A major U.S. insurer is scaling AI across the software development lifecycle

Summary

Sector: Insurance

The challenge: Following early adoption of AI-powered development tools, a leading U.S. insurer aimed to scale its investments and turn them into measurable value. The open questions were where AI could realistically be applied across the software development lifecycle (SDLC), leveraging each tool’s strengths, how to reach sustained adoption across hundreds of engineers, what governance to put around deployment, and how to harvest tangible benefits for the organization.

The solution: The insurer partnered with Wavestone to build a structured approach set the direction, get buy-in from senior management on business cases, implement key capabilities, and demonstrate first return on investments prior to scaling.

Key results:

Approximately 30% average time savings per developer
First measurable value in approximately 4 business days
Over 100 employees trained across the organization
An estimated $3-5M in annual cost avoidance across selected future initiatives

30% average time savings per developer
100 employees trained across the organization
$3-5M in annual cost avoidance

The challenge

Moving beyond AI experiments

The insurer had made the decision to invest early in AI development tools, but struggled to answer fundamental questions:

Where and when should AI be deployed across the SDLC?
How do we satisfy enterprise security, compliance, legal and sourcing requirements once tools are selected?
How do we measure improvement across the SDLC?
How can these improvements be converted into tangible gains?

Successful AI-enabled SDLC programs require more than technology deployment. They require a transformation framework that connects AI initiatives directly to business objectives and measurable outcomes.

The process

Six workstreams that accelerated AI for SDLC transformation at a major U.S. insurer

Wavstone worked with the insurer to define and launch six workstreams. The design reflected the organization’s regulatory environment, sourcing model and engineering footprint. The workstreams have proven portable to other large enterprises, though the emphasis across them varies for each organization.

The program was anchored on evidence from the field. A 12-week pilot involving more than 30 participants evaluated GitHub Copilot and Amazon Q, alongside MS 365 Copilot and SeaLights. Every recommendation was validated against what those teams experienced.

The six workstreams

1. Strategy & planning

Established a transformation roadmap aligned to business priorities.

Use cases were mapped across the software development lifecycle, scored on value and feasibility, and sequenced into an actionable roadmap. This enabled leadership to focus investments on the highest-impact opportunities.

2. Value measurement

Created a fact-based framework to prove AI’s business impact.

Productivity baselines were established before deployment and measured through delivery metrics, surveys and activity data. This allowed benefits to be quantified rather than assumed.

3. Governance & program oversight

Provided transparency and executive oversight throughout the transformation.

Regular reporting, business-case tracking and escalation mechanisms helped leadership monitor adoption, manage risk and accelerate decision-making.

4. Tool readiness & validation

Evaluated AI solutions against security, compliance, legal, sourcing and operational requirements.

The AI solutions that were assessed included GitHub Copilot, Amazon Q, MS 365 Copilot and SeaLights. The findings were turned into deployment readiness recommendations resulting in stronger tool assessments, comparative evaluations, adoption dashboards, and readiness recommendations.

5. Change management & adoption

Drive sustainable adoption of new ways of working.

Over 100 employees were trained and ran 20 knowledge handover transfer sessions with 10+ internal stakeholders across 30 deliverables. The insurer could scale autonomously after the end of the project.

6. Pilot and rollout support

Support delivery teams during implementation.

Worked with the pilot teams for 12 weeks while continuously collecting feedback, and turning what worked into reusable templates and business cases. It resulted in pilot frameworks, use case templates, feedback loops, and rollout business cases.

Where AI was used day to day

The primary use cases deployed across development teams were code generation, unit test generation, refactoring, performance optimization, debugging, code understanding and documentation. The value came from the combination rather than from any single capability.

The results

From productivity gains to business outcomes

For this insurer, value main value opportunities were found in four phases: planning, requirements, development and testing. The program focused there rather than transforming the full lifecycle at once. Wavestone has seen a similar concentration in other large engineering organizations, though the mix depends on tooling maturity and where delivery bottlenecks sit.

Productivity was only the starting point. The insurer converted those gains through two pathways.

Pathway 1: Competitive advantage through operating leverage

AI allowed the insurer to do more with its existing internal teams:

Approximately 30% average time savings per developer
First measurable value in approximately 4 business days of experimentation, which removed time-to-value as an objection early
More than 80 qualified business use cases identified
Over 100 employees trained, with positive adoption momentum measured through Net Promoter Score

Instead of reducing headcount, the insurer redirected capacity toward architecture modernization, technical debt reduction, innovation initiatives, complex problem solving and faster product delivery.

Pathway 2: Cost optimization through reduced delivery costs

AI also created opportunities to optimize external delivery spend:

An estimated $3-5M in cost avoidance across selected future initiatives
Emerging opportunities for 10-30% savings during IT services contract renewals
Involving managed service providers from the beginning to help understand lessons learned from their teams directly

As productivity improvements matured, the insurer redesigned vendor relationships through contract rebasing, scope optimization, gain-sharing models, and improved delivery efficiency. The result is measurable financial impact that reaches operating margins and increased the focus on outcome based delivery model rather than capacity based.

What this means for other organizations

Deploying AI coding assistants was only the starting point. The critical work is to build a structured transformation program that systematically improves how software is planned, developed, tested and delivered.

Planning, requirements, development and testing are also the phases agentic tools are already reaching, and that changes what drives value. It stops being individual developer speed and becomes orchestration: how work is decomposed across agents, how output is verified before it reaches production, and where human judgment stays in the loop. Review capacity becomes the constraint rather than authoring capacity.

When AI systems execute multi-step work with less supervision, the organizations that benefit already know how to measure output, govern deployment, manage the token cost of running agents at scale, and drive adoption across large engineering populations. Left unmanaged, that compute spend erodes the savings the program was built to deliver. The capability built for assistive tools is what makes agentic tools safe to scale, which leaves organizations still running pilots behind before the shift arrives.

Every wave of AI in software delivery moves the bottleneck somewhere new. The next move is toward specification and verification. The advantage belongs to organizations that spot where the constraint has moved and reorganize around it first.