Insurers are under increasing pressure—from rising claims and evolving risks to growing regulatory complexity and heightened customer expectations. Traditional software development lifecycles, often measured in months, can no longer keep pace.
In this episode of CGI Conversations, insurance and AI leaders from CGI discuss how artificial intelligence is transforming the entire software development lifecycle (SDLC). From requirements and architecture to development, testing and operations, AI is helping insurers deliver change faster while strengthening quality, resilience and compliance.
This episode is designed for insurance leaders across P&C, Life, Group Benefits, Commercial, and Specialty lines, including CIOs, CTOs, digital leaders, and heads of delivery, architecture, QA and transformation.
- Chapter 1: Introduction and Industry Context
-
Introduction
Derek Marinos (Host):
Hello and welcome to a CGI Conversation about AI and emerging technology in Canada. I’m your host, Derek Marinos.
Today, we’re exploring how AI is transforming the software development lifecycle in the insurance industry. Insurers are under intense pressure, claims are rising, risks are growing, regulations are becoming more complex, and customer expectations keep climbing. Traditional development cycles simply can’t keep pace.
AI is changing that by accelerating delivery from months to weeks, improving quality, strengthening compliance, and reshaping how insurers build and operate technology.
Joining me today are three CGI experts: Tarun Dehariya, Industry Insurance Lead for Canada; Chris Juryn, Head of AI and Emerging Technologies for Canada; and Guillaume Brincin, expert in AI and immersive technologies.
- Chapter 2: Why Traditional SDLC Models Are Breaking Down
-
Why traditional SDLC models are no longer sustainable
Tarun Dehariya:
Insurers are getting squeezed from all sides, and there are several factors driving this.
First is the business environment. It’s moving much faster than it used to. A major storm can hit and suddenly claims volume rises through the roof, while insurance systems are expected to scale instantly. Fraud is getting smarter because of increased online interactions, and regulations keep changing, IFRS 17, privacy requirements, and new AI guidelines. Models, rules, and reports need to be updated far more frequently.
Second is customer expectation. We want quicker answers and better digital experiences. We expect products tailored to our needs, such as usage-based insurance or real-time underwriting. But many insurers are still running on legacy systems that weren’t designed for rapid change.
Third is competition. Insurtechs and large technology players are raising the bar with new digital experiences. At the same time, cyber risk is increasing and talent is in short supply.
When you put all of this together, long, linear development cycles simply don’t fit anymore.
Speed, risk, and why long cycles no longer work
Chris Juryn:
You can’t wait 18, 12, or even six months to make changes anymore. It’s like steering a ship where you turn the wheel, but the rudder reacts months later.
Technology is changing at lightning speed. New models are constantly emerging, risks are evolving, and competition is moving fast. While long release cycles can feel safer, they actually compound risk. The longer organizations wait to release, test, and learn, the more uncertainty and exposure they introduce.
End-to-end SDLC thinking
Guillaume Brincin:
The SDLC is an interdependent chain, not a collection of independent activities.
If you improve development productivity by 20 percent using AI, developers will produce code faster, but if QA remains manual, testing becomes the bottleneck. Even if testing is accelerated, documentation can quickly become the next constraint.
Each phase produces outputs that become inputs for the next phase. That’s why partial AI adoption is risky. The real opportunity lies in transforming the entire flow so gains compound across the lifecycle.
- Chapter 3: AI Impact and Real-World Results
-
Real-world ROI from AI in the SDLC
Tarun Dehariya:
What’s compelling is the impact we’re seeing across the insurance value chain. AI is removing much of the manual, repetitive work from the SDLC, updating rules, checking underwriting logic, reviewing long documents, and maintaining test cases.
When AI is introduced into these steps, work that once took days or weeks can be completed much faster, allowing teams to focus on decisions that require human judgment.
Some large insurers have rolled out tools like GitHub Copilot to their developers, where AI acts like a pair programmer, helping write code and suggest fixes. Others are using AI-powered testing platforms that automatically create and update test scripts and regression steps when UI changes occur.
At CGI, we’ve seen this impact firsthand. For a large life insurer, we modernized testing and QA using our NAVI tool. Test preparation time dropped by about 45 percent, approximately 58 to 60 days were saved across 11 projects running in parallel, and test coverage improved to between 70 and 85 percent. QA teams shifted from repetitive checking to automation and improvement.
In another case, we built a digital claims portal that went live in six weeks instead of five months. That kind of speed matters during severe weather events, regulatory changes, and when customers expect faster service.
It’s not just about speed. AI helps catch gaps earlier and apply rules more consistently, improving quality, while humans remain responsible for review, validation, and final decisions.
How AI is transforming documentation, QA, and architecture
Guillaume Brincin
Each phase of the SDLC leverages AI differently.
For documentation, AI maintains living documentation that updates as code changes, ensuring developers always have current information. This makes maintenance easier, deepens system understanding, and accelerates onboarding.
For QA, AI generates comprehensive unit tests, standardizes test cases, and automates execution across the codebase. Tasks that once took weeks can now happen continuously, helping teams catch bugs earlier and maintain consistent quality.
Architecture remains underutilized today. Many organizations haven’t integrated AI into architectural design and planning, making it a weak link and a potential bottleneck. When every phase stabilizes AI workflows, a multiplier effect emerges, with superior inputs compounding gains across the lifecycle.
- Chapter 4: Scaling, Governance, and Adoption
-
Scaling and governance
Chris Juryn:
When executives see these results, the next question is how to scale. Proofs of concept are easy; scaling across teams, product lines, and lines of business is much harder.
True ROI comes from compressing the entire SDLC, not just one part of it. Faster delivery increases decision velocity, teams can test, learn, and adjust more quickly. But scaling also exposes governance gaps. Applying old governance models at scale can introduce risk.
Organizations need clear guardrails, including human-in-the-loop validation, approval processes, and clarity around data usage and residency. It’s also critical to invest in people, not only delivery teams, but also risk, legal, and compliance functions, because the way of working is fundamentally changing.
How insurers should start: the first 90 days
Chris Juryn:
The first 90 days shouldn’t be about proving that AI works. What matters is proving that it delivers value.
Insurers should define success metrics upfront, such as reduced testing time, fewer defects reaching production, or time freed for exploratory testing. Teams should identify bottlenecks across the SDLC and remove them in a way that creates repeatable, scalable processes.
Guardrails must be in place from day one, around data usage, privacy, compliance, and model risk, so early success doesn’t introduce downstream issues. After 90 days, organizations should have a repeatable approach that can be rolled out across teams and lines of business.
Mindset and adoption
Tarun Dehariya:
The biggest shift isn’t about technology, it’s about mindset. Teams need to move from seeing AI as a threat to seeing it as a helper.
When a testing team sees AI generate most of their regression test cases, or when a claims team sees AI extract key details from long documents, fear drops quickly. Starting small, showing value, and building confidence allows organizations to scale responsibly.
Chris Juryn
It’s also important to highlight early success stories. Bringing forward people who are getting value from AI and sharing their experiences helps normalize adoption and reinforces the behaviors organizations want to scale.
Avoiding silos
Guillaume Brincin:
Starting small is the right approach, but it has to be done strategically. Pilots should establish stabilized workflows, templates, standards, and quality controls that can be replicated.
Design pilots for interoperability from day one. Consider what the next phase needs from this one. Prove one standardized workflow end-to-end, then replicate it across other phases so they connect seamlessly. That’s how experiments become scalable transformation.
Reliability, resilience, and customer experience
Chris Juryn
Traditional systems fail in predictable ways; AI systems can fail probabilistically. Outputs may vary, which creates new challenges for support teams, model drift, data quality issues, and changing system behavior.
If organizations accelerate delivery but slow down recovery and support, customer experience suffers. Customers want fast change and a great experience that works every time. That means balancing speed with resilience, explainability, and operational readiness.
- Chapter 5: The Future of AI-Driven SDLC
-
What’s next: AI agents and orchestration
Guillaume Brincin
We’re moving from handoffs to continuous intelligence. AI agents will orchestrate workflows autonomously, coordinating specifications, stakeholder validation, and transforming requirements into development stories, with humans intervening at critical approval points.
Model Context Protocols enable tools to exchange not just data, but context and capabilities. This creates a fluid ecosystem where information flows naturally and humans focus on high-value decisions while AI manages execution.
When speed becomes the norm
Tarun Dehariya
When delivery acceleration becomes normal rather than a differentiator, the advantage shifts to how insurers use that speed.
If two insurers can release changes weekly, the winner will be the one that can roll out new coverage, such as flood coverage after a major event, or improve claims journeys based on real customer feedback. Strategy shifts from delivering quickly to delivering better outcomes continuously and safely.
Looking ahead to 2027
Chris Juryn:
When we look ahead to 2027, the biggest advantage will come from insurers that use AI to continuously adapt and update their software and operations together. It won’t be one or the other. If organizations focus only on delivery speed without improving operational resilience, it won’t make a meaningful difference for customers or the business. The winners will be those that can build faster, recover faster, and respond faster, using AI to balance speed with resilience.
Tarun Dehariya
The biggest impact from AI will show up in advisor and customer experience. Insurers are already asking how they can improve both. We’ll see more agentic AI quietly handling background work, reading documents, pulling data from different systems, and preparing cases for underwriting or claims. Instead of advisors spending days chasing paperwork, AI will do most of the preparation, allowing advisors to focus on explaining options and delivering real advice. For customers, that means quicker answers, fewer back-and-forth interactions, and smoother journeys, with the right guardrails and human oversight in place.
Guillaume Brincin
Another major advantage will be in architectural design and technical decision-making. This phase is currently underutilized, but it will become a breakthrough area. AI agents will increasingly generate system architectures, evaluate trade-offs, and simulate implementation impacts, creating cascading improvements in quality across every downstream phase of the SDLC.
Derek Marinos (Host):
AI isn’t just speeding things up, it’s changing how software gets built across the entire lifecycle. Thanks to Tarun, Chris, and Guillaume for the discussion, and thank you for listening.
