Top agentic AI use cases for e-commerce and D2C in H2 2026

Product feeds ready for checkout in Google AI Mode and ChatGPT

In January 2026 Google launched the Universal Commerce Protocol, built with Shopify, Etsy, Wayfair, Target and Walmart, to power checkout from eligible US retailers in AI Mode and the Gemini app, with the retailer staying seller of record. OpenAI launched Instant Checkout in ChatGPT in September 2025 on the Agentic Commerce Protocol it built with Stripe, then in March 2026 let merchants use their own checkout while it focuses on product discovery. MIRA with DRIS gets the catalogue ready for both.

the workflow, in one room

1. ai teammatespeople
2. TriggerEach month, audit the catalogue and the Merchant Center feed for variant-level price, stock, shipping, return policy and GTIN fields.
3. Flag variants where the feed, the PDP and checkout disagree, since an agent buying for a shopper acts on the data it is given.
4. Check eligibility for checkout in Google AI Mode and the Gemini app, and for ChatGPT product discovery, market by market.
5. Draft the data fixes and a channel plan stating what each programme needs from the store platform.
6. Approved by Head of ecommerceThe head of ecommerce approves the data fixes and the agent channels the brand joins.
7. Apply the fixes, enrol in the approved programmes and monitor orders and errors arriving from agent channels.

Guardrail. Give agents the same price, delivery and return terms the site shows, with no charge added at the end; drip pricing has been banned in the UK since 6 April 2025.

Amazon listings that answer Alexa for Shopping, formerly Rufus

Amazon renamed its Rufus shopping assistant Alexa for Shopping on 13 May 2026, and it answers shoppers from listing details, customer reviews and community Q&A as well as information from across the web. FOLIO finds the shopper questions a listing fails to answer and drafts the missing facts for the brand’s own listing content.

the workflow, in one room

1. ai teammatespeople
2. TriggerEach week, put a fixed set of shopper questions per product to Alexa for Shopping in each marketplace where it is offered.
3. Record answers that are vague, wrong or drawn only from reviews and community questions rather than the listing.
4. Trace each gap to a missing or unclear fact in the title, bullets, A+ content or specifications.
5. Draft the missing facts from the approved specification sheet, with no claim the product cannot support.
6. Approved by Amazon channel managerThe Amazon channel manager approves each listing change before it is submitted in Seller Central.
7. Submit, put the same questions again after the listing updates and log which answers changed.

Guardrail. Every new listing fact must match the product and its spec sheet, and the brand never plants community questions, answers or reviews to steer an assistant that draws on all three.

Accurate brand answers in ChatGPT, Gemini and Google AI Mode

Pew Research Center found Google users clicked a traditional result in 8% of visits when an AI summary appeared, against 15% when none did, so what AI engines say about a brand’s products, prices and policies now shapes the shortlist before any visit. DRIS runs a fixed set of shopper prompts each week, finds wrong or missing answers and fixes the sources the brand controls.

the workflow, in one room

1. ai teammatespeople
2. TriggerEach week, run the agreed category and product prompts across ChatGPT, Gemini, Perplexity and Google AI Mode.
3. Record which brands are named, which pages are cited and any wrong price, policy or product fact.
4. Trace each wrong answer to its likely source, such as an old help page, a marketplace listing or a third-party review site.
5. Draft corrections to brand-owned pages and structured data, and clearer buying guides where the brand is missing.
6. Approved by SEO leadThe SEO lead approves each change and its product claims before it is published.
7. Publish, run the affected prompts again and log whether the answers and citations changed.

Guardrail. Fix the source, never the answer: no planted reviews, seeded forum posts or claims the product cannot support to sway what the engines say.

A site shopping assistant that says it is AI and knows the stock

Shoppers now expect to chat with a brand before buying: in January 2026 Google launched Business Agent, a branded chat in Search that eligible US retailers set up in Merchant Center. From 2 August 2026, Article 50 of the EU AI Act requires that people are told when they are talking to an AI system unless it is obvious. Convexa answers product, sizing and delivery questions on site from approved content, says it is AI and hands over to a person on request.

the workflow, in one room

1. ai teammatespeople
2. TriggerA shopper opens chat on a PDP, in the cart or on WhatsApp.
3. Disclose in the first message that the assistant is AI, then answer only from approved product, delivery and returns content and live stock data.
4. Recommend products from the catalogue against the shopper’s stated needs, and never invent a discount.
5. Hand over to a person on request, or when the question concerns health, a complaint or a refund.
6. Approved by Head of digitalThe head of digital approves the answer library, recommendation rules and handover triggers before any change goes live.
7. Log transcripts to the help desk, and share the approved answers so the brand’s Business Agent in Merchant Center says the same.

Guardrail. Disclose AI at first contact as EU AI Act Article 50 requires from 2 August 2026, and never let the assistant promise a price, delivery date or health benefit outside the approved library.

Drip pricing and prior-price audit for UK and India sale pages

The UK DMCC Act has banned drip pricing since 6 April 2025: headline prices must include every fixed mandatory charge, with penalties of up to 10% of global turnover. India’s Consumer Protection (E-Commerce) (Amendment) Rules, 2026, notified on 9 September 2026 and in force from 1 January 2027, require a reduced price to be shown with the prior price, the lowest of the preceding 30 days, and a yearly dark patterns self-audit. FOLIO audits every price display against both.

the workflow, in one room

1. ai teammatespeople
2. TriggerBefore each sale and every quarter, crawl collection pages, PDPs, the cart and every checkout step on each storefront.
3. Flag mandatory charges that appear only late in checkout, and headline prices that leave out fixed fees.
4. For India, compare each struck-through price with the lowest price of the preceding 30 days.
5. Flag false countdowns, pre-ticked add-ons and other dark patterns, with dated screenshots.
6. Approved by Head of ecommerce and Legal counselThe head of ecommerce and legal counsel approve the fixes and the self-audit record.
7. Apply the fixes, crawl again to confirm them and file the dated audit for the yearly compliance certificate.

Guardrail. In the UK, include every fixed mandatory charge in the headline price; in India, from 1 January 2027, show the prior price beside any reduction and keep the yearly dark patterns self-audit.

AI-generated product images and copy labelled for Merchant Center

Merchant Center requires images made with generative AI to carry IPTC metadata marking them as AI-generated, and AI-written titles and descriptions to be submitted in the structured title and structured description attributes. From 2 August 2026 the EU AI Act also requires deployers to disclose deep fakes. RUPA generates lifestyle imagery and CREA writes copy with those labels kept intact.

the workflow, in one room

1. ai teammatespeople
2. TriggerA new range needs lifestyle images or product copy at catalogue scale.
3. Generate images from approved product photography, keeping the IPTC digital source type metadata that marks them as AI-generated.
4. Check each image against the real product for colour, size, materials and included parts.
5. Draft titles and descriptions, and submit AI-written text in structured_title and structured_description with the trained_algorithmic_media value.
6. Approved by Creative leadThe creative lead approves each image and text before it is added to the feed or the site.
7. Publish, and audit the feed monthly for AI-generated assets missing their labels.

Guardrail. Never strip the AI-generated metadata or show a product that differs from the real one; Merchant Center requires the labels, and EU AI Act Article 50 requires deep fakes to be disclosed from 2 August 2026.

Consent ledger for email, SMS and WhatsApp ahead of India’s DPDP Rules

MeitY notified the Digital Personal Data Protection Rules on 13 November 2025: consent manager rules apply from 13 November 2026 and most notice and consent duties from 13 May 2027. CORA builds one record per customer of what they agreed to, on which channel and when, and keeps the email platform, SMS tool, WhatsApp and ad audiences in step with it.

the workflow, in one room

1. ai teammatespeople
2. TriggerEach night, pull consent and opt-out records from checkout, sign-up forms, the email and SMS platforms and WhatsApp opt-ins.
3. Reconcile them into one ledger per customer with source, purpose, channel, notice version and date.
4. Flag customers messaged on a channel or for a purpose they never agreed to, and records past their retention period.
5. Draft suppressions, erasures and plain-language notice wording for every sign-up point.
6. Approved by Data protection leadThe data protection lead approves suppressions, erasures and notice wording before they are applied.
7. Push the approved changes to every connected tool and to ad audience uploads.

Guardrail. Treat the DPDP Rules 2025 dates as hard deadlines: consent manager rules from 13 November 2026, and notice and consent duties from 13 May 2027.

WhatsApp marketing re-planned for per-message pricing

Meta moved the WhatsApp Business Platform to per-message pricing on 1 July 2025 and updated its pricing again from 1 October 2026, so every marketing message now carries its own cost. LIPI plans WhatsApp sends by expected value per message, moves low-value broadcasts to email and drafts the templates for approval.

the workflow, in one room

1. ai teammatespeople
2. TriggerEach week, pull WhatsApp results by template: delivered, read, clicked and ordered, with the cost of each message.
3. Compare cost per order on WhatsApp with email and SMS for the same kind of message.
4. Move low-value broadcasts to cheaper channels and keep WhatsApp for messages customers act on.
5. Draft next week’s templates and audience sizes, limited to customers who opted in to WhatsApp.
6. Approved by CRM leadThe CRM lead approves the templates, audiences and weekly messaging budget before sends are scheduled.
7. Submit templates to Meta for approval, schedule the sends and honour every opt-out at once.

Guardrail. Message only customers who opted in and honour every opt-out, as WhatsApp’s Business Messaging Policy requires, and budget from Meta’s current rate card rather than last year’s.

Subscription pause and cancellation handled without friction

The Eighth Circuit vacated the FTC’s click-to-cancel rule on 8 July 2025, but ROSCA and state automatic renewal laws, such as California’s and New York’s, still govern subscriptions in the US. Convexa handles subscribe-and-save cancellations in chat, offers a pre-approved skip, pause or swap once, and cancels straight away when the customer still wants to leave.

the workflow, in one room

1. ai teammatespeople
2. TriggerA subscriber asks to cancel or change a subscription in chat, by email or from the account page.
3. Ask one question about the reason and check the subscription, order history and next charge date.
4. Offer the matching pre-approved option once: skip, pause, change frequency or swap the product.
5. Approved by Subscription leadThe subscription lead approves the save options, their wording and any change to the cancellation flow before it goes live.
6. If the subscriber still wants to leave, cancel at once in the subscription app and confirm in writing.
7. Log the reason and outcome, and send monthly cancellation reasons to product and retention.

Guardrail. Never obstruct or delay a cancellation: ROSCA and state automatic renewal laws still apply in the US after the FTC’s rule was vacated, and the same standard protects customer trust in every market.

Frequently asked questions