How Worktrace AI unblocked BukuWarung AI engineers and saved weeks per workflow

BukuWarung is an Indonesian startup offering millions of micro, small, and medium-sized enterprises access to digital financial tools like bookkeeping, payments, online stores, and loans. With BukuWarung’s help, traditional small shops “warungs” have been able to digitize their operations, manage finances, and access capital.

Founded in 2019 and backed by YCombinator in 2020, BukuWarung has always been on the forefront of technology. As generative AI technologies began to evolve, BukuWarung was determined to overhaul their own approaches to continue to stay on the cutting edge.

As a transition to move AI-first, BukuWarung developed an AI team, led by BukuWarung AI Chief of Staff and equipped with AI engineers to build solutions and automations.

BukuWarung’s bottleneck in their AI-first transition was figuring exactly what to automate

CEO Abhinay Peddisetty understood that AI cannot optimize what it cannot see. He launched a company-wide exercise of surfacing AI opportunities and developing a knowledge base of how work happens at BukuWarung. This work was then to be translated into automations.

With a pipeline in place to discover AI opportunities, develop SOPs, and generate automations, BukuWarung found that their bottleneck was the first two steps. Discovery and SOP documentation is manual, requires expensive cross-functional interviewing and shadowing, and is error-prone as it often does not capture the nuanced details of how work is done. Experts in automations and experts in the work also often use different terminology, context, and communication styles, exacerbating the issues. Ultimately, discovering AI opportunities and developing SOPs consumed 3x more time compared to development, testing and release of automations.

“80% of AI engineers' time was spent in interviews, meetings, and asking questions about workflows – not in writing code.”

— AI Chief of Staff, BukuWarung

In other words: the AI team had the talent to build, but too much time was going into figuring out the real workflow details—what the process actually looked like, end-to-end—before writing a single line of automation code. This process was not only manual and error-prone, but was also not scalable, as workflows change over time, leading to drift in the SOP generation.

BukuWarung uses Worktrace AI to generate SOPs, saving weeks per automation

Worktrace AI tackles exactly the bottleneck companies like BukuWarung face in going AI first.  Furthermore, as a fintech company, maintaining strict data residency within internal systems was non-negotiable. After reviewing Worktrace AI’s agentic-first approach, lightweight architecture and enterprise-grade security posture, BukuWarung chose to deploy Worktrace over traditional non-agentic-first tools, across their operational teams.

BukuWarung uses Worktrace AI to surface and prioritize automation opportunities and generate SOPs and found it accelerates automation development by weeks per workflow. Initially, BukuWarung shadowed employees to verify Worktrace AI’s accuracy but realized that it was not needed. With Worktrace AI, the repetitive steps, time cost of each workflow and hidden dependencies on manual trackers became immediately visible for both business and AI teams.

“The workflows surfaced by Worktrace AI were 96-99% accurate, across the Operations, customer support and on-call engineering teams”

— AI Chief of Staff, BukuWarung

Now, BukuWarung AI engineers use Worktrace outputs as ground truth, which eliminates the need to follow-up with teams on questions. One engineer estimates that Worktrace AI saves him weeks of manual discovery work, per automation.

“Worktrace accelerated workflow discovery and SOP digitization for several automations from 10 days to 3 days each. With Worktrace, we don’t have to reach out to the teams every time we have a question.”

— AI Engineer, BukuWarung

The result is a cleaner, faster path from “we think this is an automation opportunity” to “we built it.” Instead of spending most of their time extracting process knowledge through interviews and meetings, the AI team can focus on execution. 

With Worktrace AI, workflow data was captured and converted into actionable insights within 1 week, dramatically shortening the discovery to automation cycle and saving engineering and product resources spent on requirement gathering. Further, BukuWarung was able to map workflows across an operations team, accelerating their automation roadmap, enabling the company to reduce 400 hours of monthly operations work down to 12 minutes a month. 

“By working with Worktrace AI, BukuWarung was able to reduce 400 hours of monthly operations work (e.g., device dispatch, logistics, activation handling) down to 12 minutes, within a few months.”

— Manager of Operations, BukuWarung

BukuWarung plans to use Worktrace AI to continuously discover more workflows to automate

In addition to generating SOPs, BukuWarung is expanding their Worktrace deployment to discover even more automation opportunities.

At BukuWarung, the transition to AI-first is not a point in time, but a transition into a new way of how operations are conducted. They plan to continue to use Worktrace AI to find and build the best automation opportunities as their own work changes and as the external AI capabilities change too. Worktrace replaced static documentation with a live, accurate representation of how work actually happens and updating workflow procedure documents automatically.

Because BukuWarung has already manually surfaced automation opportunities for many teams, they were able to validate that the prioritized opportunities surfaced by Worktrace were accurate. BukuWarung estimates that automated workflow discovery and automation has increased productivity by 40% on average and up to 90% for teams that are especially operationally-heavy.

“Worktrace AI has accelerated our discovery and automation efforts leading to increased productivity by 40% on average and up to 90% for teams that are especially operationally-heavy.”

— AI Chief of Staff, BukuWarung