Entelect x Sanlam - Customer Success Story

How Entelect helped Sanlam modernise a decades-old retirement platform with Claude Code

Building digital products that work in the real world – across complex architectures, and shifting user expectations – requires more than good design. At Entelect, we cover the full stack of digital product engineering: from strategy and user research through to channel engineering, platform integration, and customer engagement. We build products that are structured to scale, with cross-functional teams committed to continuous development and ownership.

Summary

How Entelect helped build 1,980 automated regression tests in three months

Coronation Fund Managers needed comprehensive regression coverage for a project to replace its legacy enterprise data management system with a new EDM platform. By combining a reusable automation framework with a governed, AI-assisted engineering workflow, Entelect helped the team deliver 1,980 automated regression tests in three months.

At a glance
Client	Coronation Fund Managers
Industry	Financial services
Partner	Entelect
Scope	Replacement of a legacy enterprise data management (EDM) system with a new EDM platform, including custom configuration and development and integration with internal and external systems
Technology	Claude Code with Sonnet 4.6 and Playwright MCP, with Opus 5 evaluated for more complex tasks
Headline result	1,980 automated regression tests delivered in three months

The Challenge

Modernising a critical securities data platform

Coronation manages long-term client capital across multiple asset classes. Its investment, trading, reporting, compliance and operational teams depend on accurate, consistent and timely instrument and securities data. The platform is the central source of that data for front office, middle office and back-office teams, which makes its consistency and reliability critical to the day-to-day running of the business.

The scope of the project was to replace Coronation's legacy EDM system with a new EDM platform offering greater functionality. The new platform was a commercial off-the-shelf product that already provided much of what Coronation required, but delivery still involved custom configuration and development, as well as integration with several other internal and external systems. Work ran over approximately two years, managed through sprints with functionality released incrementally, which made comprehensive regression testing increasingly important.

Once the configuration, custom development, and integration work were tested and in place, the team needed a regression test pack. Writing and maintaining those tests manually would have been slow and difficult to sustain, and a single manual regression cycle would have occupied a team of testers for weeks while still covering considerably less ground.

The Approach

Combining AI acceleration with engineering discipline

Build the foundation first
Entelect first established a structured automation framework with reusable page objects, coding standards and maintainable engineering patterns. This gave the team a consistent foundation before Claude Code was introduced, and ensured that AI was applied within an established, governed workflow rather than being expected to solve the automation challenge on its own.

Use AI to accelerate repeatable work
Claude Code with Sonnet 4.6 helped the team turn test ideas, screen flows and existing manual regression knowledge into usable automated tests. Supported by reusable prompts, shared engineering skills and Playwright MCP, it was used to generate functional and workflow-based regression tests, discover application screens, identify locators, improve consistency across the suite and extend coverage across the priority screens and workflows selected for automation.

Keep engineering judgement in control
As adoption expanded, the team added governance around prompts, skills and AI-generated outputs. Designated ownership, controlled branching, peer review and mandatory validation ensured that Claude Code accelerated repetitive automation work without replacing engineering judgement, review discipline or quality control.

Select the model for the task
Sonnet 4.6 was selected as the primary model because it offered the right balance of quality, speed and token efficiency for repeatable, high-volume regression automation. The team also evaluated Opus 5 for more complex tasks. Its deeper reasoning capability showed potential, but it was more token-intensive for the volume of automation being delivered.

The experience reinforced a practical lesson about task-based model selection. Rather than assuming that a larger model is always the better choice, the team found that the appropriate model depends on the nature and complexity of the task. Sonnet 4.6 suited generating and extending tests, applying established framework patterns and analysing application screens and workflows, while Opus and other models continued to be explored for selected, more complex activities.

The Result

Delivering months of regression automation in 3 months

The impact was most significant in the regression cycle itself. QA cycle time decreased by 93%, from more than two weeks to roughly five hours, while manual execution labour was reduced by 100%, with the full automated suite now running unattended on a Sunday evening.

Within three months, the team delivered 1,980 automated regression tests, focused on the 25 priority screens that mattered most to Coronation’s business needs rather than attempting to cover the entire application. This provided materially broader and more repeatable regression coverage than the previous manual approach, which would have occupied a team of testers for weeks while still covering considerably less ground.

The approach also proved portable. It was subsequently reused on another Coronation front-end application upgrade, where the established framework patterns, prompts, skills and governance practices enabled meaningful automated coverage to be established within the first week.

Measure	Result
Automated regression tests	1,980 delivered in three months
Regression coverage baseline	No regression test pack existed before the project; the automated suite is the first one
Regression cycle time	A manual cycle would have occupied a team of testers for weeks; the full automated suite now runs unattended on a Sunday evening in roughly five hours
Screens automated	25 priority screens automated, identified as the highest priority for Coronation's business needs
Regression coverage	Materially broader and more repeatable coverage than a manual regression cycle would have achieved
Reuse across applications	Meaningful coverage established on a second Coronation front-end application upgrade within the first week

This was an initiative we drove together. Our team set the direction and Entelect brought the testing discipline to make it real: Shiven built a repeatable model rather than a one-off test pack, drawing on the intellectual property our own team had already built up. The proof was how quickly he stood the same capability up on an upgrade project shortly afterwards. On the back of that, we have adopted an automated regression pack as a standard deliverable on every project, which is a far more valuable outcome than a set of tests for a single system.

Sarah Harber Head of Project Office, Coronation Fund Managers