A Brand Agent is an AI shopping agent on your own storefront that reasons over your catalog and policies — not a scripted FAQ bot.
- It is Microsoft's product, announced January 8, 2026, not a Shopify feature: Shopify merchants get it by installing Microsoft Clarity and joining Microsoft's waitlist, on any Shopify plan.
- It needs four inputs to work: a complete catalog, your brand voice, your policies, and your real shopper questions.
- The quality ceiling is set by catalog completeness, not by the model.
- Train on the unhappy questions — edge cases, constraints, comparisons — because that's where trust and conversions are won or lost.
When Microsoft announced Brand Agents on January 8, 2026 and opened a waitlist for Shopify stores, the first reaction from a lot of merchants was reasonable skepticism: isn't this just a chatbot with better marketing? It isn't — and the difference matters enough to be worth a clear explanation.
- Brand Agent
- Microsoft's AI shopping agent, built into Microsoft Clarity and installed on a merchant's own storefront — Shopify or WooCommerce — that reasons over the store's catalog, brand voice, and policies to answer questions, recommend products, and guide shoppers to a purchase in natural conversation. In beta as of Microsoft's July 2026 documentation.
Brand Agents is not part of Shopify's Spring '26 Edition and is not gated to Shopify Plus. Shopify's own agentic launch on June 17, 2026 was Shopify Catalog, the Universal Commerce Protocol, agentic storefronts, the Agentic plan and the Catalog, Cart and Checkout MCP servers. Copilot Checkout, announced alongside Brand Agents, is Microsoft's separate checkout inside Copilot.
How a Shopify merchant gets one
- 01
Install Microsoft Clarity
Brand Agents is built into Microsoft Clarity, so the Clarity app on your Shopify store is the entry point. Any Shopify plan works.
- 02
Join Microsoft's waitlist
Access is through Microsoft's waitlist while the product is in beta. Read the Brand Agents terms before you accept them, including what they ask you to publish to Microsoft Merchant Center.
- 03
Prepare the inputs before you switch it on
The agent is only as good as the catalog, voice, policies and questions you give it. The next sections cover those inputs.
Chatbot vs. Brand Agent
The old chatbot was a decision tree. You drew branches: if the shopper clicks "shipping", show the shipping text. It could only answer what you anticipated, in the order you anticipated it. A Brand Agent reasons. Ask it something you never scripted — "which of your two backpacks is better for a 15-inch laptop and a wet towel?" — and it works it out from your product data.
| Dimension | Chatbot | Brand Agent |
|---|---|---|
| Logic | Scripted decision tree | Reasoning over catalog + policy |
| Coverage | Only anticipated questions | Novel, constraint-heavy questions |
| Recommendations | Rule-based or none | Compares products against stated needs |
| Path to purchase | Hands off to cart page | Guides the shopper toward a purchase in the conversation |
| Failure mode | "I didn't understand that" | Wrong answer stated confidently (if catalog is thin) |
A chatbot fails safely — it says it doesn't understand. A Brand Agent on an incomplete catalog fails dangerously — it gives a confident, wrong answer. That's why catalog quality is non-negotiable before you go live.
What a Brand Agent needs to work
Four inputs, in roughly this order of impact:
- 01
A complete catalog
Every attribute a shopper might ask about, as a structured field — materials, dimensions, compatibility, use-cases, care. This is the floor and the ceiling. See the catalog playbook.
- 02
Your brand voice
How you talk. Warm and playful, or precise and technical. The agent should sound like your best salesperson on their best day — not like generic support.
- 03
Your policies
Returns, shipping, warranty, sizing, and the edge cases. The agent will be asked about all of them, and a wrong policy answer erodes trust fast.
- 04
Your real shopper questions
The 30–50 questions your category actually generates. Pull them from support tickets, reviews, and on-site search. These become your conversation flows.
Train on the unhappy questions
The instinct is to train on the easy path — "do you ship to Canada?" The high-value training is the opposite: the constraint-heavy, comparison-driven, edge-case questions where a human salesperson earns their keep.
- Comparisons — "what's the difference between these two?" Shoppers near a decision ask this constantly.
- Constraints — "vegan, under $40, ships by Friday." Multiple conditions at once; agents that handle these convert.
- Suitability — "is this good for sensitive skin / a 4-month-old / a cast-iron pan?" Category-specific trust questions.
- Objections — "why is this more expensive than the other one?" The agent should answer with value, not dodge.
Your conversion happens in the questions you didn't anticipate. Train for those, and the easy questions take care of themselves.
How to know it's working
Beyond the standard agentic metrics (conversation volume, conversion vs. site average, attributed revenue), two qualitative checks tell you whether the agent is genuinely good:
- Read 20 real transcripts a week. You'll spot off-brand phrasing and catalog gaps no dashboard surfaces.
- Track the decline rate. How often does the agent say it can't help? Every decline is a catalog or training gap with a dollar value attached.
FAQ
Is Brand Agents a Shopify feature?
- No. Brand Agents is a Microsoft product, built into Microsoft Clarity and announced on January 8, 2026. It runs on Shopify stores (and WooCommerce) through the Clarity app and Microsoft's waitlist. It was not part of Shopify's Spring '26 Edition.
Do I need Shopify Plus for a Brand Agent?
- No. Microsoft lists no Shopify plan requirement. You need the Microsoft Clarity app and access through Microsoft's waitlist. Plus matters for other things, like checkout extensibility, unlimited B2B catalogs and higher API limits, not for Brand Agents.
Will a Brand Agent hallucinate product details?
- Only if your catalog is incomplete. The agent reasons over the data you give it — if an attribute is missing, it may infer or decline. A complete, structured catalog is what keeps answers accurate, which is why we audit and enrich the catalog before going live.
Can I control the agent's tone?
- Yes — brand voice is a core training input. The agent should sound like your brand, not a generic assistant. We tune it on your voice and review real transcripts to keep it on-brand.
How many conversation flows do I need?
- Most categories are well-served by 20–50 flows covering the most-asked questions and the high-value edge cases. Quality and coverage of real shopper intents matter more than raw count.
Sources
- 1.Shopify B2B features by plan · Shopify Help CenterB2B on Basic, Grow, Advanced and Plus, not Starter or Retail; 3 active catalogs vs unlimited; Plus-only and Advanced-only features
Operators who set up agentic storefronts, Microsoft Brand Agents, Copilot Checkout, and AI-ready catalogs for mid-market merchants. We publish the frameworks we actually use with clients.
LinkedIn