Skip to main content
Indotek Digital Nusantara

AI & Its Impact

How Ready Are Indonesian Banks to Govern AI? What a New Study Found

A whitepaper by PERBANAS, IBM and K&A, based on 91 respondents from 35 financial institutions, puts AI governance maturity in Indonesian banking at 2.74 out of 5 and names data privacy as the top AI risk. What it found, the limits of the figures and what banks and suppliers can prepare.

Glass office towers in Jakarta's Sudirman Central Business District under a blue sky, with palm trees and a building entrance in the foreground
Office towers in the Sudirman Central Business District (SCBD), Jakarta, photographed on 2 April 2018. The picture is illustrative; no institution in the study is shown. Photo: Rantemario, CC BY-SA 4.0 (opens in a new tab)

How ready are Indonesian banks to govern the AI they are starting to use? A study launched in Jakarta on 27 October 2026 suggests the honest answer is: not yet. The whitepaper, Digital Sovereign AI and the Future of Trusted Banking, was prepared by the banks' association Perhimpunan Bank Nasional (PERBANAS) with IBM and K&A. Based on 91 respondents from 35 financial institutions, it puts the average maturity of AI governance at 2.74 out of 5, and finds that only 4 per cent have expanded AI across functions under formal governance. For bank boards, risk teams and their technology suppliers, the figures are a useful benchmark, as long as their limits are understood.

What the study measured

The findings were reported by Kontan, Fortune IDN and IDN Times on 27 and 28 October, IndoTelko on 30 October, Investor.id on 1 November and Medcom on 2 November. We did not read the whitepaper itself; the figures below are as these outlets reported them. Kontan, IndoTelko and Fortune IDN give the sample as 91 respondents from 35 financial institutions in Indonesia. Most reports call the respondents bankers; Fortune IDN refers to bank executives.

  • Adoption: 34 per cent were still at the pilot or limited-trial stage, 18 per cent used AI for some operational needs, 12 per cent had not yet explored AI, and 4 per cent had expanded AI across functions with formal governance.
  • Inventory: 9 per cent had a complete inventory of AI use that is updated regularly.
  • Governance maturity: an average of 2.74 out of 5, covering documentation, validation, performance monitoring, periodic evaluation and human oversight.
  • Digital sovereignty: 73 per cent said it is relevant to banking transformation in the next 12 to 24 months, with an average relevance score of 3.98 out of 5. Fortune IDN frames the 73 per cent finding differently, as executives saying the main requirement of transformation concerns AI ownership and digital sovereignty.

Privacy tops the list of AI risks

Asked which AI risk most needs to be controlled, 82 per cent named data privacy, well ahead of cybersecurity at 44 per cent and regulatory risk at 41 per cent. Investor.id and IDN Times also report explainability at 21 per cent and auditability at 12 per cent. Kontan headlined the finding as the risk of data leakage.

Awareness of the rules varied. Of the 91 respondents, 54 named the personal data protection law (UU Pelindungan Data Pribadi), 49 named POJK 11/POJK.03/2022, 33 named SEOJK 29/SEOJK.03/2022, and 43 said they knew OJK's AI governance guidance for banking. Medcom and Investor.id note that OJK issued that guidance in 2025 and followed it with PADK OJK No. 1/2026 on the use of information technology by commercial banks. In other words, fewer than half of respondents reported knowing the guidance written specifically for AI in their sector.

Sovereignty means control, not only data location

According to the reports, the study treats digital sovereignty as more than the question of where data is stored. In an interactive poll, 70 per cent of participants chose full control over data, systems and technology as the best definition; the reports do not say whether the poll drew on the same 91 respondents. Medcom, Investor.id and IndoTelko describe five areas of control: infrastructure and technology; data; the AI model lifecycle; applications and decision-making; and governance, visibility and accountability. IDN Times lists the same ground split into six items.

The whitepaper introduces a five-layer control framework and a Digital Sovereign AI Readiness Model, and recommends inventories of AI use, model monitoring, clearer accountability and capability building. Training in AI governance was the support most requested, chosen by 66 of the 91 respondents; Kontan adds readiness assessments and help building inventories of AI use and models.

Toto Prasetio, head of technology and security at PERBANAS, said the industry's priority is a shared understanding of what digital sovereignty means in practice, and that closer alignment between banks, regulators, associations and technology partners is key to responsible AI. Catherine Lian, general manager of IBM ASEAN, described AI as fast becoming a core pillar of modern banking, used in customer interaction, fraud prevention and decision-making. Juvanus Tjandra, managing partner of IBM Consulting Indonesia, told Fortune IDN that digital sovereignty is not achieved through a single product, platform or regulation.

How to read the numbers

  • A vendor is a co-author. IBM is a technology company, and Lian's remarks referred to its work in AI, hybrid cloud and security. That does not make the findings wrong, but readers should weigh them with it in mind.
  • The sample is small and self-reported. The reports do not say how respondents were selected. Investor.id notes that answers differed by role: governance staff stressed accountability, compliance staff control over data, and technical staff data foundations.
  • The adoption figures reported add up to 68 per cent; none of the reports we read accounts for the rest.
  • The percentages some outlets attach to the scores, almost 80 per cent and about 55 per cent, are simple conversions of 3.98 and 2.74 out of 5, not separate measurements.

What banks and their suppliers can do now

This section is our analysis, not legal or regulatory advice.

  • Build the AI inventory first. If only 9 per cent of institutions have a complete one, most cannot yet list every model, vendor AI feature and internal tool they run, who owns it, what data it touches and which decisions it informs.
  • Map each use to the rules that apply. For every item in the inventory, record which of the data protection law, OJK's IT rules and its AI guidance are relevant, and who is accountable.
  • Design for privacy from the start. Decide before launch what customer data each AI use may see, where it is processed and how access is logged, rather than reviewing it afterwards.
  • Keep evidence after launch. Documentation, validation, performance monitoring, periodic evaluation and human oversight are the areas the study scored; each can be built into release and operation rather than added at audit time.
  • Suppliers should expect sharper questions. If banks define sovereignty as control over models, data and infrastructure, a statement about data location alone will not answer them.

What to watch next

Watch whether PERBANAS publishes the full whitepaper with its method, and whether its readiness model is taken up by banks as a common yardstick. Watch whether OJK adds to its AI guidance for banking, given that fewer than half of respondents reported knowing it. And watch whether a repeat survey shows the 2.74 maturity score, the 9 per cent inventory figure and the 4 per cent with formal governance moving.

Sources

  1. Risiko Kebocoran Data Jadi Tantangan Utama Adopsi AI di Perbankan · Kontan(opens in a new tab)
  2. Privasi Data Dinilai Jadi Risiko Adopsi AI Bagi Bank di Indonesia · Fortune IDN(opens in a new tab)
  3. Riset IBM Ungkap Tantangan Besar AI di Perbankan RI · IDN Times(opens in a new tab)
  4. Privasi data jadi risiko AI terbesar bagi perbankan Indonesia · IndoTelko(opens in a new tab)
  5. IBM dan Perbanas Soroti Pentingnya Kedaulatan Digital dalam Transformasi Perbankan · Investor.id(opens in a new tab)
  6. Perbankan Indonesia Prioritaskan Kedaulatan Digital dan Tata Kelola AI · Medcom.id(opens in a new tab)

This briefing was prepared by the Indotek editorial desk with AI assistance, from the public sources listed above. It is general information, not legal or professional advice.

AI & Agentic AI

How Indotek can help

A governance score of 2.74 out of 5 and a 9 per cent inventory rate describe AI that is in use but not yet documented, measured or controlled. Building AI with those controls designed in is the work Indotek does in AI.

  • Start from the decision: we define the business decision or task an AI feature serves before choosing a technique, which gives each item in an AI inventory a clear purpose and owner.
  • Evaluation before scale: every AI feature we ship comes with an evaluation set, so validation and performance monitoring produce evidence rather than assumptions.
  • Humans in command: our agentic systems act inside explicit permissions, log what they do and hand off to a person when confidence is low or the stakes are high.
  • Intelligent search with access controls: we build retrieval over enterprise knowledge that returns answers with sources and respects existing access controls, so AI sees only the data a user is allowed to see.

Let's build what comes next

Whether you are modernising a legacy environment, building an AI-powered product or exploring a new digital business, let's create technology that delivers lasting value.