. An agent that can query airline, hotel, and underwriter APIs directly bypasses that aggregator layer entirely, and in doing so erodes both brand loyalty and standard renewal commissions by removing the comparison-shopping friction that used to keep customers coming back to a single portal. The likely outcome, if this trend continues, is sustained pressure on marketing leverage and transaction take-rates, pushing these intermediary models toward more specialized, bespoke enterprise services rather than broad consumer aggregation.

The broader macro picture

Three themes tie these sector stories together. First, a "deposit beta squeeze": as customer inertia drops toward zero across banking and beyond, deposit betas — how quickly customer rates move with market rates — are likely to rise faster than historical models predict, forcing institutions to compete directly on yield rather than counting on stickiness. Second, an infrastructure redistribution: capital appears to be shifting away from customer-facing portals and aggregators and toward the edge networks, identity-verification clearinghouses, and primary agent hosts that sit underneath the new machine-to-machine layer. Third, faster regulatory evolution: as financial agents see wider adoption, scrutiny of programmatic liability, Open Banking data-access rules, and autonomous transaction audit trails is likely to accelerate.