Launch white paper · Agentic commerce · July 2026

The missing merchant control plane.

Why commerce built for humans fails AI agents, and how the new channel becomes observable, reliable, and economically valuable.

Get the dispatchesRun a free scan

The thesis

Protocols make agentic commerce possible. Merchant-side observability, testing, and optimization make it operable.

Aisle owns the missing control layer between fast-moving agent ecosystems and the operational reality of a live storefront.

$3T to $5TPotential global commerceMcKinsey estimate of consumer commerce orchestrated by agents by 2030.
8Operational bottlenecksFrom product truth and access through security, observability, and aftercare.
7Control-plane jobsDiscover, simulate, diagnose, remediate, observe, optimize, and certify.

The opportunity

Make every storefront legible, transactable, and trustworthy to agents.

01 / The shift

Agentic commerce changes who controls the journey.

The interface is becoming conversational, but the deeper shift is in demand, state, authorization, and economics.

Traditional ecommerce assumes a human session: a shopper interprets visual merchandising, navigates a taxonomy, repairs errors, and confirms payment. Agentic commerce begins with a goal. The agent converts intent into constraints, searches and compares across merchants, assembles a transaction, and increasingly acts within a bounded mandate.

Core argument

The first wave is a distribution shift. The durable opportunity is infrastructure.

Three execution paths coexist

01Browser automation
Immediate reach because agents use the same sites and forms as people.

Limit: Fragile selectors, CAPTCHAs, login and MFA, dynamic state, anti-bot defenses, and prompt-injection exposure.

02Platform-native checkout
Designed flows provide clearer state, control, and confirmation.

Limit: Coverage depends on participating merchants, geographies, payment methods, and channel-specific rules.

03Protocol / API commerce
Structured discovery, checkout, authorization, and payment capabilities.

Limit: Standards are young, overlapping, unevenly adopted, and still rely on merchant operations.

An end-to-end state chain

  1. 1IntentGoal, preferences, budget, timing
  2. 2AgentPlanning, memory, tools, comparison
  3. 3TrustIdentity, mandate, authorization, audit
  4. 4DiscoverCatalog, price, inventory, policies
  5. 5ExecuteCart, tax, shipping, checkout, payment
  6. 6AftercareFulfillment, support, returns, loyalty

Three structural changes

01Machine-mediated demand
A product must be selected by an agent before it is seen by the consumer. Inclusion becomes an upstream revenue event.
02Multi-party state machine
The user, agent, merchant, payment provider, identity layer, and support systems must agree on state and authority.
03A machine interface
The storefront is not only visual. Accessibility, semantics, policy structure, and operational reliability now affect revenue.

Operating reality

The transaction is only as reliable as the weakest link. Discovery is moving faster than execution, and execution is moving faster than full delegation.

02 / The bottleneck

The bottleneck is operational reliability.

Model intelligence is no longer the only, or even the primary, constraint on a successful transaction.

The readiness gap is measurable

78%Had robots.txtMostly designed for search crawlers, not agents.
4%Declared AI usage preferencesA basic signal for controlling agent use.
3.9%Served markdown on requestA machine-efficient format that reduces parsing cost.
<15Used selected emerging standardsSelected MCP/API standards appeared on fewer than 15 sites.

Cloudflare scan of 200,000 high-traffic domains, April 2026.

Eight operational bottlenecks

01Product truth
Agents need accurate, structured, current data on variants, attributes, inventory, price, shipping, restrictions, and policies. Gaps cause wrong recommendations, silent exclusion, and checkout surprises.
02Access and legibility
Sites are designed for visual browsers and crawlers, not authenticated task-performing agents. Agents misread content, waste context, hit bot defenses, or cannot reach the needed capability.
03Protocol fragmentation
ACP, UCP, AP2, payment-network frameworks, and platform-specific integrations overlap but do not remove channel differences. Merchants face duplicative onboarding and changing specs.
04Identity and intent
A merchant must know which agent is calling, whom it represents, what was authorized, and under what limits. Unverified automation looks like abuse; weak mandates increase fraud and disputes.
05Checkout state
Tax, inventory, shipping, promotions, subscriptions, account state, and payment risk must agree in real time. A small mismatch can invalidate a cart after the agent has optimized around it.
06Payments and liability
Credentials must be tokenized and scoped, while merchants and issuers retain familiar risk and recourse controls. New initiators create ambiguity around consent, chargebacks, and evidence.
07Agent security
Browser agents ingest untrusted content and use tools with real-world permissions. Prompt injection, tool misuse, data leakage, and cascading actions can turn a shopping task into a security incident.
08Observability and aftercare
Merchants need to see what an agent attempted and carry context into fulfillment, returns, support, and loyalty. Today, failures are invisible, attribution is weak, and teams lose the original mandate.

Trust problem

Good agents and bad bots look similar.

Authorization problem

Authorization must travel with the action.

The bottleneck beneath the bottlenecks

Merchants cannot improve what they cannot observe. The rails are advancing faster than the operating tools.

03 / The merchant

The merchant paradox: more demand, less control.

Agentic commerce can compress the path from intent to purchase while moving discovery, comparison, and checkout into interfaces the merchant does not own.

Merchant objective

Maximize agent-driven demand without surrendering the merchant system. Preserve product truth, brand differentiation, commercial policy, customer recognition, risk controls, attribution, and post-purchase ownership.

01Interface control
The agent can summarize, compare, or transact without showing the storefront. Brand cues, education, merchandising, and emotional differentiation weaken.
02Integration burden
Each surface can have different data, onboarding, checkout, payment, and policy requirements. Engineering and operating costs rise before channel economics are proven.
03Discoverability
Agents rank against machine-readable attributes, evidence, availability, and user constraints. Traditional SEO, creative, and trade spend do not guarantee inclusion.
04Customer relationship
Identity and consent may be mediated by the agent platform. First-party data, loyalty recognition, and lifecycle marketing can weaken.
05Economics and attribution
Agents compare price, delivery, promotions, and substitutes in real time. Margins compress, paid placement may emerge, and existing attribution remains incomplete.
06Fraud and liability
An agent is a new transaction initiator with unfamiliar signals. Merchants need proof of authorization, defensible risk decisions, and dispute evidence.
07Operational readiness
Catalog, inventory, tax, fulfillment, returns, and support systems may disagree. Edge cases that humans repair become automated failure points.
08Organizational ownership
Commerce, marketing, product, payments, fraud, data, security, and legal all have a stake. No single team owns agent readiness or the full outcome.

The funnel becomes partially invisible

Human ecommerce

RankClickSessionCartOrder

Agentic ecommerce

InclusionWin rateTruthTask successAftercare

Invisible failure examples

  • A missing dimension removes the product from the candidate set.
  • A return policy is misread before a session begins.
  • Bot controls block the agent before the cart.
  • Tax or shipping changes after the agent optimizes.
  • Loyalty status is not recognized.

Products can be excluded before an impression or click appears in analytics.

What merchants need

A neutral layer that turns fragmentation into a readiness score, repeatable test suite, issue backlog, live dashboard, and measurable remediation.

04 / The consumer

Consumers want leverage, not blind delegation.

People are more willing to let AI reduce work than replace judgment. Trust determines how far autonomy can move.

~2/3Use or would use agentsTo save time or find better prices.
~9/10Want transparencyInto how agent decisions are made.
11%Would delegate purchasesIn Gartner's surveyed low-stakes categories.

Survey designs and populations differ; figures are directional and should not be combined into a single adoption forecast.

The delegation gradient

  1. 1Summarize
  2. 2Narrow
  3. 3Recommend
  4. 4Assemble
  5. 5Confirm
  6. 6Automatenarrow categories

Trust-building design

01Accuracy
Show source freshness, confidence, and what changed before purchase.
02Transparency
Disclose decision criteria, constraints, sponsorship, and material trade-offs.
03Control
Use risk-based thresholds, pre-approved budgets, and clear pause or override controls.
04Privacy
Minimize data sharing, separate contexts, and let users inspect or revoke access.
05Authorization
Bind approval to merchant, cart, amount, timing, and policy with an audit trail.
06Recourse
Carry the original intent, evidence, and responsible party into returns, refunds, and disputes.
07Identity and loyalty
Recognize memberships and entitlements without unnecessary disclosure.
08Preference fidelity
Let users distinguish hard constraints from soft preferences and teach by correction.

Design principle

The future is not zero confirmation; it is better delegation. Portable mandates and risk-based limits reduce unnecessary prompts. Repeat usage will depend on clear aftercare, preserved evidence, and an easy path to human recourse.

05 / The solution

Aisle is the merchant control plane.

Aisle fills the gap between fast-moving agent ecosystems and the operational reality of a live storefront.

Category definition

Aisle is not another checkout protocol. It is the layer that makes every protocol, browser path, and agent channel operationally reliable for the merchant.

The Aisle control loop

  1. 01DiscoverMap storefront, catalog, policies, and access.
  2. 02SimulateRun repeatable browser and protocol journeys.
  3. 03DiagnoseCapture evidence, classify failures, find root cause.
  4. 04RemediatePrioritize fixes, verify changes, preserve rollback.
  5. 05ObserveTrack live agents, conversion, fraud, and attribution.
  6. 06OptimizeImprove win rate, offers, loyalty, and lifetime value.
  7. 07CertifyReadiness score, conformance tests, partner-facing evidence.

How Aisle closes the loop

01Launch: map and score
Connect a storefront or URL, choose representative tasks, and receive a decomposable readiness grade with evidence for every finding.
02Simulate and monitor
Repeat cross-agent journeys as the store, agent surfaces, policies, and protocols change; catch regressions before customers do.
03Fix and verify
Turn root cause into prioritized guidance or safe remediation, then rerun the exact journey to prove the issue is resolved.
04Optimize and certify
Link technical behavior to conversion, risk, support, loyalty, and revenue; publish a trusted readiness signal for partners.

Why the position is defensible

Not just a scan; a longitudinal outcome dataset. Aisle can link agent behavior, storefront state, failure type, remediation, and commercial result. Over time, that corpus can power benchmarks, predictive issue detection, automated fixes, fraud insights, and a trusted readiness score.

06 / The wedge

Land with readiness. Expand into the system of record.

The launch wedge makes an abstract problem concrete; the long-term product owns agent-channel reliability and economics.

A staged merchant roadmap

01ReadinessKnow how agent-ready the storefront is.Scan, map, simulate, prioritize.
02ReliabilityKeep agent journeys working as the store changes.Monitoring, regression tests, alerts.
03RemediationFix the highest-value issues quickly and safely.Platform workflows, verification, rollback.
04EconomicsOptimize agent traffic, checkout, and risk.Attribution, conversion, fraud insight.
05OrchestrationMake loyalty, offers, pricing, and CRM agent-native.Partner APIs, certification, actions.

The launch wedge

A merchant-facing Agent Readiness Score.

One score summarizes five dimensions (discoverability, product truth, transaction reliability, trust and security, and observability) while always decomposing into coverage, evidence, confidence, freshness, and recommended action.

Free diagnostic → recurring monitoring → remediation → analytics and enterprise control.

Why this can compound

  • A cross-agent, commerce-specific workflow, not a single-standard validator.
  • Longitudinal data linking failure, fix, and commercial outcome.
  • An open readiness framework with a proprietary outcome corpus.
  • Platform-first distribution, agency portfolios, and a certification network.

North-star metric

Agent task success rate on revenue-relevant journeys, segmented by agent, intent, channel, and storefront version.

The 12-month test

Can Aisle prove a failing journey, fix it, and measure the commercial improvement?

That is the category test. If yes, protocols make agentic commerce possible, and Aisle makes it observable, reliable, and economically valuable for merchants.

The dispatches

Follow the buildout.

Occasional letters on agentic commerce: readiness patterns from real scans, what actually breaks agents on storefronts, and what we ship next. Joining also holds your place in line for an Aisle invite.

No spam. Unsubscribe any time.

Selected sources

[1] McKinsey & Company, The agentic commerce opportunity, Oct. 2025.

[2] OpenAI, Instant Checkout and the Agentic Commerce Protocol, Sept. 2025.

[4] Google Developers, Universal Commerce Protocol, Jan. 2026.

[5] Google Developers, Guide to AI Agent Protocols / AP2, Mar. 2026.

[6] Shopify Help Center, Shopify agentic storefronts, accessed July 2026.

[9] Visa, Trusted Agent Protocol, Oct. 2025.

[10] Cloudflare, Introducing the Agent Readiness score, Apr. 2026.

[12] Mastercard, Verifiable Intent in agentic commerce, Mar. 2026.

[13] Visa, Earning Consumer Trust in the Age of Agentic Commerce, 2026.

[14] Gartner, Consumers Want AI Shopping Help, May 2026.

[18] Deloitte, Agentic Commerce in CPG: Winning the Algorithmic Shelf, June 2026.

[22] Aisle internal product roadmap and go-to-market priorities, July 2026.

Research note: consumer and merchant surveys reflect different populations, markets, dates, and question wording. Forward-looking sections represent strategic interpretation of current infrastructure and Aisle plans.

The Missing Merchant Control Plane · Aisle