TheJakartaPost
Please Update your browser
Your browser is out of date, and may not be compatible with our website. A list of the most popular web browsers can be found below.
Just click on the icons to get to the download page.
Popular Reads
Top Results
No results found. Please check your search term and try again
Can't find what you're looking for?
View all search results
Popular Reads
Top Results
No results found. Please check your search term and try again
Can't find what you're looking for?
View all search results‘Vibe Coding’: How Asia’s finance teams are rewriting rules
AI coding agents allow finance teams to build solutions without needing an engineering background, as they only need to define the logic and desired outputs, while the AI writes the code, tests it and connects it to relevant data sources.
Change text size
Gift Premium Articles
to Anyone
Share the best of The Jakarta Post with friends, family, or colleagues. As a subscriber, you can gift 3 to 5 articles each month that anyone can read—no subscription needed!
A message reading “AI artificial intelligence,“ a keyboard and robot hands are seen in this illustration created on Jan. 27, 2025. (Reuters/Dado Ruvic)
‘Vibe coding’ is now firmly a part of mainstream business language. But understanding what that means in practice, how it applies to a heavily regulated function like finance and what the next evolution holds are more complex questions.
Finance functions have traditionally been cautious about adopting new technologies. Since financial outputs flow directly into audited results, there is little room for error. Precision, traceability and data security are paramount.
At Boston Consulting Group (BCG) we are now witnessing a major shift across the industry. AI is becoming ubiquitous across finance teams and wider corporate functions.
Enter vibe coding. This term describes a new wave of AI-driven custom application development that promises faster development and more efficient processes, if the right guardrails are in place.
Traditionally, end-user computing required IT support or complex spreadsheet engineering. Today, finance teams across the Asia Pacific region are using cloud-based AI models to build their own tools using natural language. They are learning which problems are best solved with AI, creating a credible path to build targeted applications without waiting for long IT development queues.
This unlocks a powerful new opportunity for finance functions in Indonesia and beyond.
Promise of efficiency
Organizations have invested heavily in new enterprise software for decades, inspired by the promise of enhancing efficiency and shifting talent towards more strategic roles. Yet, paradoxically, finance departments have continued to grow.
Vibe coding presents a unique opportunity to break this cycle by rapidly upskilling finance professionals to use AI coding agents.
Automating workflows and building custom tools on the fly allows analysts to drastically cut the manual effort required to keep the department running. This efficiency gain frees up the finance team to focus on interpreting data, managing risk and driving the corporate value agenda forward.
How Vibe Coding works
AI coding agents allow finance teams to build solutions without needing an engineering background. The analyst defines the logic and desired outputs, while the AI writes the code, tests it and connects it to relevant data sources.
We are already seeing compelling, practical use cases within Indonesia.
First, answering questions from source data. A large Indonesian financial services company recently connected an AI coding agent directly to its database. Operating within strict, read-only limits, the agent translates plain language questions into charted answers in minutes. The analyst retains full judgement, automating only the data retrieval.
Second, building internal tools rapidly. A global Indonesian investment company now uses AI agents to build working tools in weeks instead of months. Analysts describe their required dashboards in natural language. These applications sit on top of the system of record and run securely in sandboxed environments to avoid the risk of creating undocumented ‘shadow code’.
Crucially, these custom applications do not replace traditional finance systems. Core platforms for accounting and planning remain essential. AI acts as an augmentation layer. It sits on top of the system of record, making insights easier to interpret without compromising the underlying data.
Added value
Vibe coding provides a doorway to a wide range of process enhancements.
Finance teams can deploy these tools to handle tasks requiring interpretation, localized insight or anomaly detection. When examining how coding agents add value, the improvements over traditional baselines are striking across multiple finance functions.
In general accounting, teams historically relied on manual sampling or static rules within their enterprise resource planning (ERP) systems. Analysts often reviewed only flagged items or trial balance rollups.
Coding agents change this dynamic entirely by generating Python scripts on the fly. These scripts scan full transaction data for duplicates, round-dollar entries, off-hours postings and outlier amounts. This approach creates full-population testing instead of manual sampling. It quickly surfaces hidden issues that remain invisible at the trial balance level.
Order-to-cash workflows see similar transformative benefits, particularly in the area of revenue leakage detection. Previously, teams depended on spot audits or annual revenue assurance reviews. Now, AI seamlessly cross-checks contracts, billings and shipping records. The agent instantly flags under-billed orders, missed price escalations and unbilled services. By surfacing these systemic pricing and contract gaps, companies recover vital margin previously lost to billing errors.
The value extends into financial planning and analysis (FP&A). Past forecasting involved tedious spreadsheet rebuilds and manual driver updates, which led to fragmented scenario analysis. Coding agents now build localized forecasting tools split by geography or sub-business unit. This targeted approach improves forecast accuracy and delivers sharper insights for regional analysts.
Finally, treasury teams can revolutionize their short-term cash position forecasting. A treasury analyst might typically spend one to two days per cycle rebuilding static, rolled-forward spreadsheets. An AI agent simply pulls live accounts receivable and payable ageing, alongside historical payment behavior. It models expected daily inflows and outflows automatically. As a result, treasury teams generate accurate 13-week forecasts in minutes and significantly reduce idle cash.
Setting the right guardrails
While the accessibility of vibe coding is empowering, it must be carefully managed. These tools still require human review before outputs feed reporting. A next evolution is the widespread adoption of ‘agentic AI’, which uses predefined structures to autonomously execute a task. Assign it a high-level objective, such as reconciling accounts, and the system independently navigates software, manages the complete workflow, and delivers a final output.
Organizations need to be proactive in avoiding ‘AI sprawl’. A CFO who lets their team build tools without boundaries risks trading shadow spreadsheets for shadow code. Undocumented scripts operating outside official systems can cause chaos, with this shadow code incredibly difficult to spot until something breaks.
An informed shift towards agentic AI will be pivotal. In this near future, the true value will lie in verification, security and continuous monitoring of agents, rather than just the code generation itself.
Leaders must establish clear guardrails to scale safely. In Indonesia, these guardrails should be closely aligned with OJK's IT and data-risk rules, as well as the Personal Data Protection Law No. 27/2022. Governance must be embedded directly into how these tools operate, restricting custom AI applications to analysis and exception-handling in controlled environments, while core financial reporting remains safely within strictly audited platforms.
What finance leaders should do
The opportunity is real, but realizing the value of AI coding requires disciplined execution. At BCG, we have identified some crucial next steps to capture this opportunity.
One, build a review checkpoint, not just a review function. Every analyst-built tool that touches financial data should pass through a verification step before its outputs are trusted. That means additional human review of logic as well as compliance.
Two, make ownership and documentation non-negotiable. Every custom tool needs a named owner and a record of what it does and what data it touches. The difference between an asset and shadow code is whether others can find and understand it.
Three, draw a hard line at the system of record. Let analysts build freely for analysis and exception-handling, but keep AI-generated code out of anything that writes directly into audited financial reporting.
AI coding agents give finance teams a fast, credible way to build the applications they have always needed. By empowering teams with these capabilities and enforcing the right guardrails, organizations can achieve better productivity, faster insights and a leaner, more agile finance function.
*****
Tushar Agarwal is a managing director and partner at Boston Consulting Group (BCG), while Daymas Dipo is a principal at BCG.
