Every private bank and family office has a GenAI roadmap in 2026. Far fewer have asked the harder question first: what is that AI actually trained on?
The OECD has now answered it. In its new report, Artificial Intelligence and Open Finance (OECD AI Papers No. 61), the intergovernmental body representing 38 economies delivers a clear verdict: Open Finance — not the model — is the foundational data enabler for AI in finance. Data is, in the OECD's own words, “the bloodline of AI innovation,” and Open Finance is “the connective tissue enabling data flows and interoperability across the financial sector.”
While wealth executives race to launch copilots, autonomous portfolio analytics and generative advisory tools, the report is unambiguous on one point: these systems are only as good as the data feeding them. Model performance depends on the quality, structure and real-time accessibility of a consolidated, multi-custody wealth data foundation — not on which LLM sits on top of it.
Data First, Algorithm Second: Why Wealth Data Aggregation Has to Come First
Building LLMs or advisory agents on top of fragmented, unstandardised custodial data doesn't just underperform — it creates operational overhead and compliance exposure that compounds with every new custodian added. High-performing financial AI needs structured, high-frequency, fully reconciled inputs. Establish a clean multi-custody aggregation layer first, and every downstream model — today's and next year's — inherits accurate, hyper-personalised portfolio intelligence by default.
It's the same principle behind our own webinar with co-founders Álvaro Morales and Joaquim de la Cruz: Data First, AI Second. The OECD just gave that principle a 38-country policy stamp.
AI-Ready Data Governance: The Bank of Korea Blueprint
To separate market reality from vendor hype, the OECD highlights the Bank of Korea (Box 2, p.14) as the operational benchmark for AI transformation. When the central bank set out to launch its domain-specific AI platform, it found that data architecture built for human inspection simply couldn't support machine-learning workflows.
To close that gap, the Bank converted over 1.4 million internal documents into structured, machine-readable JSON — complete with enriched contextual metadata, semantic ontologies and automated pipeline lineage. That's what “AI-ready” actually means in practice, and it's the exact operational shift wealth management C-suites now face:
Traditional vs “AI-ready” data governance
The takeaway for wealth institutions: without clean, structured, continuous data streams sourced across every custodian, AI models generate inaccurate analytics and compliance vulnerabilities by design.
Proof at Scale: What Open Wealth Infrastructure Delivers
Skeptical this pays off commercially? The OECD's analysis of Brazil's Open Finance framework (Box 3, p.17) puts numbers behind it. In the first half of 2025 alone:
- 35 million clients actively used account-aggregation solutions built on Open Finance data
- BRL 14 billion in transactions were generated through 2 million tailored investment proposals
- 18 million accounts received automated, real-time alerts to optimise fund performance
- Account opening timelines fell from 32 hours to just 2 hours and 10 minutes
Mastering multi-custody data aggregation isn't a compliance cost centre — it's a client-retention and product-origination engine, at national scale.
The Non-Negotiables: AISP-Grade Security for Multi-Custody Portfolio Reconciliation
Integrating AI with broad financial data sharing raises the security bar, not lowers it. The OECD is explicit: scaling data pipelines has to be matched by strict data protection, explicit consent management and zero-trust architecture. In practice, that means four things your infrastructure partner should already have:
- Regulatory licensing: operating as a licensed Account Information Service Provider (AISP), supervised by a central banking authority — in Flanks' case, the Bank of Spain under PSD2.
- Industrial-grade encryption: RSA and AES-256 dual encryption protecting credentials and assets at rest, in transit and in use.
- Independent certification: SOC2 Type II and SOC3 audits providing continuous, third-party assurance.
- Read-only data access: Data Feeds and Power of Attorney (PoA) mechanisms that create automated, read-only flows, removing transactional execution risk entirely.
Key Strategic Takeaways
- OECD confirmation: OECD AI Papers No. 61 names Open Finance the primary data enabler for AI in financial services.
- Data first, always: model performance depends on clean, standardised, aggregated financial data — not on the algorithm layer alone.
- “AI-ready” governance is the bar: raw custodial files need to become structured, machine-readable schemas before any wealth AI copilot can be trusted.
- Proven at scale: Brazil's Open Finance rollout drove engagement across 35M+ active clients and billions in tailored investment proposals.
- Security is the foundation, not an add-on: AISP status, SOC2 Type II, AES-256 encryption and read-only Data Feeds are what make AI-driven wealth data safe to scale.
Frequently Asked Questions
What does the OECD say about AI and Open Finance?
In OECD AI Papers No. 61, the OECD concludes that Open Finance is the foundational data enabler for AI in financial services: it supplies the richer, more diverse and interoperable datasets that AI models need for training, fine-tuning and hyper-personalised output.
What is “AI-ready” data governance?
AI-ready data governance transforms raw, human-readable financial records into structured, machine-readable formats — complete with contextual metadata, semantic ontologies and pipeline lineage — so both humans and AI systems can interpret and act on the data consistently.
What is multi-custody portfolio reconciliation?
It's the automated process of aggregating, cleaning, validating and enriching portfolio data sourced from multiple custodian banks into a single, reconciled source of truth — the prerequisite for any accurate downstream AI analytics or advisory output.
What is an AISP?
An Account Information Service Provider (AISP) is a regulated entity authorised under frameworks like PSD2 to access financial account data on a customer's behalf, under strict, read-only, consent-based conditions and central-bank supervision.
Accelerate Your Wealth AI Strategy
If your institution is evaluating or scaling a generative AI roadmap, start here:
--> Gen AI Strategic Guide for Advisors: Navigating GenAI Projects in Wealth Management: a decision-making framework for CTOs, CISOs and Heads of Strategy on architecture, vendor compliance and ROI.
--> On-demand webinar, “Data First, AI Second”: Flanks co-founders Álvaro Morales and Joaquim de la Cruz unpack why leading private banks and family offices build multi-custody data infrastructure before deploying AI algorithms.
Access the whitepaper:
About Flanks
Flanks is a wealth management technology company (wealthtech) that is redefining the industry through automation and data-driven insights. Its modular and all-in-one solution empowers global financial institutions, including banks, family offices, asset managers, pension plan providers, and technology companies, to offer faster, higher-quality, and personalised advice by transforming complex and fragmented wealth data into valuable insights.
Flanks was founded in 2019 in Barcelona by Joaquim de la Cruz, Sergi Lao, and Álvaro Morales, former Global Head of Santander Private Banking. Currently, the company aggregates data from 600+ connections with global financial institutions and processes more than 500,000 portfolios per month in over 33 countries, managing assets worth more than €39 billion. For more information, visit flanks.io.
.png)
.png)
.png)
.png)