practically each trade — together with finance. The truth is, the monetary sector has lengthy been an adopter of what we now name “conventional machine studying,” utilizing it for predictive modeling, credit score scoring, and threat analytics.
However with the present hype round Massive Language Fashions (LLMs) and Agentic AI, I begin to query: how can this trade really make the most of this new know-how? In contrast to many different sectors, finance operates beneath strict rules, information privateness guidelines, and governance buildings — circumstances that, for me, appear a bit contradictory to the autonomy idea of Agentic AI.
To fulfill that curiosity, I attended the Agentic AI for Finance Convention [1] held in Jakarta, Indonesia, on October 16, 2025. The occasion — organized by Algoritma Knowledge Science College [2] — introduced collectively main practitioners from banks, insurance coverage corporations, fintechs, authorities establishments, and AI startups in Indonesia to discover how Agentic AI may reshape the monetary sector.
Though the discussions centered totally on Indonesia, most of the insights mirrored the challenges and alternatives confronted by the broader panorama — from Southeast Asia and even the worldwide finance trade. On this article, I’ll share my key takeaways, insights, and reflections from the occasion.
Disclaimer: I’ve no affiliation with any of the businesses or organizations talked about on this article. They’re referenced purely for readability and illustration. For extra details about them, please consult with the references part on the finish of this text.
1. ROI of AI
Each time an organization plans to start out a brand new undertaking or undertake a brand new know-how, the query of Return on Funding (ROI) inevitably arises. That is pure as a result of to start out one thing new — particularly with AI — the funding required to construct or combine such programs isn’t small. Measuring how a lot monetary return the corporate will get in alternate for that funding is crucial.
Nonetheless, within the case of AI adoption, precisely measuring ROI will be difficult. This occurs as a result of more often than not, when an organization integrates an AI resolution into present operations, it’s usually tough to isolate the worth generated purely by AI. The affect is normally distributed throughout a number of groups and applied sciences, making attribution removed from simple.
One strategy I discovered from audio system to higher perceive that is to have a look at how organizations undertake AI at completely different maturity ranges:
Boosting productiveness
Reaching technical excellence
Enhancing or creating income streams
By taking a look at these ranges, we all know that ROI doesn’t at all times seize the complete image of AI’s affect. Many organizations at the moment are starting to enhance ROI with Return on Worth(ROV) [3], a extra holistic strategy that measures not solely monetary return but additionally goals to reply questions like:
Did AI enhance determination high quality?
Did it improve buyer satisfaction?
Did it enhance inside productiveness?
One other equally necessary perspective is the Price of Inaction (COI). This represents the potential losses an organization incurs by not adopting, or by delaying the adoption of, AI. Based on Forbes [4], there are 4 key areas the place an organization will face challenges if it chooses to “wait and see”: widening information gaps, issue attracting prime AI expertise, missed studying alternatives, and rising operational inefficiencies in comparison with opponents which have already turn into AI-enabled.
To conclude, whereas ROI will at all times stay a elementary part within the decision-making strategy of AI adoption, corporations want to enhance it with different views equivalent to Return on Worth (ROV) and Price of Inaction (COI) to seize the complete image of AI’s affect and strategic significance.
2. Problem as Regulated Sectors
The second key takeaway that I wish to focus on right here is safety.
As I discussed earlier, one main distinction between monetary and different sectors lies in the strict rules and excessive degree of information safety. For example, the Indonesian Monetary Providers Authority (OJK) requires banks to maintain their information facilities and catastrophe restoration facilities inside Indonesia’s borders [5].
As the results of this regulation, monetary establishments can’t freely undertake cloud-based programs. They need to be certain that all information stays safe and follows native rules. That’s the reason many organizations desire to run their AI programs on on-premise or hybridinfrastructures relatively than relying totally on the cloud.
As well as, with the variety of information breaches and phishing assaults, the monetary sector should additional strengthen their cybersecurity frameworks. One of many audio system emphasised that when integrating AI, the whole lot have to be safe and compliant earlier than deployment — or it’s higher to not deploy in any respect. The price of failure on this space, whether or not monetary or reputational, will be far higher than the price of delay.
3. Agentic AI in Motion
We’ve mentioned two most necessary elements earlier than implementing AI within the group. Now, let’s see a few of the use-cases of Agentic AI which were talked about by audio system of the occasion.
Humanless Monetary Reporting
Historically, monetary analysts depend on giant and numerous inputs — equivalent to market costs, firm filings, and information sentiment — to construct their evaluation. This course of calls for each velocity and accuracy, because the monetary panorama adjustments quickly.
With Agentic AI, this workflow will be reimagined. By connecting to dependable and real-time information sources, AI brokers with completely different specializations (e.g., market analysis, firm reporting and information, historic information evaluation, and report designer) collaborate autonomously to generate concise and data-driven report.
Determination-makers or analysts can merely ask query in pure language, after which multi-agent system orchestrates these specialised brokers to ship the report in PDF or slides format inside seconds.
As well as, for recurring reporting duties (equivalent to every day or weekly updates), the consumer can schedule the system to generate the report with the newest market information.
In my opinion, the reliability of the information supply is essentially the most essential issue on this use case. To make the report reliable, we can not let the specialised agent make up the information — what we referred to as as hallucinating. So as an alternative of utilizing their very own coaching information, we have to feed them with a curated, verified dataset.
Particularly for market evaluation functions, platforms like Sectors.app present a listing of API endpoints that may be accessed by AI brokers to retrieve precise market information. Utilizing a trusted supply equivalent to this — or some other verified sources — helps reduce hallucinations, guaranteeing accuracy whereas enhancing analyst productiveness.
A few months in the past, I began my journey on Agentic AI utilizing Sectors.app API and OpenAI Brokers SDK. I create a easy Streamlit app that enables customers to work together with AI Brokers and ask questions associated to corporations listed on Indonesia Inventory Change (IDX).
I’ve revealed this undertaking on my GitHub, and the hyperlink to the repository will be discovered on the finish of this text.
AI-powered Audit Course of
One dialogue that stunned me got here from The Audit Board of Indonesia (BPK) — the federal government establishment chargeable for auditing the administration and accountability of state funds. What caught my consideration was how far they’ve already gone in adopting AI options inside their operations, one thing I didn’t fairly anticipate from a authorities physique.
In collaboration with Supertype, BPK built-in AI options to their BIDICS platform, remodeling an enormous quantity of audit paperwork right into a queryable information base to help the audit course of — from looking out, analyzing, and visualizing key information contained in studies [7].
They leverage the LLM for information extraction and categorization of paperwork, in addition to for producing preliminary analytical insights. The AI-driven insights help BPK auditors in planning, threat evaluation, and decision-making earlier than conducting detailed subject audits [8].
One of many key challenges talked about by the speaker was guaranteeing that the information entry is restricted solely to licensed auditors, to forestall any potential misuse. That is significantly necessary since BPK has authority to gather and handle a considerable amount of delicate monetary information and paperwork.
As well as, BPK maintains a human-in-the-loop strategy, that means that every one last selections should contain human oversight — an important safeguard given the affect of each motion taken by the establishment.
Studying from BPK, as one of many extremely regulated establishment in Indonesia, and seeing how they’ve adopted AI whereas collaborating with third events equivalent to Supertype and several other universities, made me understand that regulation doesn’t need to be a roadblock to innovation. As an alternative, it may function a framework that guides accountable and impactful adoption of latest applied sciences.
How AI Can Assist Notaries
One other fascinating use case of AI comes from NOTAPOS [9], a platform developed to handle end-to-end authorized paperwork and assists notaries and authorized professionals.
Authorized processes — particularly in Indonesia — are sometimes handbook and time-consuming, from handbook information entry, verification doc uploads, and so forth.
Pushed by this truth, NOTAPOS leverages AI to streamline and automate these workflows. Based on them, the platform can scale back the processes that sometimes take 18 hours to a number of days right down to simply half-hour.
Sounds magical, proper?
However that’s not the primary level I wish to spotlight right here. Within the subsequent part, I’d wish to recap what the founder shared in regards to the struggles they confronted through the growth — classes that I imagine are priceless for us to discover from.
4. When AI Strikes Too Quick
That is in all probability my favourite perception that I can gathered from the occasion when the specialists shared one other aspect of AI pleasure.
So let’s proceed the dialogue about NOTAPOS from earlier than. Again of their early interval of the event, constructing a customized mannequin that might perceive domain-specific information — on this case Indonesia’s authorized context — required handbook fine-tuning. They needed to feed the mannequin with a whole lot of authorized paperwork so it may be taught the required context and terminology.
Now, with the fast development of LLMs with bigger and extra various coaching information, a lot of this context already exists as a part of their information. Duties that beforehand required in depth handbook fine-tuning can now be completed instantly, with out extra coaching and setup, making growth turns into sooner and considerably less expensive.
This and different related instances that occurred as a result of fast development of AI increase a brand new dilemma:
Ought to we preserve constructing now, or look forward to the following large leap that may make at the moment’s efforts out of date?
On a private reflection, this query connects again to the sooner dialogue of this text — the price of inaction. In such a fast-evolving subject, ready may seem to be a approach to keep away from “pointless” value, however the true threat lies in failing behind — lacking the possibility to be taught, experiment, and adapt to the know-how because it evolves.
We can not simply wait. Because the speaker properly famous, “the hot button is to have the ability to predict the place the know-how will probably be within the subsequent six months“. It’s not at all times about chasing each new development, however about staying adaptable and positioning ourselves to grab every alternative because it comes.
5. The Final Query: Will AI Change People?
One factor I actually respect about this convention is the way it introduced collectively individuals from completely different backgrounds. This variety was mirrored in how the they responded barely in another way to the identical large query — will AI change people?
After all, this query can’t be answered by a easy sure or no. Many of the corporations imagine that AI isn’t right here essentially to interchange people, however relatively to empower them. To make sure this, organizations want to take a position closely in focused coaching applications to assist staff leverage AI — at the very least to boost productiveness.
Nonetheless, coaching alone doesn’t assure that folks will undertake AI successfully, particularly in corporations with various generations and cultures. There’ll at all times be a gaggle resistant to alter, viewing AI as a menace, or just too complicated to make use of. That is the place management turns into essential — to information and shift mindsets.
Nonetheless, if AI can carry out sure repetitive or clerical duties — in all probability higher than people — what occurs to these whose roles are outlined primarily by this work alone? One firm shared that they’ve shifted various back-office roles towards extra customer-facing positions as AI took over administrative duties. (Sadly, I didn’t get an opportunity to get extra readability on why from back-office to customer-facing roles.)
Finally, each corporations and staff share the duty to make sure steady reskillling and upskilling to remain related within the ongoing AI transformation within the office.
Conclusion
Agentic AI opens up many alternatives; boosting productiveness, reaching technical excellence, to creating fully a brand new stream of enterprise. Nonetheless, none of this may occur with out a strong basis on regulation, information safety, infrastructure, and human readiness.
On this article, we’ve mentioned a few of my key learnings from the convention. I’m grateful for the possibility to see how completely different organizations are reacting to this new period of transformation.
After all, this isn’t the tip of this journey. The truth is, elements of what we mentioned right here quickly to be not related. Nonetheless, that’s the truth of AI’s fast evolution. But, we can not afford to easily wait and see.
Yow will discover the supply code and my early studying milestone on this GitHub repository. On this undertaking, I used OpenAI Brokers SDK, Sectors.app API, and Streamlit to construct a easy interactive monetary app that enables customers to work together with AI Brokers and ask questions associated to corporations listed on Indonesia Inventory Change (IDX).
