The First Wave of Agentic Verticals: Banking, Telecom, Utilities, and Why They Adopt First

Most discussions of agent-to-agent commerce treat the consumer-brand relationship as a single category. From a market-formation perspective, this is a mistake. The category will land in specific verticals first, and the verticals where it lands first share specific structural characteristics that other verticals do not. Building for the average consumer-brand relationship produces a product that is wrong for every specific relationship; building for the verticals where the structural conditions are aligned produces a product that has actual customers.

This paper identifies the four verticals where we expect the first wave of agent-to-agent commerce to land: banking, telecom, utilities, and insurance, and explains the structural conditions that put them at the front of the adoption curve. We also identify the verticals that are commonly cited as adoption candidates but whose structural conditions are misaligned, and which will trail by years rather than months.

The five conditions

Five structural conditions, in our analysis, predict early adoption of agent-to-agent commerce in a vertical. Verticals that score high on most of these conditions will adopt early; verticals that score low on most will trail.

High-frequency recurring transactions. Verticals where the consumer interacts with the brand many times per year produce more opportunities for agent-to-agent value creation per customer. A monthly billing relationship produces twelve opportunities; a one-time annual purchase produces one.

Significant friction in the current process. Verticals where the existing customer-experience model is widely understood to be hostile or inefficient produce strong consumer-side demand for an advocate. Telecom is a textbook example; restaurant ordering is not.

Material economic stakes per interaction. Verticals where individual transactions involve hundreds or thousands of dollars produce ROI for the agent in a way that low-stakes verticals do not. A bill negotiation that recovers $40/month is worth pursuing; a $0.50 quibble is not.

Structured, legible product taxonomy. Verticals whose products and pricing can be cleanly machine-read enable agents to compare offers, understand terms, and negotiate. Insurance policies, mobile-phone plans, and electricity contracts are structured; bespoke services are not.

Existing regulatory pressure on customer-experience practices. Verticals where regulators are already attentive to the customer-experience model produce both buyer demand for compliance-improving tools and policy support for agent-to-agent intervention. Click-to-cancel, dark-patterns enforcement, and bill-shock regulations all create regulatory tailwinds for agentic deployment.

Why banking, telecom, utilities, and insurance score high

Each of the four verticals we expect to lead the first wave scores high on four or five of these conditions. Briefly:

Banking. Recurring transactions: very high (account fees, card transactions, service charges, recurring assessments occur monthly or more often). Friction: high (consumers report the financial-services customer-experience as among the most frustrating). Stakes: very high (overdraft fees, denied disputes, mispriced loans accumulate to material annual sums per household). Taxonomy: structured (account types, fee schedules, interest rates are all machine-readable). Regulation: significant and growing (CFPB, regional bank regulators, deposit-insurance frameworks all create compliance pressure).

Telecom. Recurring transactions: very high (monthly billing). Friction: extreme (telecom customer-service is the textbook case of engineered hostility). Stakes: significant (typical household telecom spend is $1,500–$3,000/year, with negotiable elements throughout). Taxonomy: structured (plans, features, promotions are machine-readable). Regulation: significant and rising (FCC, state utility commissions, click-to-cancel obligations).

Utilities. Recurring transactions: very high (monthly billing, demand-side fluctuation). Friction: moderate to high (limited consumer choice in some markets, but high friction where competitive markets exist). Stakes: very high (annual utility spend in the US averages over $4,000/household). Taxonomy: structured (rate plans, time-of-use schedules, demand charges). Regulation: extensive (state utility commissions, energy-policy frameworks, evolving consumer-protection rules).

Insurance. Recurring transactions: moderate (annual or semi-annual policy renewals, claims as needed). Friction: very high (claims processes are widely reported as adversarial, renewal pricing is opaque). Stakes: very high (auto, home, health insurance involve thousands per year and can produce five-figure outcomes during claims). Taxonomy: structured (policy terms are highly formalized). Regulation: extensive (state insurance commissioners, federal frameworks for specific lines).

These four verticals collectively account for a significant share of the typical household's recurring vendor relationships and the bulk of the annual dollars left on the table by friction. They are also the verticals where consumer-side advocacy services: Rocket Money, Billshark, Trim, Hello Trove, have already proven that demand exists for the analog version of what agent-to-agent commerce will do at machine speed.

Why retail e-commerce trails

The most commonly assumed first vertical for agentic commerce is retail e-commerce. Almost all of the early consumer-agent companies: Phia, Daydream, Operator, have positioned themselves there. We expect retail e-commerce to be a meaningful market for shopping agents specifically, but to trail the four high-friction verticals above as a market for agent-to-agent commerce more broadly.

The reason is that retail e-commerce is structurally well-served by the existing internet. The friction is real but bounded: a consumer can compare prices on three sites in five minutes, read reviews, and make a decision without calling anyone. The economics per transaction are low: most e-commerce purchases involve $50–$200, which limits the agent's capacity to capture value for either side. The taxonomy is structured but already exposed to consumers through search, comparison shopping, and price-tracking tools.

Retail e-commerce is a great market for shopping discovery and price comparison. It is a less compelling first market for agent-to-agent negotiation, dispute, retention, and the high-friction conversation flows where agentic commerce produces the most value. The companies building consumer agents specifically for shopping should expect their TAM to be smaller than the agentic-commerce TAM as a whole, even though the consumer adoption may be faster.

What this means for go-to-market

The implication for vendors is to plan vertical-by-vertical, not horizontal-everywhere. Specifically:

Brand-side vendors should win one of the four leading verticals first. A brand-side agent platform that wins banking, telecom, utilities, or insurance has a credible path to defining the agent standard for that vertical and then extending. A platform that pursues all four simultaneously will dilute its product and lose to a focused competitor in each.

Consumer-side vendors should build for the high-friction verticals first. A consumer advocate that handles banking, telecom, utilities, and insurance well has a stronger product than one that handles every category mediocrely. The advocate that demonstrates real recovery in the high-friction verticals will earn the trust to extend into the lower-friction ones.

Marketplace operators should sequence their participant onboarding by vertical. A marketplace that achieves liquidity in one vertical first will compound network effects there before extending. The temptation to onboard horizontally will produce a marketplace that has many participants but no liquid sub-market, which is the worst outcome for two-sided network economics.

Investors should size verticals carefully. The total addressable market for agent-to-agent commerce should be summed across the verticals where adoption will actually occur, weighted by the addressable sub-market within each. The flat-line "all consumer-brand interactions" framing produces TAM numbers that look impressive and predict little. The vertical-specific framing produces smaller numbers but better forecasts.

Sequencing within the wave

Within the first wave, we expect the verticals to adopt in roughly the following order: telecom (highest friction, well-established adoption pattern through Rocket-Money-class products), banking (high stakes, strong regulatory tailwinds), insurance (extreme friction in claims specifically, slower in renewals), utilities (highest stakes per dollar, but slowest because consumer choice is limited in many markets).

Beyond the first wave, we expect healthcare-billing-and-benefits, subscription services more broadly, and government-services interactions to be the second wave, arriving in 2027–2028. Retail e-commerce, hospitality, and bespoke services will be the third wave, arriving 2028–2030.

These timelines are predictions, not commitments. They will accelerate or slow based on regulatory signals, capital availability, and the speed at which the trust-layer infrastructure matures. The structural ordering, however, is robust to those variables.

Closing

The vertical-by-vertical analysis is the most useful predictive frame we have for understanding where the agentic-commerce category will land first. It is the frame we use internally for product decisions, partnership prioritization, and capital planning. We expect to revise the analysis as new data accumulates from the first deployments at scale, but the structural conditions: high frequency, significant friction, material stakes, structured taxonomy, regulatory pressure, will continue to predict adoption order for the foreseeable future.

Builders, investors, and brand-side strategists who internalize the vertical-specific frame will make better bets than those who treat agentic commerce as a single horizontal market. The opportunity is large, but it is shaped, and the shape determines who wins.