Transaction data in banking has never been better managed. Real-time processing, fraud scoring at scale, millisecond reconciliation: the infrastructure built to handle billions of card transactions annually is sophisticated, well-funded, and continuously refined.
The payment network knowledge that governs those transactions is a different story.
Every rule governing how a payment can be processed originates from payment networks: interchange rates, fraud liability thresholds, dispute timelines, product certification requirements, security mandates. Visa, Mastercard, and regional schemes publish thousands of pages of documentation each year: compliance bulletins, operational mandates, technical specifications, and standards updates.
This documentation arrives as PDFs distributed through scheme portals. It requires expert interpretation. It is jurisdiction-specific and affects teams across the organisation: compliance, operations, IT, finance, and product management. But it typically lands on one team's desk initially.
By Rivero's analysis, Visa and Mastercard issued approximately 25% more bulletins in 2025 than in the prior year. Compliance teams have not grown proportionately. The gap between the volume of scheme knowledge being produced and the organisational capacity to absorb it has been widening.
This is not an operational problem that additional headcount will solve. It is a knowledge architecture problem.
The gap that AI investment is now exposing
Banks across Europe are making significant investments in payments AI: predictive fraud models, automation for dispute workflows, intelligent assistants for operations teams. These initiatives share a common dependency: clean, structured, machine-readable data.
- According to RGP's 2025 analysis of AI in financial services, over 85% of financial firms are actively deploying AI, yet the report warns that AI oversight, risk management, and compliance must be embedded from the earliest stages of development, not bolted on afterward.
- The regulatory environment reflects that same pressure. Alvarez & Marsal's Q4 2025 regulatory update highlights AI governance, disclosure controls, and vendor oversight as growing compliance pressures for institutions deploying AI.
- Underneath both is a data problem. BCG's 2025 banking AI report finds that most AI failures in banking are not about the models but about slow, incomplete, or fragmented data, compounded by outdated systems and the absence of efficient data integration across environments.
Payment network documentation fits that description precisely. It is among the most technically complex unstructured data in financial services: multi-region, multi-product, inconsistently formatted, and filled with technical nuance that general-purpose language models cannot reliably interpret without expert context.
The implication is direct: an organisation cannot build reliable AI tools on top of an unstructured knowledge base. Banks that have attempted to apply machine learning to raw scheme PDFs have encountered this. The outputs are operationally unreliable and legally insufficient. The problem is not the AI, but the data foundation.
What the current model actually costs
The costs of leaving scheme knowledge unstructured accumulate across every function that depends on it.
For compliance teams, BCG's 2025 bank compliance benchmark study describes the current model as financially strained, with rising compliance costs that cannot be solved simply by adding more staff. Teams managing payment network knowledge across multiple card schemes spend disproportionate time processing, interpreting, and distributing updates.
The coordination challenge compounds when multiple networks are involved. Visa, Mastercard, other premium networks and regional schemes each have their own update cadence, their own document structure, and their own affected functions. Without a structured approach, oversight risk is not occasional; it is built into the operating model.
The enforcement context is also shifting. Visa consolidated several separate acquirer monitoring programmes into a single global framework. Card networks can impose fines for non-compliance. In some programmes, penalties range from tens of thousands to over a million dollars, depending on the violation and jurisdiction. The direction of travel for major card networks is tighter obligations, shorter response windows, and more systematised enforcement.
For product teams, the cost is less visible but equally material. Scheme updates affecting card product features, technical infrastructure, or fee structures arrive late or in filtered form. Roadmaps become reactive. Product decisions are made without the payment network intelligence that should inform them.
For IT, the pattern is similar. Compliance-driven system requirements reach technical teams after the fact. Planning windows shrink. Reactive project work and missed dependencies accumulate. These costs are rarely attributed back to the data problem that caused them.
From documentation to data asset
Organisations that treat scheme documentation as a structured data challenge, rather than a compliance processing task, gain something qualitatively different.
A structured, expert-tagged compliance dataset is queryable. It can be searched across networks, filtered by impacted functions, and cross-referenced with internal systems. It can feed the data warehouse that AI models depend on. It supports forecasting compliance workloads by mapping the flow of incoming mandates over time, rather than reacting to annual spikes.
It also changes what different teams can do independently. Product managers can access scheme updates that affect their roadmaps without waiting for compliance summaries. Finance and payment network economics teams can identify bulletins with interchange or fee implications without depending on manual distribution. IT can plan infrastructure changes ahead of compliance deadlines rather than scrambling to meet them.
The compliance function stops operating as a knowledge bottleneck and becomes a structured data source with value across the organisation.
For executives framing the investment, the context matters. BCG's analysis puts compliance costs at financial institutions under serious pressure. The question is not whether the investment in managing payment network knowledge is necessary; it clearly is. The question is whether that investment also produces a strategic data asset, or whether it simply manages a recurring process.
The organisations building this infrastructure now
The banks extracting the most value from payment network compliance today are not necessarily the ones with the largest compliance teams. They are the ones who have decided to treat scheme knowledge as a structured data asset first.
Kajo was built on that premise: transforming card network bulletins and mandates into expert-tagged, machine-readable compliance datasets that support AI-powered assistants, data warehouse integrations, and structured compliance tracking. Teams using Kajo report saving over 30 hours of manual processing per week, with manual work reductions of up to 70%. That time can instead be invested in applying their expertise and knowledge where it adds the most value.
The structured knowledge base is not an add-on. It is the infrastructure that makes everything downstream possible: proactive compliance planning and cross-functional access to scheme intelligence that was previously locked within a single team or scattered across payment network portals.
The question worth sitting with is not whether your organisation can manage scheme knowledge manually today. Most can. The question is what becomes possible once that knowledge is structured, and how quickly the compounding advantages of organisations that have done this become commercially visible.