The definitive guide to agentic commerce for ecommerce retailers

AI shopping agents are changing how consumers discover, compare, and buy. Here is everything you need to know, and how to make sure your delivery operations are ready for the agents that will decide whether shoppers choose your store.

Published: April 2026

Agentic commerce architecture and AI shopping agent delivery network

What is agentic commerce?

Understand the shift from traditional ecommerce to AI-driven autonomous shopping, and why your delivery operations are now a competitive variable.

Read more →

How agents choose delivery

AI shopping agents evaluate carrier breadth, delivery accuracy, returns policy, and checkout API compatibility. Here is what that means for your store.

Read more →

Make your store agent-ready

nShift connects your store to 1,000+ carriers, automates delivery promises, and gives agents the structured data they need to recommend and complete purchases on your behalf.

Read more →

What is agentic commerce?

Agentic commerce is the next evolution of online retail, in which AI-powered software agents search, compare, and complete purchases autonomously on behalf of consumers, without requiring manual input at each step.

In traditional ecommerce, a shopper opens a browser, searches for a product, compares options, selects a delivery method, and completes checkout. Each step requires human time and attention.

In agentic commerce, AI software agents handle the entire journey. A consumer sets a goal: "find me the best running shoes under $150 with next-day delivery and an easy returns policy", and the agent takes over. It queries product databases and inventory APIs, evaluates delivery options in real time, checks returns terms, and completes the purchase. The shopper never opens a checkout page.

This is happening now. Google, Perplexity, and a growing ecosystem of third-party agents are building agentic purchase capabilities. 38% of consumers already use AI when shopping, and 80% expect to use it more. Shopping-related generative AI searches grew 4,700% between July 2024 and July 2025.

For ecommerce retailers, agentic commerce changes the fundamental rules of competition. Today you compete on search ranking, homepage design, and ad spend. Tomorrow you also compete on machine readability: whether an AI agent can access your delivery options, evaluate them accurately, and trust your fulfilment capabilities enough to recommend your store over a competitor.

Delivery is no longer just the last step in the purchase journey. For AI shopping agents, it is a ranking criterion applied before the shopper ever sees your store.

Agent-readiness has four components: carrier breadth, delivery accuracy, returns automation, and checkout API compatibility. In this guide, we'll explore how to address all four through a single, modular integration.

Retailers who build agent-ready delivery infrastructure now gain a durable advantage: not just with AI agents, but across every dimension of delivery performance, from checkout conversion to post-purchase retention.

Get tailored advice on agentic commerce and your delivery operations:
Agentic commerce tracker

The year AI started shopping

A live view of the milestones, adoption signals, and market forces shaping agentic commerce in 2026, and what they mean for delivery.

45%
of consumers already use AI in their buying journey (IBM, Jan 2026)
75%
of surveyed retailers implementing or planning agentic commerce (NRF)
$3-5T
in global retail spend agentic commerce could redirect by 2030 (McKinsey)
~50%
of online shoppers projected to use AI agents by 2030 (Morgan Stanley)

2026 milestones

Click any milestone to expand. Scroll to see more.

January 2026
IBM: 45% of consumers use AI to buy
Research
IBM's Institute for Business Value reports that nearly half of consumers now use AI in at least part of their purchasing journey, up sharply from 2024 baselines. The finding signals that agent-assisted shopping has crossed from early-adopter into mainstream territory.
January 2026
Google launches Universal Commerce Protocol at NRF
Protocol
Google announces UCP as an open standard for agentic commerce covering discovery, buying, and post-purchase support. Initial launch includes US Etsy sellers, with 1M+ Shopify merchants (Glossier, SKIMS, Spanx, Vuori) announced as incoming. A direct attempt to set the infrastructure layer.
January 2026
75% of NRF retailers planning agentic initiatives
Industry
A survey at the National Retail Federation conference finds three-quarters of retailers are either actively implementing or formally planning agentic commerce projects. The remaining 25% risk falling behind as agent-ready competitors capture AI-routed demand.
February 2026
Microsoft ships Copilot Checkout
Platform
Microsoft launches Copilot Checkout, enabling US Copilot users to purchase from Etsy, Urban Outfitters, and Anthropologie without leaving the chat interface. This brings agentic purchasing to the Microsoft 365 ecosystem and its hundreds of millions of users.
March 2026
OpenAI and Shopify refine checkout flows
Platform
OpenAI communicates it is rethinking its strategy for checkout in agentic experiences. Simultaneously, Shopify details its approach for using ChatGPT within the flow toward checkout for merchants. The checkout layer is being actively redesigned around agent workflows.
March 2026
Regulatory gap flagged for AI purchasing agents
Regulation
The Center for Data Innovation highlights that consumer-protection frameworks were written for humans, not autonomous purchasing agents. Open questions around liability, consent, and dispute resolution remain unresolved, creating uncertainty for retailers scaling agent-based transactions.
March 2026
Walmart evolves its agentic commerce position
Retail
Walmart publicly details how its view of agentic commerce is evolving, signaling that the world's largest retailer is actively adapting operations and technology to prepare for AI-driven purchasing at scale.
2026 ongoing
Multi-item cart support becomes standard
Infrastructure
Agent platforms evolve from single-product transactions to full multi-item cart support. When a user tells an agent to "order everything I need for taco night," the agent now builds a complete cart and executes as a single transaction. This raises the bar for structured product and delivery data.

Trend data

Adoption curves and market projections shaping the agentic commerce trajectory.

Consumer AI adoption in shopping
% of consumers using AI in their buying journey. IBM IBV (Jan 2026), Morgan Stanley projections
Bar chart showing year-over-year growth in consumer AI use during shopping, rising from 12% in 2023 to 50% projected by 2030.
Data points, estimated or projected consumer AI adoption in shopping: 2023, 12%. 2024, 22%. 2025, 34%. January 2026, 45%. 2028 projection, 48%. 2030 projection, 50%.
Retailer readiness
NRF 2026 survey: agentic commerce initiative status
Doughnut chart showing retailer readiness for agentic commerce, with 75% implementing or planning initiatives and 25% not yet active.
Retailer readiness breakdown: 75% implementing or planning agentic commerce initiatives. 25% not yet active. Center highlight: 75% already acting on agentic commerce.
Projected global retail spend redirected by agentic commerce
McKinsey estimate, $ trillions, conservative to upper range
Area chart showing projected global retail spend redirected by agentic commerce, growing from 0.05 to 0.1 trillion dollars in 2024 to 3 to 5 trillion dollars by 2030.
Projected global retail spend redirected by agentic commerce in trillions of dollars. Upper range: 2024, 0.1. 2025, 0.4. 2026, 0.9. 2027, 1.6. 2028, 2.6. 2029, 3.7. 2030, 5.0. Conservative range: 2024, 0.05. 2025, 0.2. 2026, 0.5. 2027, 1.0. 2028, 1.6. 2029, 2.2. 2030, 3.0.

4,700%

growth in shopping-related generative AI searches (2025)

- eMarketer

38%

of consumers already use AI when shopping


- Salesforce

80%

expect to increase their use of AI for purchases


- Salesforce

57%

of ecommerce businesses exploring AI agent use cases

- Pattern Group

How AI shopping agents choose delivery

Summary

Summary

When a human shopper selects a delivery option, they weigh speed, cost, and convenience based on what they can see on screen. When an AI agent makes the same decision, it queries APIs, processes structured data, and applies a ruleset optimised for the consumer's stated preferences. The logic is consistent, instantaneous, and scale-free. Understanding what agents evaluate is the first step to building a delivery operation they will recommend.

Real-time delivery accuracy

Agents do not guess at delivery windows. They query your carrier APIs and expect machine-readable, accurate, real-time data. A vague "3–5 business days" response is a disqualifying signal. Agents need exact delivery date commitments tied to current carrier availability, origin warehouse location, and service-specific cut-off times. A delivery management platform that maintains live carrier data and exposes it through structured APIs is a prerequisite.

Carrier breadth

Agents optimize for each individual consumer's preferences, which vary significantly: same-day delivery in their city, a carbon-neutral option, a specific carrier they trust, or simply the lowest available cost. A retailer operating with two or three carrier connections will be filtered out of a large share of agent evaluations. Breadth, meaning access to carriers covering every delivery preference, is now a competitive necessity, not a premium feature.

Returns policy as a ranking signal

Returns ease is increasingly a binary filter in agent decision-making. McKinsey's research on agentic commerce identifies clear, automated, low-friction returns as a measurable signal of retailer reliability. An agent advising a consumer on a discretionary purchase will weight returns policy heavily, because a poor returns experience damages the consumer's trust in the agent as much as in the retailer.

Checkout API compatibility

Agents interact with your checkout via APIs, not via your website's visual interface. OpenAI's agentic commerce checkout specification uses five REST endpoints (create session, update session, get state, complete purchase, and cancel) with all responses returning JSON including line items, totals, and structured delivery options. If your checkout does not expose this data in a machine-readable format, agents cannot complete purchases through your store, regardless of how strong your product offering is.

Post-purchase visibility

Agentic commerce does not stop at checkout. After purchase, agents also need reliable, machine-readable shipment status and delivery events, which makes post-purchase visibility a critical part of the agent-ready stack.

Sustainability signals

Eco-conscious delivery is increasingly weighted in agent decisions, particularly in European markets where consumer preferences and regulatory pressure point in the same direction. Agents can filter for lower-emission delivery options, and retailers who surface carbon data through structured APIs have a measurable competitive advantage. The nShift platform surfaces sustainability data alongside carrier options, enabling both agents and consumers to make informed choices.

How an agentic commerce transaction works

From the moment a consumer sets a goal to the post-purchase tracking loop, here is what happens when an AI agent shops on someone's behalf.

  1. Step 1
    Consumer activates the agent

    A shopper sets a shopping goal: product type, budget ceiling, preferred delivery speed, and returns requirements.

  2. Step 2
    Agent queries product and delivery APIs

    The agent queries compatible retailers simultaneously, pulling structured inventory, pricing, and carrier data in real time.

  3. Step 3
    Agent evaluates and ranks options

    Delivery speed, carrier, cost, returns policy, and sustainability score are weighted against the consumer's stated preferences.

  4. Step 4
    Agent selects the winning retailer

    The retailer with the best combined score is selected. If your delivery data is inaccessible, inaccurate, or too narrow, your store is not in the ranking.

  5. Step 5
    Agent completes the purchase via checkout API

    Payment and order confirmation are processed automatically. The consumer receives confirmation without visiting your website.

  6. Step 6
    Post-purchase tracking loop begins

    The agent monitors delivery progress through tracking APIs and alerts the consumer to milestones and any delays, building trust in both the retailer and the agent for the next purchase.

The protocol layer:
What UCP means for delivery

The Universal Commerce Protocol (UCP), co-developed by Shopify and Google, standardizes how AI agents interact with merchant infrastructure. It defines a manifest that declares what a merchant's systems support: checkout sessions, order lifecycle events, identity linking, and payment token exchange.

For delivery, the implication is direct. UCP's checkout capability generates sessions that include fulfillment options. If your delivery layer cannot serve structured carrier data, estimated delivery times, and sustainability scores through an API, your store is invisible to UCP-compliant agents. The protocol does not replace your delivery management platform. It creates the standard interface that your platform needs to feed.

Universal Commerce Protocol (UCP) diagram showing how AI agents query checkouts and delivery endpoints
logo-millesima-us-blanc

“A driving force for our growth”

Millesima used nShift to quickly connect to carriers across Europe, avoiding the time and cost of setting up each carrier individually and accelerating its expansion into new markets. 

Read how they did it

The delivery gap:
Why most retailers are not agent-ready today

Despite speed of adoption, most ecommerce retailers are not operationally prepared for agentic commerce. A 2025 study by Pattern Group found that 57% of ecommerce businesses are exploring AI agent use cases, but only 33% are actively preparing for deployment. The gap between awareness and readiness is still wide and poses a significant advantage for early adopters.

Gap #1

Delivery data quality

AI agents abandon carts if delivery data is unstructured.

Retailers must move from static text estimates to machine-readable data schemas. Agents require exact, carrier-specific delivery commitments generated from live service rules, warehouse cut-off times, and current availability.

A delivery management platform maintains this live carrier data to ensure the accuracy required for automated checkout.

Gap #2

Carrier network limitations

AI shopping agents require a broad carrier mix to satisfy diverse delivery goals, including same-day, sustainable, and cross-border options.

Most mid-size retailers lack this breadth because they rely on a small number of direct carrier integrations. Building these connections individually is slow and expensive, as each requires a separate contract, technical integration, and ongoing maintenance.

A unified delivery network provides immediate access to the carrier breadth that agents demand.

Gap #3

Checkout architecture

Legacy checkouts built exclusively for human visual scrolling block AI agents from completing purchases.

An API-first checkout exposes structured delivery data directly to the agent, including accurate dates, carrier options, and sustainability scores.

This architecture is a foundational requirement for agentic commerce and simultaneously improves the standard human conversion rate by reducing friction.

Gap #4

Returns infrastructure

AI agents bypass retailers whose return policies are hidden in unstructured text.

Returns data must be structured, machine-readable, and accessible via API for an agent to confidently represent the terms to the consumer. While humans manually scan the fine print, agents require immediate data verification.

Automated, API-accessible returns handling turns this basic requirement into a measurable competitive advantage.

 

Agent-readiness maturity curve

The five stages of agent-ready delivery

As carrier choice, delivery promises, tracking, and returns become more structured and connected, your store becomes more visible to shopping agents.

Maturity model chart showing the five stages of agent-ready delivery, from Reactive to Agent-preferred.

As delivery maturity moves from Reactive to Agent-preferred, visibility to shopping agents increases. The model progresses through Reactive, Connected, Structured, Agent-ready, and Agent-preferred.

Visibility to shopping agents Delivery maturity
  1. Stage 1

    Reactive

    Delivery works, but the signals agents need remain fragmented.

    • Limited carrier choice
    • Manual or inconsistent returns

    Agents see: Low confidence.

  2. Stage 2

    Connected

    Carrier connectivity and automation are improving, but delivery is still uneven across channels.

    • More delivery options, including PUDO
    • Early tracking visibility

    Agents see: Usable signals.

  3. Stage 3

    Structured

    Delivery promises, tracking, and returns become more consistent and machine-readable.

    • More accurate delivery dates
    • Unified tracking signals

    Agents see: Trusted signals.

  4. Stage 4

    Agent-ready

    Delivery operations are connected well enough for agents to evaluate and complete purchases.

    • Structured checkout API responses
    • Clearer returns logic

    Agents see: Purchase confidence.

  5. Stage 5

    Agent-preferred

    Delivery becomes a competitive edge in agent-led retail.

    • Margin-aware delivery rules
    • Sustainability signals surfaced

    Agents see: Ranking edge.

The API delivery infrastructure
that makes retailers agent-ready

nShift is the global leader in delivery and experience management. The nShift platform connects more than 22,000 retailers, brands, manufacturers, and logistics providers to over 1,000 carriers across 190 countries, supporting nearly one billion annual shipments. It is the delivery infrastructure that AI shopping agents need to find beneath your store.
Checkout

Checkout: the agent-compatible delivery layer

nShift Checkout surfaces delivery options at the moment an AI agent evaluates your store. Its Smart Rules Engine takes basket data (product type, order value, destination) and dynamically presents carrier options, real delivery dates, and sustainability information in a structured, machine-readable format. A/B testing of delivery configurations lets retailers continuously improve both human conversion rates and agent compatibility scores.

Under the hood, the Checkout API operates as a REST interface returning structured JSON. When an agent sends a request with cart contents and a destination postcode, the API responds with carrier options, real-time estimated delivery times, pricing, and available pick-up and drop-off locations, each tied to a unique connection ID linked to the retailer's configuration. Retailers using nShift do not need to build a separate agent-facing checkout layer. The infrastructure already supports programmatic access.

Deliver: intelligent carrier selection at scale

nShift connects your operations to 1,000+ carrier services across 190+ countries through a single platform. Carrier selection, label generation, customs documentation, and return label printing are automated across the suite, covering nShift Ship for high-volume multi-carrier shipping, nShift Delivery for businesses that need centralized booking and labeling across multiple carriers, nShift TMS for transport planning and route optimization, Transsmart for ERP-connected warehouse and manufacturing flows, and nShift Go for API-first platforms that need checkout and tracking built in.

For agentic commerce, the critical capability is always-current delivery data. nShift maintains live connections to its full carrier library, so the options exposed through your checkout API reflect real-time availability, pricing, and service levels. When carriers update their rules or service offerings, nShift applies the changes automatically. Your team never manages carrier integrations manually.

Track: visibility that agents can query

From the moment a parcel is dispatched, nShift Track consolidates carrier event data into a single, structured tracking feed. Consumers receive proactive notifications at key milestones (dispatch, out for delivery, delivered, exception), and the same structured data is accessible to AI agents monitoring fulfilment on a consumer's behalf.

The feed normalizes shipment events across 1,000+ carriers into a single, consistent status model. An agent monitoring delivery on behalf of a buyer can query it for live progress updates and exception alerts without polling a branded tracking page. This is the signal that allows agents to proactively notify consumers of delays or confirm successful delivery, closing the loop on the transaction they initiated.

Returns: the signal agents trust

nShift Returns powers branded, self-service return experiences that give consumers confidence before they buy and give AI agents a structured data point to factor into their retailer ranking. The portal integrates directly with nShift Checkout, automates refund processing, and provides analytics to identify the return reasons that matter most.

For AI agents, this goes beyond the portal experience. The Returns API exposes return conditions through both a Pull endpoint (for querying item eligibility, refund rules, and available exchange options by product category) and a Push endpoint (for triggering return workflows from external systems). The returns widget URL can also be embedded in structured product data, giving agents a machine-readable entry point to confirm the policy without parsing a web page.

Icon representing business rules that prevent AI agents from exploiting expensive carrier options

Maintaining margin control when AI does the shopping

When an AI agent completes a purchase on behalf of a consumer, it optimizes for the buyer, not the retailer. That creates three specific margin risks that most fulfillment setups are not built to handle.

The nShift Rule Engine addresses each of these by filtering what the Checkout API exposes before an agent ever sees the options. Operations teams set conditions based on cart value, product category, parcel weight, and destination. Orders below the free shipping threshold do not receive free tiers. Heavy or oversized items are routed to carriers with appropriate rate cards. Premium add-ons can be surfaced as default selections rather than optional upgrades. The agent still optimizes for the buyer, but only within a set of options the retailer has already validated for profitability.

Cost-blind carrier selection

An agent comparing delivery options will gravitate toward the cheapest or fastest choice. If a flat-rate shipping tier is available, the agent will select it regardless of parcel weight, dimensions, or destination surcharges. A heavy, bulky order shipped at a subsidized flat rate can turn a profitable sale into a loss.

Free shipping threshold exploitation

Many retailers offer free shipping above a cart value. An AI agent acting on a price-sensitive brief will bundle items to cross that threshold, then select the most expensive carrier option available at zero cost to the buyer. The retailer absorbs the full fulfilment expense.

Loss of delivery upsell

Human shoppers can be nudged toward premium options (gift wrapping, named-day delivery, sustainable shipping) through visual merchandising in the checkout. An agent skips that entirely. Without a mechanism to surface margin-positive options to automated buyers, those revenue streams disappear.

Retailers already building agent-ready delivery

Millesima: fast carrier expansion across Europe

Millesima, the wine retailer, needed to connect quickly to multiple carriers across European markets. Building each integration independently would have cost significantly in development time and operational overhead. By connecting through nShift, Millesima gained access to the carriers it needed through a single integration, reducing time-to-market for each new market and enabling the carrier breadth that agent-era competition demands.

Read their story
millesima-nshift-ecommerce-outcomes

Imerco: scaling delivery through peak demand

Imerco, Denmark's leading kitchenware and home accessories retailer, ships high-value, fragile products to 165 stores and direct to consumers. During peak season, orders increase by as much as 240%. With nShift Ship integrated into its AutoStore-powered robotic warehouse, Imerco scaled from their annual outbound shipments by 14% without operational disruption. 

Read how they did it
Imerco-nShift-customer-outcomes-ecommerce

Connecting you to every carrier

Connect to 1,000+ carriers across 190+ countries and 70+ PUDO networks (1.2M+ locations). No bespoke per-carrier builds. 
Explore available carriers
nShift-carrier-network

The agent-readiness gap

AI shopping agents evaluate retailers across eight delivery capabilities before recommending a purchase: carrier breadth, delivery promise accuracy, returns experience, post-purchase tracking, checkout API compatibility, cross-border coverage, carrier update management, and sustainability data. Most retailers without a delivery management platform fall short on the majority of these criteria.

The table below compares the typical state of each capability without nShift and with nShift in place.

What agents evaluate

Without nShift

With nShift

Carrier network breadth

2–5 carriers, limited per market

1,000+ carrier services, 190 countries

Delivery promise accuracy

Estimated windows, not carrier-verified

Live, carrier-specific delivery dates via API

Returns experience

Manual or limited returns process

Automated, branded, self-service returns portal

Post-purchase tracking

Fragmented, carrier-by-carrier

Unified tracking feed across all carriers

Checkout API compatibility

Visual UI checkout, not machine-readable

Structured delivery data exposed via REST API

Cross-border coverage

Requires individual carrier contracts per market

190 countries, single platform integration

Carrier update management

Manual technical effort per carrier change

Centrally managed by nShift, zero effort for retailer

Sustainability data

Not surfaced at checkout

Carbon data available per delivery option

Want to make your delivery operations agent-ready?

Join more than 22,000 businesses using nShift to future-proof fulfilment for the era of agentic commerce. Our delivery experts will review your current carrier setup and show you exactly what it takes to become agent-ready.
Talk to an expert

Frequently asked questions

What is agentic commerce?

Agentic commerce is a model of online retail in which AI-powered software agents autonomously search for, evaluate, and purchase products on behalf of consumers, without requiring manual input at each step. The consumer sets a goal (product type, budget, delivery preferences, returns requirements), and the agent queries APIs, ranks options, and completes the transaction. 

How do AI shopping agents choose a delivery provider?

AI shopping agents evaluate delivery options based on structured, machine-readable data. They assess carrier availability, delivery speed, cost, accuracy of the delivery promise, returns policy, and sustainability score, all pulled from APIs in real time. Agents prioritise retailers whose delivery data is accurate, complete, and accessible via a structured checkout API. Retailers with narrow carrier options, imprecise delivery windows, or inaccessible checkout architectures are filtered out of agent recommendations.

What makes an ecommerce retailer agent-ready?

Agent-readiness has four components: carrier breadth (access to a wide range of carriers and services, covering every consumer preference), delivery accuracy (real-time, carrier-verified delivery promises), returns automation (a clear, structured, self-service returns process), and checkout API compatibility (a machine-readable checkout that agents can complete on the consumer's behalf). The nShift platform addresses all four through a single, modular integration.

How do I make my delivery options visible to AI shopping agents?

AI agents cannot visually scroll through shipping options. To make delivery choices visible, retailers must expose structured, machine-readable delivery commitments via an API-first checkout. This requires a delivery management platform that translates live carrier data (including exact dates, costs, and sustainability scores) into a format the agent can immediately query and evaluate.

Why do AI shopping agents abandon carts during delivery selection?

Agents abandon transactions when they encounter unstructured delivery data, such as vague '3-5 day' text estimates, or when a retailer lacks the specific carrier options the consumer's prompt requires (e.g., guaranteed next-day or sustainable delivery). If the agent cannot programmatically verify the delivery commitment against the user's constraints, it will abandon the cart and purchase from a competitor whose checkout data is accessible.

How do retailers protect profit margins when AI agents select delivery?

Retailers must enforce strict business rules within their delivery management platform before exposing options to an agent. By configuring cost-blind carrier selection rules, enforcing free shipping thresholds automatically, and mapping delivery zones accurately via API, retailers can ensure that an agent always selects the most cost-effective compliant option without compromising the retailer's margin.