Gemini Code Assist Cut Their Dev Time by 37%. (ComplyAdvantage)

ComplyAdvantage sits in an uncomfortable position in the financial crime space. Its platform has to process sanctions updates within minutes of publication, scan global news feeds continuously, and surface risks that compliance teams at banks would otherwise miss entirely. The product is genuinely impressive. The engineering challenge of keeping it that way is less glamorous: a codebase that grew fast, original developers who moved on, and documentation that never kept pace with the code itself.

By the time ComplyAdvantage was ready to ship a major new product offering, that technical debt had become a real drag. New engineers and contractors couldn’t navigate the codebase without pulling senior developers away from building. Senior developers spent chunks of their time explaining code instead of writing it. On a fixed delivery deadline, that’s a compounding problem with a predictable outcome.

Running Their Gemini Code Assist Pilot

What makes the ComplyAdvantage rollout compelling is how carefully they measured the impact. Rather than rolling out Gemini Code Assist to everyone and asking people how they felt about it afterward, they ran a structured pilot with 20 senior developers across three global development centers. Jellyfish tracked development lifecycle metrics directly from Jira and GitLab throughout the pilot, giving them an objective comparison between Code Assist users and everyone else. That methodology matters. A lot of developer productivity claims are based on surveys. This one is based on commit data.

The pilot group improved median development time by 42%. When the rollout expanded to the full 170-person engineering org, the figure settled at 37% across the board. Commits per developer up 50%. Merge requests per developer doubled. The new offering shipped on the original timeline.

The documentation problem also got addressed, somewhat incidentally. Developers started using Gemini Code Assist for what the team calls “brownfield code navigation”, asking it to explain existing code, trace how components connect, identify dependencies. The institutional knowledge that had been locked inside the heads of a handful of long-tenured engineers became something any developer could surface on demand. That’s a different kind of value than raw productivity, and arguably a more durable one.

Support, Too

Engineering wasn’t the only function that changed. ComplyAdvantage ran more than 500 support requests through Gemini to automatically triage them into genuine support issues versus bugs requiring engineering attention. Spot-check accuracy came in at 75%. More importantly, root-cause identification for customer-reported bugs went from hours of manual investigation to minutes. Jennifer Bursack, Sr. Director of Technology Operations, described one specific instance where a developer found a bug’s root cause in minutes that would previously have taken hours to track down. Multiply that across hundreds of support cases and you have a meaningful shift in how quickly customers get resolutions.

Why the Infrastructure Choice Matters Here

ComplyAdvantage is a multi-cloud company. They didn’t default to Google Cloud because of inertia. They chose it for this workload specifically because of data sovereignty requirements, security posture, and confidence in the infrastructure. This is a company that processes sensitive financial data for regulated institutions, and the banks and financial services firms they sell to ask hard questions about where data goes and who controls it. Running Gemini Code Assist on Google Cloud gave ComplyAdvantage answers they could actually give to those customers.

That context changes what the 37% productivity number represents. It’s not just a speed improvement. It’s a speed improvement that happened without relaxing any of the security or compliance constraints that make selling to regulated institutions possible in the first place. That combination is harder to achieve than it sounds.

What Comes Next

The Code Assist rollout was phase one. ComplyAdvantage is planning to fine-tune Gemini models for customer-facing features and extend AI-assisted root cause analysis further into the support workflow. The developer productivity story opened the door. The customer experience story is where this goes from internal efficiency gain to product differentiation.

For GCP sellers working the fintech space: ComplyAdvantage sells to the same banks and financial institutions that are evaluating their own AI infrastructure decisions. A fintech that built on Google Cloud, measured the results publicly with real commit data, and attributed a successful product launch to Gemini tooling is a reference that travels well in those conversations. The fact that they’re a multi-cloud shop who still chose Google Cloud for this is the part worth leading with.

For ISVs considering Gemini Code Assist as part of their own stack: the benchmark here is useful precisely because it wasn’t run by Google. ComplyAdvantage designed the measurement methodology, chose Jellyfish as the tracking tool, and published the numbers as part of a public case study. The 37% figure has provenance. The 42% pilot result shows the ceiling when you start with experienced engineers who adopt the tool quickly. And the doubling of merge requests per developer is probably the most interesting number in the set, because it suggests developers aren’t just working faster on the same tasks. They are completing more discrete units of work per cycle, which compounds across a 170-person org in ways that a simple time-savings estimate doesn’t fully capture.

Want to go deeper?