Agentic commerce is shopping where an AI agent finds products, compares them and completes checkout on a person’s behalf. Protocols such as Google’s Universal Commerce Protocol (UCP) and OpenAI’s Agentic Commerce Protocol define how agents talk to stores. For a retailer, costs sit in product data quality, integration work and payment and fraud setup; ROI is measured as agent-driven orders, their margin and their effect on other channels.
What do agentic commerce and UCP mean in practice?
Instead of a shopper clicking through a site, an assistant such as ChatGPT, Gemini or Google’s AI Mode receives a request like “buy a waterproof hiking jacket under 150 euros in size M”. The agent searches product data from merchants, presents options and, with the user’s approval, places the order.
For that to work, stores need a standard way to expose products, prices, stock, shipping options and checkout to agents. OpenAI introduced the Agentic Commerce Protocol with Stripe to power Instant Checkout in ChatGPT. Google announced the Universal Commerce Protocol as an open standard for agent-driven checkout across its surfaces, with support from retail and payment partners. Platforms such as Shopify have been building support so merchants can take part without custom engineering. Details and availability vary by market and are still changing at the time of writing.
Where do the costs come from?
Product data
Agents rely on structured data more than people do. A human can guess from a photo that a jacket has a hood; an agent needs the attribute. Expect work on titles, attributes, variants, sizes, materials, shipping times and return policies, usually through the product feed used for Google Merchant Center and similar channels.
Integration
On hosted platforms, joining a programme may be a settings change plus app installs. On custom stacks, developers must implement the protocol endpoints for catalogue, cart, checkout and order updates, and test them against the specification.
Payments and fraud
Agent-placed orders use delegated or tokenised payments through providers such as Stripe, Google Pay or others supported by the protocol. Fraud rules may need tuning, since order patterns from agents differ from normal sessions.
Fees and operations
Some programmes charge merchant fees on agent-driven orders, and customer service must handle questions about orders a person did not place on your site directly. Check current fee terms before estimating.
How do you build a budget?
Separate one-off readiness costs from ongoing costs.
- One-off: feed and attribute cleanup, integration or app setup, fraud rule review, testing.
- Ongoing: programme fees per order where applicable, feed maintenance, monitoring and customer service time.
- Opportunity cost: developer time that could go to other site work.
For example, a mid-sized Shopify fashion store might find most of its effort is attribute cleanup across a few thousand variants, with integration handled by the platform. A retailer on a custom stack may face the opposite: clean data but weeks of engineering.
How do you measure ROI?
The first challenge is attribution. Orders placed through agents should be identifiable in your order system, usually by a sales channel or source tag the protocol provides. Make sure they are recorded separately from the start.
KPIs to track
- Number of agent-driven orders and revenue per week.
- Average order value and margin compared with web orders.
- Return rate for agent-driven orders, since mismatched expectations show up here.
- Share of new customers versus existing ones.
- Customer service contacts per agent order.
- Effect on other channels: are agent orders new demand or replacing web visits?
The last point matters most. If agent orders mostly replace orders that would have happened on your site, the gain is smaller than the revenue number suggests. Compare total sales trends, not just the agent channel.
What are the risks?
Agents compress the shopping journey. Brand storytelling, cross-selling and upsells on your site may never be seen. Price comparison becomes easier, which pressures margins on commodity products. Data errors also become more expensive: a wrong size chart may lead an agent to recommend the wrong product confidently.
There is also dependency. Once a share of sales comes through one assistant, its rules, fees and ranking logic matter a great deal. Keep direct channels healthy.
How Alien Road applies agentic commerce readiness for e-commerce clients
A typical engagement starts with a feed and catalogue audit. The team checks attribute completeness against what agent shopping surfaces need, compares Merchant Center diagnostics and reviews return and shipping policies for clarity.
Setup covers attribute cleanup, feed rules, joining the programmes available on the client’s platform and tagging agent-driven orders so they appear separately in reports. Where custom integration is needed, the team prepares the requirements and works alongside the client’s developers.
Measurement is a monthly report on agent order volume, margin, returns and overall sales trend, with commentary on whether the channel is adding demand or shifting it.
A short scenario: an outdoor gear retailer
Consider an outdoor gear retailer selling jackets, boots and tents. Its site copy is rich, but its feed has gaps: waterproof ratings sit only in description text, sizes are inconsistent across brands and shipping times are missing for some regions. A shopper asking an agent for “a waterproof jacket for hiking in rain, size M, delivered by Friday” would rarely see its products, because the agent cannot confirm any of those details.
Filling those attributes is the main readiness cost, and it pays off beyond agent channels, since the same feed powers Shopping ads and free listings. After enabling agent checkout, the retailer tags those orders separately and compares their margin and returns with web orders for three months before deciding how much more effort the channel deserves.
Where to start this week
Open Google Merchant Center and review the diagnostics for missing or invalid attributes on your top products. Fix the ones on your best sellers first. Then check if your e-commerce platform has announced support for agent checkout programmes in your market. Those two steps cost little and prepare you for whichever protocol your customers end up using. Add a source tag for agent orders as soon as one is available, so the first sale is already measurable.