91������

The Challenge

Building Enterprise-Grade Quality Without an Enterprise-Sized QA Team

Wealth Connect is not a simple CRM. It covers the full client lifecycle for financial advice firms — agreements, service delivery, financial modelling, and document generation — running on top of Salesforce Financial Services Cloud with thousands of custom components.��

With a 10-person team split between developers and functional design personnel, and no dedicated QA testers, the pressure to maintain quality across every 12-week release cycle was immense. ��The only way to get adequate coverage before a quarterly release was to pull everyone in.

"All of our BAs and business units would do a stint of UAT work. During our release cycle all of our developers would split their time 50-50 between regression testing and development. We just weren't getting the coverage we needed," mentions Pierce Wittke, CTO of Wealth Connect at GBST.��

But, this all-hands-on-deck approach still wasn't enough. Bulk regression testing consistently fell to the end of the cycle — and when it competed with bug fixes and UAT for new features, it was routinely pushed, rushed, or dropped entirely. Features that were too bug-ridden had to be descoped or pulled from a release altogether.

"With manual regression testing, releases could be delayed multiple days because we found fundamental failures we hadn't picked up earlier. We were finding solid errors at the last minute — which meant the team had to work overtime to make those release dates,” recalls Wittke.

Beyond regression, the UAT process itself was a source of friction. There was no consistent way to record and report failures. When a bug was found, developers and business users would spend significant time trying to identify whether it was a user error, a configuration issue, or a genuine defect. Without a clear record, developers spent enormous time just trying to replicate issues.

The team also faced a 70–80% nightly package build failure rate, which required an automated pipeline just to identify which incremental changes were causing issues each morning.

Faced with a perpetual manual UAT cycle and growing product complexity, GBST had two options: hire a dedicated regression testing team, or invest in a robotic testing tool. The maths made the decision straightforward: enter 91������ Robotic Testing.

The Solution

Choosing Automation Over Headcount

Faced with a choice between building a large regression testing team or implementing a robotic testing tool, GBST chose the latter.�� GBST ran a formal business case comparing multiple tools, including the Selenium-based stack used elsewhere in the GBST group for their AWS/React/Golang product. They also evaluated other commercial options. Two requirements drove the decision: technical depth for developers and business-user accessibility for UAT.

The decision was straightforward: the annual cost of 91������ Robotic Testing is roughly equivalent to the cost of one tester — approximately $100,000 AUD — but the output is easily ten times that of a single human. An effective manual regression team would have required a minimum of five people. "We chose 91������ Robotic Testing as the more cost-effective and repeatable solution," says Wittke. "It would be financially unviable to replace it with human resources."

GBST self-deployed 91������ Robotic Testing, connecting it to their Azure DevOps repository within a few weeks. The rollout was staggered: the development team came first, building regression tests and establishing the automation structure. Once that was stable, the BAs and business users were onboarded to 91������ Explorer for UAT.

CRT is now embedded directly into GBST's nightly CI/CD pipeline. Every morning, alongside the fresh beta package build, the full regression suite runs automatically. The team went from running approximately two tests per day (only during release cycles) to running 100+ tests every single day. And 91������ Robotic Testing's low-code capabilities eliminate the need for the GBST to write duplicate tests twice. “Having 91������ Robotic Testing built into our pipelines is incredibly valuable," comments Wittke.

Explorer: Bridging the Gap Between Business and Development

The 91������ Explorer tool proved to be a standout capability. Business users and BAs now record every UAT session using Explorer. When a workflow is deemed suitable for a regression test, it is flagged, and the development team converts those recordings into automated regression tests on a weekly basis — eliminating the need to write tests twice.

"The Explorer tool fundamentally changed how UAT is conducted," says Wittke. "It provides evidence of passes and a clear record when bugs are found — stored in a standard, repeatable way."

The recorded videos allow developers to immediately understand whether an issue is user error, a configuration problem, or a genuine defect — eliminating 60–70% of the back-and-forth communication that previously surrounded every bug report.

Beyond testing, the team uses Explorer as a documentation tool to record how-to guides and training videos, exporting documents with steps and screenshots stored in Confluence via Vimeo. The recordings also provide valuable insight into how end-users actually navigate the product — often very differently from how a developer would approach it. Instead of spending four hours manually writing step-by-step guides with screenshots, a team member records the screen flow in Explorer and exports it — generating 90% of the formatted documentation automatically.

The Outcome

From 2 Tests a Day to 100+ — Every Single Day

The impact on testing volume has been dramatic. Before 91������, the team might run two tests in a day — and only during the release cycle. Today, the automated pipeline builds a new beta package every night and runs the full regression suite over it, executing over 100 tests daily. This continuous testing allows the team to find bugs incrementally throughout the 12-week development cycle, rather than discovering them in a high-pressure rush at the end.

"Because the product is incrementally tested throughout the development cycle, we have more time to fix glitches and avoid major errors," says Wittke. "The increase in testing confidence has helped us meet deadlines more closely, leading to a higher quality of release and building customer confidence."

The result is a measurably better product. Clients have noticed fewer patch releases in the weeks following a major release — a direct consequence of catching and fixing issues earlier in the cycle. ��"Our customers have noticed a closer commitment to meeting deadlines and a higher quality of what's come out — and that builds confidence," comments Wittke.

Regression test coverage now stands at 45–50% of the product. The team's goal is to reach 80% automation by the end of the calendar year — and Wittke is confident that 100% is achievable. In a recent deprecation cycle — removing old code that had accumulated over four to five years — the regression suite caught a cascade of interconnected failures that no manual process would have found.

And since implementing automated testing, the stress levels during release retrospectives have dramatically dropped. "When we do our retros after a release, the stress levels, confidence and work-life balance are far more level than before,” Wittke mentions. “That makes a huge impact on the team."��

The Future

Testing the Next Frontier: Agentforce

Wealth Connect holds a distinction that sets it apart in the financial services ISV landscape: it is the first wealth ISV to be certified as an Agentforce partner. The team is now designing pre-built agent templates — slightly agnostic by design, so customers can customize them without losing the flexibility Salesforce provides — to help financial advisors gather client information, surface insights, and prepare for client interactions.

"Everything we build will be put through 91������ Robotic Testing,” comments Wittke.��

In a heavily regulated industry, the biggest concern with AI agents is repeatability. 91������ Robotic Testing is central to GBST's strategy for validating agent behaviour, because it ensures deterministic outcomes — a non-negotiable requirement for compliance-level activities. The fact that Agentforce now includes the ability to make certain agents deterministic was a major factor in GBST's decision to move forward with the technology.

"91������ Robotic Testing is how we validate that our agents behave the way they're supposed to — every time. In a regulated environment, repeatability isn't optional," Wittke assures.

‍

Standardising the Way the Business Works with 91������ AI

Looking ahead, GBST is also exploring 91������ AI — not for test automation, but to standardise business-side processes. The team is interested in using 91������ AI to bring consistency to requirements gathering, document writing, and user story structuring. Today, design teams write stories in very different ways, making it difficult to maintain a consistent standard of acceptance criteria. 91������ AI represents an opportunity to solve that problem at the source.