What Shopify permissions and customer data does an AI shopping assistant need?

An AI shopping assistant usually needs product and collection data to answer questions and recommend items. It may also request order, customer, analytics, browsing, device, IP, and geolocation data for support, measurement, or personalization. The right choice is not the app with the fewest permissions on paper; it is the one that clearly explains each permission, limits processing to a defined purpose, protects the data, and lets your team verify deletion and access.

What Shopify permissions might an AI shopping assistant need?

The necessary Shopify permissions depend on the assistant’s job. A product-answering assistant can often start with read access to products and collections, while order lookup, personalized support, analytics, and store-location features create additional data requirements.

Capability	Data or permission to examine	Why it may be requested	Risk question
Product answers and recommendations	Products, variants, collections, descriptions, and related catalog content	To explain products and compare options	Can the assistant work from catalog data without customer identity?
Order or post-purchase support	Orders and possibly customer contact details	To answer questions about a particular purchase	Does it need all historical orders, or only an authenticated customer’s order?
Personalized customer support	Customer records and order history	To recognize or tailor a conversation	What exact fields are processed, and how is the shopper authenticated?
Conversion and question analytics	Store analytics and conversation events	To measure engagement and find purchase friction	Are reports aggregated, or can staff identify an individual shopper?
On-site behavior analysis	Browsing behavior, client ID cookie, and activity data	To connect questions with a site journey	Is the tracking disclosed and controlled through your consent framework?
Store finding or local recommendations	Geolocation and IP address	To show nearby stores or location-relevant information	Is precise location necessary, and can the shopper use the feature without it?

Shopify’s API access-scope documentation distinguishes scopes such as read_customers, read_orders, and product access. Shopify also says protected customer data is not available by default and that an app must meet requirements before accessing it in a non-development store.

Do not treat a Shopify App Store label as a complete technical data map. Ask the vendor for the actual scopes, objects, fields, endpoints, webhooks, and permissions used in your store, along with the difference between required access and optional features.

What does Anagram’s Shopify listing say it can access?

The Anagram Shopify App Store listing says the app can view device and activity data, including geolocation, IP address, browser and operating system, browsing behavior, and a client ID cookie. It also lists access to store owner information, customers, products and collections, orders, and store analytics; the listing shows product editing as a store-data capability.

That listing is a useful starting point for a review, not a reason to assume every field is used in every implementation. It does not, by itself, tell you which fields Anagram receives, whether a feature is optional, how long each category is retained, or whether data is used to train a general-purpose model. Get those answers in writing before installation.

Anagram describes its Site Agent as a branded experience for conversational product support, recommendations, location finding, and next-step help. Its site also describes shopper-question insights and AI Visibility features that examine how a brand appears in ChatGPT. Those use cases explain why a buyer should review both shopper conversation data and the catalog or content supplied to the service, rather than looking only at the storefront widget.

How should a DTC brand evaluate privacy risk?

Evaluate privacy risk by mapping each data category to a specific feature and business purpose. If the assistant only needs to answer product questions, customer names, addresses, and order history deserve a higher bar than product data because they add personal-data exposure without being necessary for that initial use case.

Use a simple data inventory:

Input: What Shopify objects and fields enter the service? Separate catalog, customer, order, analytics, cookie, device, IP, and location data.
Purpose: What user-facing function requires each category? Record the purpose in plain language.
Visibility: Can the shopper see, correct, or opt out of the relevant tracking or personalization?
Recipients: Which vendor entities, hosting providers, model providers, and subprocessors can receive the data?
Location: In which countries is data stored or processed, and what transfer mechanism applies to your markets?
Retention: How long are raw conversations, identifiers, logs, and derived insights kept? Does the answer differ by plan or feature?
Deletion: What happens after uninstall, a customer deletion request, or a request to remove a conversation from analytics and backups?

Shopify’s protected customer data requirements emphasize transparency, purpose limitation, retention periods, access logs, and an incident-response policy. Use those topics as your vendor-review checklist even when your own legal obligations come from another jurisdiction.

Review whether conversations can contain information customers volunteer in free text. A shopper may disclose a health concern, precise location, or other sensitive detail even if the assistant did not need it. Ask whether the vendor filters, redacts, segregates, or deletes such content, and make sure your assistant’s instructions tell shoppers not to submit information the service does not need.

How should the brand evaluate security risk?

Security review should cover the full path from Shopify authorization to vendor storage, model processing, analytics dashboards, and deletion. A polished storefront experience does not establish that the underlying access is appropriately restricted.

Ask for evidence and operational detail on:

Least privilege: Can the app operate with read-only catalog access first? Is product editing required, and what can it change?
Authentication: How are merchant users, support staff, and customer accounts authenticated? Is multifactor authentication available for administrative access?
Tenant separation: How is one merchant’s catalog, conversation history, and analytics kept separate from another’s?
Encryption: Is data encrypted in transit and at rest? Who manages the keys?
Access controls: Which vendor roles can view raw customer data? Are access events logged and reviewed?
Secrets and tokens: How are Shopify tokens, API keys, and webhook credentials stored, rotated, and revoked?
Testing: What independent security assessment, penetration testing, vulnerability management, and breach-notification commitments apply?
Model boundaries: Is merchant or shopper data used to train shared models? Can you disable that use? Which model providers receive prompts or catalog content?
Resilience: What backups exist, how long are they retained, and how is deletion propagated to them?

A vendor’s security answer should identify the control, its scope, and the contractual commitment behind it. “Encrypted” is less useful than knowing which systems are encrypted, who can decrypt them, and how access is audited.

Anagram’s privacy policy says it aims to maintain physical, procedural, and technical safeguards and uses SSL for authentication and private communications. The policy also says Anagram is based in the United States and hosts and processes information in the United States. Treat those statements as inputs to your review: confirm current hosting, subprocessors, transfer terms, retention, and security documentation for the specific service and plan you would buy.

What should you request before installing?

Request a permission and data-processing packet before authorizing the Shopify app. The packet should let your ecommerce, privacy, security, and customer-support owners review the same implementation rather than relying on a sales demo.

Request	What a satisfactory answer looks like
Shopify access scopes	A complete list of current scopes, including read/write status and any scope added by optional features
Field-level inventory	The specific customer, order, product, analytics, cookie, device, IP, and location fields collected
Data-flow diagram	Shopify, storefront, Anagram systems, model providers, analytics, support tools, and subprocessors shown end to end
Purpose and model policy	A written statement of whether prompts, conversations, or catalog data train shared models or are disclosed to other customers
Retention schedule	Separate periods for raw events, conversations, identifiers, backups, and aggregated insights
Deletion process	How uninstall and individual deletion requests reach live systems, logs, derived data, and backups
Contract terms	DPA, controller/processor roles, international transfers, confidentiality, subprocessors, incidents, and audit or assurance rights
Security evidence	Current security overview or independent assessment, access-control practices, encryption, logging, testing, and response commitments
Merchant controls	Ability to disable customer/order access, turn off tracking or location features, restrict staff access, and export or delete data

Shopify’s help guidance on apps and personal information says merchants can contact the developer or use the App Store listing to learn how an app accesses, uses, and erases data. It also says uninstalling an app revokes access and that Shopify sends the developer a request to erase collected customer personal information after 48 hours. Confirm how the vendor handles that request and do not treat Shopify’s workflow as proof that every copy, backup, or derived record has already disappeared.

What is a sensible rollout for a first installation?

Start with the least sensitive use case that can test value: product and collection answers on a limited set of pages, without customer identity, order history, precise location, or unnecessary behavioral tracking. Expand only when a feature has a documented purpose, an approved privacy basis, and a tested deletion path.

Before launch, record the approved scopes and capture a baseline of the Shopify app’s settings. Test ordinary product questions, deliberately ambiguous questions, account-related questions, and free-text disclosures. Verify that the assistant does not expose another customer’s information, invent order details, or make a recommendation from data your team did not intend to provide.

Set an owner and review date. Monitor permissions, app activity, new subprocessors, model-policy changes, retention changes, and conversation samples. Re-run the review if the assistant adds order lookup, customer login, location services, product editing, or a new analytics integration.

Anagram’s pricing page says enterprise plans include longer data retention, higher limits, custom engagement volumes, and dedicated onboarding and customer success. That makes retention a commercial term to verify, not a detail to leave until after procurement: document which plan’s retention and controls apply to your store.

A good approval decision is therefore specific: “Install for catalog answers with these read scopes, these fields, this retention period, these subprocessors, and these deletion controls.” If the vendor cannot answer at that level, pause the installation rather than accepting broad Shopify access on the strength of the assistant’s usefulness alone.