Agentic AI in Finance: Opportunities and Challenges for Indonesia The financial sector has long embraced data-driven technologies, from predictive modeling and credit scoring to risk analytics. Yet, with the rise of Large Language Models and Agentic AI, a new wave of transformation is unfolding—one that promises to redefine how financial institutions operate. However, this shift brings unique challenges, especially in regulated environments like Indonesia. To explore these dynamics, I attended the Agentic AI for Finance Conference in Jakarta on October 16, 2025, organized by Algoritma Data Science School. The event gathered leaders from banks, fintechs, insurance firms, government bodies, and AI startups across Indonesia. While focused on the local context, the insights reflected broader regional and global trends in AI adoption within finance. One of the most pressing questions in AI adoption is Return on Investment (ROI). While financial metrics matter, measuring AI’s impact is complex because its benefits are often distributed across teams and systems. Many organizations now supplement ROI with Return on Value (ROV), which considers non-financial outcomes such as improved decision-making, faster processes, and enhanced customer experiences. Equally important is the Cost of Inaction (COI)—the risk of falling behind competitors who are already leveraging AI. Delaying adoption can lead to talent loss, knowledge gaps, and operational inefficiencies. Regulation remains a major hurdle. Indonesia’s Financial Services Authority (OJK) mandates that banks keep data centers and disaster recovery systems within the country’s borders. This restriction limits cloud-based AI deployment and pushes institutions toward on-premise or hybrid infrastructures. Security and compliance are non-negotiable—any AI system must be secure before deployment, as failures can result in severe financial and reputational damage. Despite these constraints, Agentic AI is already making inroads. One compelling use case is automated financial reporting. Instead of analysts manually compiling data from market feeds, filings, and news, AI agents with specialized roles—market research, data analysis, report generation—can collaborate autonomously. Users simply ask a natural language question, and the system delivers a polished report in seconds. For recurring reports, scheduling ensures up-to-date outputs. The key to trust here is data quality: agents must be fed verified, real-time data to avoid hallucinations. Platforms like Sectors.app offer reliable APIs for market data, making this approach viable and accurate. Another powerful example comes from the Audit Board of Indonesia (BPK), which has integrated AI into its BIDICS platform. In collaboration with Supertype, BPK transformed vast volumes of audit documents into a searchable, analyzable knowledge base. LLMs extract and categorize data, generate preliminary insights, and support risk assessment—while maintaining a human-in-the-loop model to ensure accountability. This shows that even highly regulated institutions can innovate responsibly. NOTAPOS, a legal tech platform, uses AI to automate document management for notaries. What once took days can now be done in 30 minutes. But the journey wasn’t easy. Early on, the team had to manually fine-tune models with legal documents. Today, with more advanced LLMs, much of that context is already embedded, reducing development time and cost. This rapid progress raises a critical dilemma: should we build now, or wait for the next breakthrough? The answer lies in adaptability. As one speaker put it, “The key is to predict where the technology will be in six months.” Waiting may seem safe, but the real cost is missing the chance to learn, experiment, and lead. Finally, the question of whether AI will replace humans remains complex. Most organizations see AI as a tool to empower, not replace. But adoption depends on culture and leadership. Training is essential, but so is mindset change. Some roles are shifting—back-office staff are being retrained for customer-facing positions, reflecting a broader trend of reskilling. In conclusion, Agentic AI offers immense potential in finance—boosting productivity, enabling innovation, and creating new value. But success depends on a strong foundation: compliance, data security, infrastructure, and human readiness. The technology evolves fast, and the window to act is now. The future of finance in Indonesia—and beyond—won’t be defined by the AI we have today, but by how we learn, adapt, and lead in the face of change.