AI-Powered Contact Center Transformation for Pet Retail

petco · E-commerce · 2024 · source Agentic: no · promotional bias: med · source-backed · extracted with prompt 0.5.4 (schema v2)

Generated from extraction.json — do not edit by hand. Every claim cites verbatim source evidence; stated means the quote directly asserts the claim, implied means it supports it indirectly.

Problem

Domain: Retail customer care

Petco needed to improve customer-care efficiency and insight across high-volume pet retail interactions without misrouting emotionally sensitive conversations to automation.

Constraints:

More than 10,000 daily customer contacts
Pet-related issues can be emotionally high-stakes
Knowledge content and governance must be ready for customer-facing AI

Architecture

Petco consolidated contact-center AI on Amazon Connect so calls, IVR, chat, Contact Lens transcripts, summaries, QA rules, Q in Connect assistance, order-status self-service, and chatbot workflows sit on one platform. The architecture favors native transcript and analytics access, rapid IVR/chat iteration, and selective automation for routine interactions while preserving humans for emotionally sensitive pet-care conversations.

Diagram confidence: reconstructed (reconstructed from prose).

flowchart LR
  amazon-connect-platform["amazon-connect-platform"]
  contact-lens-transcripts["contact-lens-transcripts"]
  automated-qa-rules["automated-qa-rules"]
  q-in-connect-agent-assist["q-in-connect-agent-assist"]
  knowledge-base["knowledge-base"]
  chatbot-flows["chatbot-flows"]
  ivr-order-status["ivr-order-status"]
  supervisor-reporting["supervisor-reporting"]
  amazon-connect-platform --> contact-lens-transcripts
  amazon-connect-platform --> automated-qa-rules
  amazon-connect-platform --> q-in-connect-agent-assist
  amazon-connect-platform --> chatbot-flows
  amazon-connect-platform --> ivr-order-status
  contact-lens-transcripts --> q-in-connect-agent-assist
  contact-lens-transcripts --> automated-qa-rules
  q-in-connect-agent-assist --> knowledge-base
  chatbot-flows --> knowledge-base
  automated-qa-rules --> supervisor-reporting

Drag to pan · scroll or use +/− to zoom · reconstructed from the source's described component topology

Components

Component	Role	Interacts with
`amazon-connect-platform`	Unified platform for calls, IVR, chat, transcripts, analytics, summaries, routing, QA, and chatbot/self-service flows.	`contact-lens-transcripts`, `automated-qa-rules`, `q-in-connect-agent-assist`, `chatbot-flows`, `ivr-order-status`
`contact-lens-transcripts`	Conversation transcripts and summaries that feed case notes, supervisor review, insights, and downstream agent assistance.	`amazon-connect-platform`, `q-in-connect-agent-assist`, `automated-qa-rules`
`automated-qa-rules`	Rules that evaluate every interaction for quality risks such as long holds and expose coaching opportunities.	`contact-lens-transcripts`
`q-in-connect-agent-assist`	Surfaces paragraph answers and knowledge directly to agents based on the current transcript.	`contact-lens-transcripts`, `knowledge-base`
`knowledge-base`	Initial curated article set used by the chatbot and Q in Connect for high-frequency questions.	`q-in-connect-agent-assist`, `chatbot-flows`
`chatbot-flows`	Amazon Connect chat launch using flows for a small set of high-frequency use cases such as order status.	`amazon-connect-platform`, `knowledge-base`
`ivr-order-status`	Self-service IVR path that recognizes likely order-status intent and can read back order information before agent escalation.	`amazon-connect-platform`
`supervisor-reporting`	Uses summaries, QA findings, and contact-center analytics to identify coaching, routing, and customer-experience issues.	`automated-qa-rules`

Data flow

Customer contacts enter Amazon Connect through phone, IVR, or chat. Contact Lens captures transcripts and call summaries, QA rules flag interaction patterns such as long holds, Q in Connect surfaces answers from conversation context, and self-service flows handle order-status or grooming/vet routing before escalating higher-value conversations to care team members.

“we've got calls, we have IVR, we have chat all running through the same platform. And so we get to benefit”transcript, stated

“to have Amazon Connect and Q and Connect just pop a paragraph and answer based on the transcript that's running.”transcript, stated

Why this architecture

Petco chose consolidation because disconnected tools would consume scarce engineering capacity and make transcript/insight extraction harder. It also avoided automating all contacts: the architecture routes routine, low-impact work to self-service while reserving human attention for interactions where pet owners' emotions or loyalty matter.

“the team has a strong perspective that we want fewer integrations. We don't have an unlimited technology budget.”transcript, stated

“we all get excited about the shiny object and the technology. But what is it that we're here to do?”transcript, stated

Models used: Amazon Connect AI features · Amazon Q in Connect

Scale

Petco operates more than 1,500 stores in the US and Mexico.
“with over 1,500 stores in the US and Mexico.”transcript, stated
The contact center has roughly 400 care team members.
“So we have roughly 400 care team members. And we service everything.”transcript, stated
Petco speaks with more than 10,000 customers per day.
“We talk to more than 10,000 customers a day. That's equivalent to more than 30 of our retail stores.”transcript, stated
The grooming-salon routing pilot began with six salons and could be built in weeks rather than months.
“And we started with like six grooming salons.”transcript, implied

Alternatives considered (roads not taken)

Keeping many stitched-together contact-center applications

Rejected because: Petco wanted fewer integrations and easier access to shared transcripts, calls, IVR, and chat data.

“So it's one of the reasons Amazon Connect works really well for us because we have contact lens with the transcripts, we've got calls, we have IVR, we have chat all running through the same platform.”transcript, stated

Automating every customer-care interaction

Rejected because: Petco segments routine, low-impact work from high-impact conversations where human care protects loyalty and satisfaction.

“So how do we get the best service, best quality of service, best customer satisfaction for the lowest cost?”transcript, implied

Large, slow routing projects for store and vet workflows

Rejected because: The team used Amazon Connect to build IVRs and reroute calls in weeks instead of turning it into a months-long production effort.

“the IT team was able to build those IVRs, reroute the calls within a couple weeks versus a couple months and having it be a big production.”transcript, stated

Patterns

`human-in-the-loop-gate` (routing)

Petco routes low-impact order-status questions toward self-service while preserving human care for pet loss, retention, and high-lifetime-value situations.

“Those are high impact conversations that we wanna invest in a great human interaction. Whereas maybe some of the other ones we're happy to use AI or some other self-help experience”transcript, stated

`context-compaction` (agent-assist)

AI-generated call summaries are stored in case notes and make later review faster than reading full transcripts.

“we're doing call summaries, we have that turned on in Amazon Connect. The benefit of that, initially the big focus was we take those and we put it right in the case notes.”transcript, stated

`observability-tracing` (quality-management)

Every contact can be checked against QA rules, sentiment, and agent/team averages instead of relying only on sampled coach reviews.

“So we created a rule in Amazon Connect that happens on every single call to flag, okay, this agent is putting customers on hold way too much”transcript, stated

`rag-retrieval` (knowledge-assist)

Q in Connect surfaces knowledge-base answers to agents and powers a small chatbot launch for high-frequency questions.

“we've loaded 20 KB articles into what is before reInvent, Q and Connect, loaded it and are getting gen AI answers to those.”transcript, stated

Tradeoffs

Where to apply automation.

Chose AI/self-service for low-impact routine inquiries over automating every customer-care conversation.

Rationale: Petco explicitly separates order-status-style interactions from high-impact emotional or retention conversations.

“But differentiating these is the way that we think about balancing that trade off.”transcript, implied

How to structure the contact-center platform.

Chose fewer native integrations on Amazon Connect over many stitched-together applications.

Rationale: Finite technology resources and cross-channel insight needs made integration overhead a core concern.

“we want fewer integrations. We don't have an unlimited technology budget. We have very finite resources. We don't wanna spend that time managing a bunch of integrations”transcript, stated

Gotchas (what bit them)

Knowledge articles written for humans did not automatically work well for AI.

Root cause: Article format and content changed the quality of both customer-facing answers and agent assist.

Resolution: Petco worked with Amazon on AI-specific article formatting and continues to tune content.

“we had them formatted in the right way which was different than how we had them formatted with agents. And so we've spent a lot of time in that.”transcript, stated

The initial chatbot launch missed real customer utterances.

Root cause: The team thought they had covered all utterances but production traffic exposed gaps.

Resolution: They added missed utterances in real time and prioritized reporting for ongoing tuning.

“I will tell you, big lesson learned, we thought we had all of the utterances covered and we didn't. And so we're learning a lot and we're adding those in real time.”transcript, stated

Customer-facing AI required organizational governance and safeguards.

Root cause: AI use in customer care felt risky and was complicated by state-specific legal requirements.

Resolution: Petco used an AI Governance Council and explained goals, usage, and safeguards to stakeholders.

“We went through our AI Governance Council and then we had a bunch of stakeholders on board and said this is what we're trying to accomplish. This is how we're using AI.”transcript, stated

Evaluation

Petco uses journey-level containment thinking, chatbot reporting, contact summaries, QA rules, hold-time flags, sentiment analysis, and supervisor dashboards.

Metrics: journey containment · missed utterances · drop-off/abandonment · hold time over two minutes · sentiment · CSAT

“How are you measuring containment across your entire ecosystem? If a customer's making a phone call, but they chat you 30 minutes later, was that really successful?”transcript, stated

“what are the utterances that are getting missed? Where are customers dropping out or abandoning in that experience?”transcript, stated

Outcomes

Petco launched call summaries, automated QA rules, agent-assist pilots, chat with a small knowledge base, and IVR order status.
The grooming and veterinary routing pilots could be built in weeks instead of months.

Petco reports more than 10,000 daily customer contacts, roughly equivalent to more than 30 retail stores' worth of customer voice.

“We talk to more than 10,000 customers a day. That's equivalent to more than 30 of our retail stores.”transcript, stated