How Ziina Cut Regression Testing Time by 90%
How Georgii Troshin, Senior Software Engineer, Mobile Platform & Quality at Ziina, automated 69% of their E2E suite and empowered the team to double their release cadence.
By Jason Steinberg · August 2026 · 5 min read
What Ziina does
Ziina is a UAE financial platform for consumers and businesses, designed to make everyday financial activity simpler and more accessible. It provides a fast and secure way to spend, receive, and manage money, combining award-winning design with products built for everyday financial needs.
Its ecosystem brings consumer and business financial products together within a single platform, spanning payments, wallets, cards, and business financial services.
Built in the UAE around how the region actually manages money, Ziina is licensed by the Central Bank of the UAE (CBUAE). Its mission is to enable financial freedom across the Arab world.
Before Autosana
Ziina had no dedicated QA engineers. Instead, the team rotated a release engineer role among mobile developers. The assigned engineer manually verified the entire application before each release, stepping through every flow and screen. The process required roughly 12 hours of active effort spread across two business days, interrupted by meetings, context switches, and other responsibilities.
For the engineer on rotation, feature development stopped entirely.
Georgii, who owns testing infrastructure across Ziina's mobile teams, requires every engineer to complete the full manual regression cycle before contributing to automation. The experience is intentional.
It's motivation. This is how I try to motivate people, to help me, to fix problems, to automate more things.
One engineer recently needed three full days to complete their first cycle, requiring constant coordination with other team members to understand features they had not built.
Before evaluating Autosana, the team considered native testing frameworks including Espresso and XCTest. They never adopted them. The maintenance cost was evident before they committed. Every minor UI modification broke deterministic test scripts. Engineers needed to add accessibility identifiers throughout the codebase. Verifying a single code change required full application rebuilds.
It's totally opposite from native tests. In native tests, you need to redo everything [after changes]. You need to program it because it's very explicit.
How Autosana changed the workflow
Instead of rigid scripts tied to element IDs, Autosana's agent interacts with the application the way a user does. When a developer adds a button, shifts a layout, or changes a label, the agent adapts without requiring test modifications.
It's not a deterministic tool, which is very good because then it gives you flexibility, and then the flow will succeed even if it's with minor changes.
Three capabilities made the integration effective at Ziina's pace:
Hooks for test setup. Ziina uses Autosana's hooks to programmatically pre-create test data and configure account preconditions before flows execute. The test environment is fully prepared when the flow begins, eliminating repeated setup time from every run.
Nightly regression runs. The full regression suite executes every morning before the team begins work. Four critical smoke tests run on every pull request. By the time engineers start their day, test results are already available.
Failure analysis with AI. After a run completes in GitHub Actions, Georgii reviews failures using Codex, which extracts annotated screenshots, compares them against previous successful runs, and delivers a structured summary of each failure.
You can basically delegate this work of analysis, checking, verification, and this last step of failures to the agent, focus on something more important, and then come back and see the results.
What Autosana catches
Autosana delivers a level of precision that manual testing cannot sustain across a 12-hour regression window.
It's helped us to find some invisible bugs. They're not really invisible, but … so hard to notice them by eye.
Running the full suite every night has also exposed infrastructure problems that sequential manual testing never reached. A broken Intercom SDK credential, for example, was caught during a routine nightly run. Issues like this now surface weekly.
Results
| Metric | Before Autosana | With Autosana |
|---|---|---|
| Active manual testing time | ~12 hours across 2 business days | Approximately 1 hour |
| Full regression run time (automated) | No automation | 1 hour 20 minutes including build |
| Time to detect a broken flow | At release testing, up to 2 weeks later | Same day |
| E2E automation coverage | None | 69% (targeting 80%) |
Across both measures, manual regression effort dropped by roughly 90%.
Of the test cases not yet automated, most can still be converted to Autosana flows. The remainder depends on physical hardware, such as tap-to-pay flows with real cards. Georgii's target is 80% automation coverage, with the final 20% requiring manual verification by design.
Impact on engineering velocity
What previously required two full business days of manual effort now runs overnight, with approximately one hour of manual review the following morning. Georgii describes the current state as relaxed mode: low intensity, low risk, with the automated suite handling the majority of verification before the team starts their day.
The testing bottleneck that constrained Ziina to a bi-weekly release cycle is dissolving. Without the combined effect of automated regression and reduced manual overhead, transitioning to a weekly cycle would not be feasible.
We can release as many times as we want, because it's very safe and easy for us.
Multiple engineers are now writing and updating flows independently. The objective is full team autonomy, where any engineer can expand test coverage for new features without relying on a single person.
It's helped us to dramatically decrease manual effort and accelerate our release process. It almost completely eliminated manual testing for us, as well as it helps you to find bugs, old issues, new issues, very early on if you run this often.
Get started
Get started at autosana.ai.
Prefer a walkthrough first? Book a demo.