Virtual Assistant for Subscriptions and Pet-Care Questions at a DTC Pet Supplies Retailer
Let's Connect
Overview
What we built
A DTC pet supplies retailer's support inbox was buried in routine questions its own customers could have answered themselves. We built a virtual assistant that answers most of them directly.
In plain terms: this retailer runs 15 regional stores alongside a fast-growing autoship subscription programme, and its support team was buried. Roughly two-thirds of contacts were routine matters, delivery dates, skipping a shipment, swapping a food size, yet each one queued for hours in a shared inbox regardless of how simple the question was. Peak-season response times stretched past a day, and subscription pauses that should have kept customers quietly turned into outright cancellations.
We built a virtual assistant embedded directly in the website and mobile app, connected to the retailer's order, subscription and product systems rather than a static help page. Customers reschedule or skip autoship deliveries, track orders and get feeding and pet-care answers from an approved knowledge base, with full-context handover to a human agent for anything sensitive or health-related. Within 3 months, 62% of routine support enquiries were resolved end-to-end by the assistant, and first-response time for the tickets that still needed a human fell from 9 hours to under 45 minutes.
The Problem
Support tickets outpacing headcount
The retailer's autoship subscription programme was the fastest-growing part of the business, and every new subscriber added a steady trickle of small questions into the same support inbox as everything else: delivery dates, skipping a shipment, swapping a food size. None of these needed real judgement, but each one queued in the shared inbox exactly like a complex complaint.
Roughly two-thirds of all contacts were this kind of routine matter, yet the support team had no way to separate them from the questions that genuinely needed a person's attention. Every enquiry waited its turn in the same queue, so a simple delivery-date question sat behind whatever had arrived before it, and response times stretched accordingly.
Peak season made the queue worse and the consequences sharper. Response times stretched past a day, and subscribers who wanted to pause a delivery rather than receive it could not get a quick answer, so a share of those pauses quietly turned into cancellations instead of the temporary skip the customer actually wanted.
Routine questions, one queue
Roughly two-thirds of contacts were simple matters, delivery dates, skipping a shipment, swapping a food size, yet each waited in the same shared inbox as complex issues.
Hours for a simple answer
Every enquiry queued for hours regardless of complexity, so a customer asking a delivery date waited as long as one raising a genuine problem.
Peak season blowout
Response times stretched past a day during peak season, when the retailer's 15 stores and autoship programme both generated the most contact volume at once.
Pauses becoming cancellations
Subscribers who wanted to pause a delivery could not get a quick answer, so pauses that should have been temporary quietly turned into cancellations instead.
What it was costing them
Every routine question stuck in the shared inbox competed with a genuinely complex one, so support time went into answering delivery-date questions instead of the sensitive or health-related issues that actually needed a person. Peak-season delays past a day pushed subscribers toward the easiest exit, cancelling, and each cancellation erased recurring revenue the fast-growing autoship programme had only just won.
The Solution
Subscription-aware virtual assistant
We built the virtual assistant to live where customers already are, embedded directly in the website and the mobile app rather than as a separate help centre. It connects directly to the retailer's order, subscription and product systems, so it can act on a request instead of just describing what the customer should do next.
For the two-thirds of contacts that were routine, customers can reschedule or skip an autoship delivery, track an order, or get an answer to a common feeding or pet-care question, drawn from an approved knowledge base, without waiting in the shared inbox at all. Skipping a delivery became as easy as pausing one, addressing the exact confusion that had been turning pauses into cancellations.
Anything sensitive or health-related still goes to a person, with full context handed over so the customer never has to repeat themselves. That handover is deliberate: the assistant is built to resolve the routine and recognise its own limits, not to keep a worried pet owner talking to a bot when what they need is a human.
Key decisions
Meet customers in-app
The assistant is embedded directly in the website and mobile app, rather than a separate help centre customers would have to seek out.
Connect to live systems
It connects directly to order, subscription and product systems, so it can reschedule or skip a delivery instead of only explaining how to.
Make skipping as easy as pausing
Autoship rescheduling and skipping sit in the same simple flow, addressing the exact gap that had been turning pauses into cancellations.
Answer only from approved content
Feeding and pet-care questions are answered from an approved knowledge base, so customers never receive improvised advice about their pet.
Hand off sensitive issues with context
Anything sensitive or health-related transfers to a human agent with full conversation context, so the customer is never asked to start over.
Measurable Impact
What changed after launch
The shared inbox stopped being the bottleneck for routine questions. Within 3 months, 62% of routine support enquiries were resolved end-to-end by the assistant, and first-response time for the tickets that still needed a human fell from 9 hours to under 45 minutes, because those tickets were no longer queued behind delivery-date questions.
The subscription programme felt the change directly: cancellations fell 17% as customers switched to skipping or rescheduling deliveries instead of cancelling outright. Across more than 30,000 assistant-handled chats, customer satisfaction averaged 4.5 out of 5, evidence that resolving a routine question instantly beats waiting hours for the same answer.
Routine enquiries
Queued for hours in a shared inbox
62% resolved end-to-end by the assistant
Human first response
An average of 9 hours to any reply
Under 45 minutes within 3 months
Subscription pauses
Quietly turning into outright cancellations
Cancellations down 17% as customers skip instead
Customer satisfaction
Untracked across a buried support queue
4.5 out of 5 across 30,000+ chats
Headline results
62% of routine support enquiries resolved end-to-end by the assistant within 3 months
First-response time for human-handled tickets cut from 9 hours to under 45 minutes
Subscription cancellations fell 17% as customers switched to skipping or rescheduling deliveries
CSAT on assistant-handled conversations averaged 4.5 out of 5 across 30,000+ chats
Tech & Tools Used
What powered the build
Every tool below earned its place in this engagement. Here is the part each one played.
Python (FastAPI)
Serves the assistant's core logic as an internal API, handling requests from both the website and the mobile app through one consistent service.
LangChain
Manages retrieval from the approved knowledge base and enforces the handover rules that send sensitive or health-related questions to a human agent.
OpenAI API
Generates the assistant's answers to feeding and pet-care questions, constrained to the approved knowledge base so guidance stays consistent and accurate.
PostgreSQL
Stores conversation history and resolution outcomes, giving the support team one record of what customers asked and how each contact was resolved.
Redis
Holds live session state so a customer can move between the website and the mobile app without losing their place mid-conversation.
Next.js
Renders the website experience where the assistant appears, keeping the conversation available wherever a customer already is on the site.
React Native
Builds the assistant experience inside the mobile app, so customers can reschedule or skip an autoship delivery from their phone.
Zendesk API
Hands sensitive or health-related conversations to human agents with full context attached, so tickets that need a person never lose the thread.
Stripe Subscriptions
Executes the reschedule and skip actions against the autoship programme's billing records, so a skipped delivery is reflected immediately.
AWS ECS
Hosts and scales the assistant's services in production, keeping response times fast for both routine chats and peak-season volume.
Ready to Build your Pet Retail & E-Commerce Business with Chatbots & Virtual Assistants
Ask Byte
Ask Byte
Typically replies instantly
just Now
Hi! I'm OrganByte's assistant. How can I help you today?
AI-generated content may be incorrect
