Instagram DMs for European Shopify brands
Tyra puts WhatsApp, Instagram, Facebook Messenger, and SMS in one inbox, hosted in Ireland and wired into your Shopify catalog so the AI agent answers from real stock instead of guesses.
The problems you know too well
Every Shopify brand we onboard has some version of these challenges.
Your support lead is tab-switching between Gmail, the WhatsApp web app, and Instagram on her phone
That is not a workflow. That is a person remembering, by hand, that the customer who emailed about a wrong size on Tuesday is the same one who DM'd on Sunday about the original order. She gets it right most of the time. The cost of the times she does not is a refund, a one-star review, or a customer who quietly stops opening the newsletter. Every Shopify brand we onboard has some version of this person, and every one of them is doing heroic work to compensate for software that treats four channels as four problems.
The DM tab gets checked at 9am and at 6pm, and that is the whole reply SLA
Meta's own merchant research keeps landing on the same number: shoppers expect a reply inside an hour on social, and most brands answer somewhere between four hours and the next morning. The gap is where the cart gets abandoned. The DM that asked "is this back in stock in 38?" at 2pm gets answered at 9am the next day, by which point the customer has either bought from someone else or moved on. Notifications are not an inbox. A notification badge that reads 47 is a synonym for "nobody has triaged this since lunch."
Story mentions vanish in 24 hours and so does the goodwill
A customer reshares your product to her story. That is the cheapest, warmest UGC you will ever get, and the window to reply is the lifetime of the story. Miss it and you have lost the moment to send a thank-you, a code, a repost permission ask. Most brands miss it because story mentions live in a separate notifications tray that nobody opens before the 24 hours are up. We see this on almost every audit: dozens of mentions a month, near-zero replies, no UGC pipeline downstream.
A reel goes viral and the inbox is unusable for a week
A creator posts about you, or one of your own reels takes off. Comments pile up, half of them ask "where can I buy this?", a third are friend tags, the rest are influencer pitches and the occasional scam. Real buying intent gets buried under noise. The team either burns a day reading every comment, or gives up and lets the post coast. Both options cost money. There is no manual moderation strategy that scales past a few hundred comments on a single post, and "we will get to it on Monday" is not a strategy at all.
How Tyra solves this
Four pillars that make Instagram customer support actually work.
EU-built infrastructure
Tyra runs in AWS Ireland, eu-west-1. Conversation bodies, customer attributes, order history, agent transcripts, audit logs: all stored in the EU, no replication to US regions for storage. Meta still processes the underlying Instagram traffic under its own DPA before it reaches us via the Graph API, and we are precise about that distinction in our own DPA so your DPO does not have to translate marketing copy into a real answer. Right-to-be-forgotten requests are scrubbed from Tyra's systems inside 24 hours. Sub-processors are listed, named, and dated on a public page; when one changes you get an email rather than a footnote update. For a Shopify brand selling into Germany, France, or the Nordics, this is the difference between a procurement review that closes in a week and one that drags into the next quarter.
- AWS Ireland (eu-west-1) hosting
- Right-to-be-forgotten within 24 hours
- Public sub-processor list with email notifications
- GDPR-compliant data processing
EU infrastructure with Ireland hosting
WhatsApp and Instagram unified
Instagram is the discovery channel. WhatsApp is where the conversation often continues, especially in Germany, Italy, Spain, the Netherlands. A customer DMs on Instagram, then a day later messages your WhatsApp number from the link in your bio. In most tools those are two unconnected threads with two unconnected histories. In Tyra they are the same customer record with both channels stacked in one timeline, and the next reply can go on whichever channel the customer last used. We are an official BSP through 360dialog, so WhatsApp templates, interactive buttons, product messages, and the 24-hour service window all work the way Meta intends. Outbound order updates, abandoned cart nudges, and post-purchase flows go on WhatsApp where they get opened, not buried in promotional email. None of this is bolted on after the fact: WhatsApp was the first channel we built, and Instagram was added once the inbox primitives could carry it properly.
- Same customer record across channels
- Official BSP through 360dialog
- WhatsApp templates and interactive buttons
- Reply on whichever channel the customer last used
WhatsApp and Instagram threads unified
Shopify-native integration
When a DM comes in, the right-hand panel renders the live Shopify customer: past orders, lifetime value, current cart, address, tags. When the AI agent or a human sends a product card, the image, price, and variant stock are pulled from Shopify at send time, not from a cached export that went stale last Thursday. If a variant drops to zero between selection and send, Tyra rechecks and either swaps the variant or offers the customer a back-in-stock opt-in. Instagram's Messaging API caps product carousels at 10 cards. We respect that cap. We also link the "buy" action to your real Shopify checkout rather than an in-app shopping surface that is not available in most EU countries. Refunds, order edits, and cancellations execute against Shopify with one click from the side panel, and the action shows up in the conversation timeline so the next agent or the AI knows it happened.
- Live customer data in side panel
- Real-time inventory checks on product cards
- One-click refunds, order edits, cancellations
- Back-in-stock opt-in for out-of-stock variants
Shopify side panel with live inventory
AI grounded in your catalog
The agent is grounded in your Shopify catalog, your help center, your shipping policy, and your return policy. It is not grounded in the open internet, and it cannot name a product that does not exist in the connected store with positive stock. We learned that lesson the hard way in the first build, when the agent confidently recommended a SKU we had discontinued six months earlier. The rewrite fails closed: no catalog match, no product name, full stop. Intent classification routes refunds, complaints, legal mentions, and any message where the model's own confidence drops below threshold straight to a human queue. The agent replies in the language the customer wrote in, across the twelve EU languages we currently support. It will not promise a shipping date it cannot verify, will not argue with an upset customer, and will not extend a conversation past a second clarifying question without offering a human handoff. On the brands we have measured, it closes between 40 and 60 percent of inbound DMs without escalation, and the share is higher for stores with clean catalog data and a real returns page.
- 40-60% of DMs closed without escalation
- 12 EU languages supported
- Automatic escalation on low confidence
- Refuses to promise unverifiable shipping dates
AI agent reply with confidence controls
Every message goes to the right place
The AI classifies inbound DMs so your team focuses on what matters.
Customer questions
Full agent treatment with catalog grounding and live Shopify data
Press & collab pitches
Routed to a dedicated queue for humans to handle on their schedule
Fan messages
Templated thank-you responses that acknowledge without overloading
Spam & scams
Filtered out automatically so they never reach your team
Catch story mentions before they expire
Tyra catches the story mention webhook in real time and triggers whatever flow you configure. The reply lands while the story is still live, which is the only window where any of it matters.
- Thank-you with a discount code
- Permission-to-repost request for UGC pipeline
- Follow-up offering a referral code
- Custom flows based on customer tags
Story mention automation flow
Frequently asked questions
Does Tyra work for personal Instagram accounts?
No. Meta requires an Instagram Business or Creator account linked to a Facebook Page for Messaging API access. If yours is not set up yet we walk you through the conversion during onboarding, which usually takes about ten minutes.
How does Tyra stay inside Instagram's 24-hour messaging window?
We track the window per conversation and show the remaining time in the inbox. After 24 hours we will not let you send a free-form message; the only outbound options are approved message tags, mainly HUMAN_AGENT for verified support cases.
Can Tyra send cold DMs or broadcast promotions outside the window?
No, and any tool promising this is either lying or actively risking your account. Comment-to-DM automation works because the user's comment opens a messaging window. Cold DMs to people who have not engaged are a policy violation.
What happens when a viral post or campaign floods the inbox with influencer pitches?
The agent classifies inbound DMs into buckets like real customer question, press or collab pitch, fan message, and spam. Customer questions get the full agent treatment with catalog grounding. Press pitches route to a dedicated queue for a human to handle on their own schedule. Fan messages get a templated thank-you. The point is to keep the buying-intent DMs from drowning.
How do story mentions get handled?
Tyra catches the story mention webhook in real time and triggers whatever flow you configure: a thank-you with a discount code, a permission-to-repost request for your UGC pipeline, or a follow-up offering a referral code. The reply lands while the story is still live.
Where exactly is customer data stored?
Conversation content, customer attributes, and order context are stored in AWS Ireland, eu-west-1, with no replication to US regions for storage. Meta processes the underlying Instagram traffic under its own DPA before it reaches us through the Graph API. We publish a current sub-processor list and notify customers by email when it changes.
Can the AI agent answer in Italian or Swedish if the customer writes in those languages?
Yes. Language is detected per message and replies match. We currently support twelve EU languages with first-class quality. If a customer switches mid-thread the agent switches with them.
Does Tyra replace Meta Business Suite or sit alongside it?
Both tools can read the same DMs since Meta's API allows multiple readers. Most teams use Tyra as the primary inbox for triage, agent replies, and Shopify actions, and dip into Business Suite occasionally for things outside our scope like ad creative review. We do not schedule content; tools like Later and Buffer own that category.
Ready to unify your Instagram inbox?
See how Tyra handles your Instagram DMs, story mentions, and comments alongside WhatsApp, Facebook Messenger, and SMS.