Contact Center Transformation with AI-Powered Customer Service and Agent Assistance

canada_life · Insurance · 2025 · source Agentic: yes · promotional bias: med · source-backed · extracted with prompt 0.5.5 (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: Financial-services contact center modernization and AI customer support

Canada Life needed to replace a siloed, manual, phone-heavy contact-center ecosystem with a scalable, omnichannel, AI-enabled platform that could reduce wait times, improve routing, support agents, and preserve secure human handoffs.

Constraints:

Architecture

Canada Life migrated from a legacy contact-center platform to Amazon Connect, supported by Salesforce, Calabrio, Bedrock, Lex, Lambda, knowledge bases, API integrations, SharePoint integrations, monitoring, metrics, and integrated data. The modernization created a unified agent desktop, automated contact flows, proficiency-based routing, call summarization, voice-to-text, automated authentication, self-service IVR, scheduled callbacks, and the Cali chatbot, with a roadmap for next-best action, conversational IVR, agent assist, custom summarization, and fraud detection.

Diagram confidence: explicit (topology described by the source).

flowchart LR
  customer-channels["customer channels"] --> amazon-connect["Amazon Connect"]
  customer-channels --> cali-chatbot["Cali chatbot"]
  amazon-connect --> unified-agent-desktop["unified agent desktop"]
  amazon-connect --> salesforce-crm["Salesforce CRM"]
  amazon-connect --> calabrio-workforce-optimization["Calabrio workforce optimization"]
  amazon-connect --> routing-and-authentication-layer["routing and authentication layer"]
  amazon-connect --> llm-intelligence-layer["LLM intelligence layer"]
  routing-and-authentication-layer --> unified-agent-desktop
  routing-and-authentication-layer --> llm-intelligence-layer
  llm-intelligence-layer --> knowledge-and-integration-layer["knowledge and integration layer"]
  llm-intelligence-layer --> cali-chatbot
  llm-intelligence-layer --> call-summarization["call summarization"]
  knowledge-and-integration-layer --> data-integration-layer["data integration layer"]
  knowledge-and-integration-layer --> monitoring-and-metrics["monitoring and metrics"]
  call-summarization --> salesforce-crm
  call-summarization --> unified-agent-desktop
  cali-chatbot --> unified-agent-desktop
  data-integration-layer --> monitoring-and-metrics
Drag to pan · scroll or use +/− to zoom · reconstructed from the source's described component topology

Components

Component Role Interacts with
customer channels Voice, chat, text, callback, and self-service entry points for customers. Amazon Connect, Cali chatbot
Amazon Connect Cloud contact-center backbone for telephony, routing, omnichannel support, and unified workflows. customer channels, unified agent desktop, Salesforce CRM, Calabrio workforce optimization, routing and authentication layer, LLM intelligence layer
unified agent desktop Single desktop used by frontline agents after migration. Amazon Connect, Salesforce CRM, call summarization
Salesforce CRM Case and customer-record system that receives summaries and supports the 360-degree customer view. Amazon Connect, unified agent desktop, call summarization
Calabrio workforce optimization Workforce optimization layer supporting contact-center operations. Amazon Connect
routing and authentication layer Routes inquiries to the right agents and automates customer authentication securely. Amazon Connect, LLM intelligence layer, unified agent desktop
LLM intelligence layer Bedrock, Lex, and Lambda-based AI/automation layer for personalized information and chatbot/IVR capabilities. Amazon Connect, knowledge and integration layer, Cali chatbot, call summarization
knowledge and integration layer Knowledge bases, API integrations, SharePoint integrations, monitoring, metrics, and backend data. LLM intelligence layer, data integration layer
data integration layer Integrated data foundation needed to sustain personalized contact-center architecture. knowledge and integration layer, monitoring and metrics
call summarization AI capability that turns call transcripts into short case summaries. LLM intelligence layer, Salesforce CRM, unified agent desktop
Cali chatbot Authenticated AI assistant chatbot resolving interactions without agent escalation where possible. customer channels, LLM intelligence layer, unified agent desktop
monitoring and metrics Measurement layer used to manage journey performance and continuous improvement. knowledge and integration layer, data integration layer

Data flow

Customer interactions enter Amazon Connect across voice/chat/IVR channels, routing and authentication determine the right flow and agent, Lex/Bedrock/Lambda and knowledge/API integrations provide intelligence and automation, transcripts feed call summarization into Salesforce cases, monitoring and metrics guide improvements, and Cali resolves authenticated chatbot interactions when possible while escalating complex interactions to agents.

3 supporting quotes
“We have our backbone of Amazon Connect along with Salesforce and Calabrio to help us with the workforce optimization.”transcript, stated
“we are looking at LLM models, Bedrock, Lex, Lambda, to provide us with that intelligent, personalized information”transcript, stated
“knowledge basis. API integrations, SharePoint integrations, monitoring and metrics, these are all key because all of us has to go ahead and integrate with our backends to provide that 360 degree view that we are looking at. Last is data. Very critical”transcript, stated

Why this architecture

Canada Life chose a modular reusable Amazon Connect blueprint because it needed one scalable foundation for many business units, secure omnichannel routing, AI automation, workforce management, and a data/metrics feedback loop rather than disconnected point solutions.

3 supporting quotes
“about disjointed solutions. That is actually a struggle in the industry right now. For us, it is a very modular, reusable architecture that we have built in, something that is designed to accelerate and help us with any innovation, be it AI or non-AI. On the top, it is all about”transcript, stated
“one that delivers scalability, reliability, the one unified customer view and workforce optimization, very critical when it comes to contact center. Now, this actually is not just a reusable foundation that has helped us to go ahead and create that”transcript, stated
“Unless you go ahead and have the right monitoring and metrics, you don't know what to improve on. And that has been a game changer for Jenny and I to go ahead and run this journey. These are like small elements that are actually going ahead and creating the”transcript, stated

Models used: Amazon Bedrock LLM models · Amazon Lex · AWS Lambda · Cali chatbot

Scale

Alternatives considered (roads not taken)

Reduce call volume as the main goal

Rejected because: Canada Life explicitly says calls are not going down and the goal is channel diversification and capacity support.

“our call volumes are not going down, and our goal is not to go ahead and take down the call volumes. It is about diversifying the channels so that we can go ahead and support that increasing demand and capacity better. And at Canada Life, we are at the”transcript, stated
“It is about diversifying the channels so that we can go ahead and support that increasing demand and capacity better.”transcript, stated

Customize every business unit separately

Rejected because: Canada Life used out-of-the-box blueprint capabilities for each business unless customization directly affected daily work or productivity.

“more than 21 business units. Our biggest friction point was having contact flows that are manual against each business unit.”transcript, stated
“more than 90% operational cost savings just by using the out-of-box features that Amazon Connect brings and the automated change management that we have actually seen through. Now these are not small wins, these are transformation. These are game changers to”transcript, stated

Treat modernization as only a technology upgrade

Rejected because: The speakers frame modernization as reimagining customer service and agent empowerment.

“in the future. So when we talk about modernization, a lot of us think modernization is just a technology upgrade, but is it really so? It is actually, for Canada Life, about reimagining how we serve our customers, how we empower our teams. We focused on building a strong foundation, and that is”transcript, stated
“is it really so? It is actually, for Canada Life, about reimagining how we serve our customers, how we empower our teams. We focused on building a strong foundation, and that is Amazon Connect, the backbone of everything that we do, which turns complexity into simplicity. Our next”transcript, stated

Pure automation without human service

Rejected because: The introduction emphasizes balancing automation with human touch and seamless handoffs.

“being able to balance automation with the human touch to really create those seamless handoffs, that more personalized CX that really needs to meet the evolving customer requirements. Now, there are a tremendous number of analysts that are talking about this”transcript, stated
“many studies that you'll be able to automate 100% of your interactions and I suggest that that human agent isn't going anywhere for a while. Great. Great, and the thing I'll tell you, too, is make sure you know your customers. Know them, this is very critical.”transcript, stated

Patterns

human-in-the-loop-gate (customer-service)

Cali handles eligible interactions while freeing agents for complex cases and preserving escalation to humans.

“83% of those interactions was answered by the chatbot themselves and didn't get escalated to the agents.”transcript, stated
“Now, we are freeing up agent time for complex interactions. We have seen a call deflection of 1,900 calls per week.”transcript, stated

latency-optimization (operations)

Routing, automation, contact-flow standardization, and AI summarization reduced wait time, speed-to-answer, and handle time.

“So we had a 94% reduction in wait time, a 10% reduction in average handle time.”transcript, stated
“And in this year, in Canada Life, in our contact centers, our average speed to answer is 18 seconds for the whole year. We reduced also our abandonment rate by 2 to 4% depending on the business unit.”transcript, stated

observability-tracing (measurement)

Canada Life built monitoring and metrics into the architecture and used a measure-it-to-manage-it philosophy.

“Last but not the least, something very close to my heart, I say always, measure it to manage it. Unless you go ahead and have the right monitoring and metrics, you don't know what to improve on.”transcript, stated
“API integrations, SharePoint integrations, monitoring and metrics, these are all key because all of us has to go ahead and integrate with our backends to provide that 360 degree view that we are looking at.”transcript, stated

tool-use-orchestration (contact-center-platform)

The platform combines Amazon Connect, Salesforce, Calabrio, Bedrock, Lex, Lambda, knowledge bases, APIs, SharePoint, and data integration.

“We have our backbone of Amazon Connect along with Salesforce and Calabrio to help us with the workforce optimization.”transcript, stated
“When it comes to the core, we are looking at LLM models, Bedrock, Lex, Lambda, to provide us with that intelligent, personalized information that is needed to go ahead and provide that experience to our customers.”transcript, stated

context-compaction (agent-assist)

Call summarization compresses transcripts into short case summaries for after-call work and future callbacks.

“through call summarization, you can take the call transcript, break it down into three or four sentences, attach it to the case.”transcript, stated
“if that customer calls us back, we can pick up where we left off.”transcript, stated

safety-policy-enforcement (financial-services)

Automated authentication secures interactions while preserving seamless customer experience.

“Every interaction when you're automating it or when you're implementing AI on, it has to be secure while they're being seamless.”transcript, stated
“automated authentication goes ahead and has taken care of 37% of authentication.”transcript, stated

Tradeoffs

Use Amazon Connect as a shared backbone.

Chose One Amazon Connect foundation with Salesforce and Calabrio over Legacy siloed contact-center systems.

Rationale: A shared backbone reduced complexity/cost and created a reusable scalable platform across business units.

“we chose the best-in-class with Amazon Connect supported by Salesforce and Calabrio”transcript, stated
“60% reduction in complexity and cost because the unification of three applications to one Amazon Connect.”transcript, stated

Favor out-of-box reusable blueprint capabilities.

Chose Reusable blueprint for all business units over Custom implementation for each unit.

Rationale: The blueprint enabled faster migration, one risk sign-off, and consistent operations across 21 units.

“We migrated 21 business units in seven months. We had 100% uptime or 0% downtime. We converted 150 toll numbers.”transcript, stated
“Now with all of this, we have reached more than 90% operational cost savings just by using the out-of-box features that Amazon Connect brings and the automated change management that we have actually seen through.”transcript, stated

Use summarization before more advanced agent-assist roadmap items.

Chose Call transcript summarization as the first AI capability over Starting with next-best action or full conversational IVR.

Rationale: Summarization produced immediate after-call-work and handle-time benefits while preserving agent workflow.

“This was called summarization. It was our first AI capability in our contact center earlier this year.”transcript, stated
“This reduces time with an after-call work at sometimes zero after-call work.”transcript, stated

Deploy a chatbot inside authenticated spaces.

Chose Cali authenticated chatbot containment over Phone-only service for simple interactions.

Rationale: The chatbot handles common interactions while keeping agent capacity for complex cases.

“We are, very recently, done an implementation of our first ever AI assistant, a chatbot called Cali. And we have seen a consistent containment rate of 83%.”transcript, stated
“83% of those interactions was answered by the chatbot themselves and didn't get escalated to the agents.”transcript, stated

Gotchas (what bit them)

Customers still call even when automation is available.

Root cause: Complex financial and service issues require explanation, trust, and human preference across age groups.

Resolution: Diversify channels and design seamless handoffs rather than treating phone calls as obsolete.

“our Gen Z's are also preferring to pick up the phone and give us a call.”transcript, stated
“It is about diversifying the channels so that we can go ahead and support that increasing demand and capacity better.”transcript, stated

Fragmented systems force customers to restart their journey.

Root cause: Escalations from self-service to agents lacked contextual information.

Resolution: Use unified customer view, integrated data, summaries, and backend integrations so context follows the interaction.

“there's no contextual information that accompanies the escalation.”transcript, stated
“It forces the customer to start their journey all over again. That is not a frictionless experience, and that's absolutely something that we need to improve upon.”transcript, stated

Manual business-unit contact flows do not scale.

Root cause: Each business unit previously had manual flows and many routing profiles, driving maintenance cost and slow seasonal changes.

Resolution: Automate common contact flows and use proficiency-based routing.

“Our biggest friction point was having contact flows that are manual against each business unit.”transcript, stated
“we went ahead and automated 101 contact flows across all business units”transcript, stated

AI contact-center architecture fails without integrated data.

Root cause: Personalized service and a 360-degree customer view require backend and data integration.

Resolution: Connect knowledge bases, APIs, SharePoint, monitoring, metrics, and data integration to the AI/contact-center stack.

“unless you have your data integrated, you won't be able to go ahead and maintain or sustain this architecture.”transcript, stated
“integrate with our backends to provide that 360 degree view that we are looking at.”transcript, stated

Evaluation

Canada Life evaluated the transformation with contact-center operating KPIs, migration resiliency metrics, AI capability metrics, customer feedback, and forecasted savings.

Metrics: 94% wait-time reduction · 10% average handle time reduction · $7.5 million saved in first half of 2025 · 92% reduction in average speed to answer · 18-second average speed to answer · 2-4% abandonment-rate reduction · 47-second AHT reduction · 25-second after-call-work reduction · 37% automated authentication · 83% chatbot containment · 1,900 calls deflected per week · 80% positive customer feedback · $43 million projected five-year savings

4 supporting quotes
“We saved 7.5 million in the first half of 2025. We had a 92% reduction rate in our average speed to answer. And in this year, in Canada Life, in our contact centers, our average speed to answer is 18 seconds for the whole year.”transcript, stated
“So first is we reduced our after, sorry, our AHT per call by 47 seconds. Now, there are compounding capabilities that came together for that 47 seconds, but there was one capability that stood really tall.”transcript, stated
“Canada Life has been in this AI journey steady state for two months, and in this two months itself, we have seen after-call work reduction of 25 seconds, our authentication, automated authentication goes ahead and has taken care of 37% of authentication.”transcript, stated
“Along with that, we have seen 80% of customer positive feedback responses. That tells us how happy customer is with the chatbot that we have recently built.”transcript, stated

Outcomes

The program serves a company with 14 million customers, touched 21 business units, 150 toll numbers, 14 major deployments, 38 teams, over 2,000 agents, 101 automated contact flows, 400 annual agent hires, and 1,900 weekly call deflections.

3 supporting quotes
“So we had a 94% reduction in wait time, a 10% reduction in average handle time.”transcript, stated
“And in this year, in Canada Life, in our contact centers, our average speed to answer is 18 seconds for the whole year. We reduced also our abandonment rate by 2 to 4% depending on the business unit.”transcript, stated
“We are looking at a reduction of 200,000 call volume, that will result into a cost savings of $43 million.”transcript, stated