A data platform as the foundation
-A robust data platform increases efficiency and reduces the risks associated with adopting AI in practice, says Joakim Valderhaug, Cloud & AI Sales Lead at Orange Business Norway.
A solid data foundation, free from silos and supported by high data quality, clear governance and built-in security, reduces the risk of an organisation’s AI initiatives failing.
Valderhaug describes the data platform as a technological and organisational framework that enables data to be collected, stored, processed, made available and analysed across the organisation. It serves as the foundation for both data-driven decision-making and the advanced use of artificial intelligence.
The principle is simple but demanding: the platform must bring together technology, people and processes – from basic reporting to advanced AI.
Three categories of AI with different platform requirements
Orange distinguishes between three main categories:
- General-purpose AI (AI primarily trained on external data) – requires little access to company data.
- Role- or process-based AI (AI that replaces an internal role or process) – requires moderate to significant access to company data, integrations and platforms.
- Transformative AI (AI that combines and replaces critical roles and processes) – requires extensive access to company data, integrations and platforms.
In most sectors, including finance, the potential is greatest in the final category – and so are the associated risks. The importance of a mature data platform increases in line with the level of ambition.
Individual projects can certainly succeed without a “perfect” platform. The challenge arises when advanced initiatives need to be operationalised. As larger or multiple initiatives move into production, the need for control, security and quality in the underlying platform increases.
Valderhaug points out that an excessive focus on security can slow down development, but insufficient focus on security is not an option. A good balance between speed and security can be achieved by running the first AI initiatives on less sensitive data while, in parallel, developing a robust platform to support the more critical initiatives.
The data platform is an investment in risk reduction, not just a cost item.
Low level of maturity and high risk of failure
Market data shows that generating value from AI remains challenging. In October 2024, 74% of companies reported that they had yet to generate tangible value from AI, according to a report by Boston Consulting Group. In January 2025, only 1% of companies reported having reached full AI maturity, according to a McKinsey report.
At the same time, analyses indicate that the lack of a solid data foundation is one of the main reasons AI initiatives fail: 63% of companies lack clear data practices for AI, and by the end of 2026, 60% of AI projects without AI-ready data will be abandoned (Gartner).
There are also positive results. In 2025, 95.1% of companies reported increased AI maturity after modernising their data platforms, supporting the view that platform modernisation can deliver measurable organisational benefits (Indicium/Pure Spectrum).
In other words, the data platform is risk-reducing infrastructure, not just a cost item.
What does it mean to have AI-ready data in banking?
AI-ready data primarily means high-quality data. The process starts by defining which data is truly business-critical – from transactions and exposures to customer interactions – and establishing the quality targets that should apply. Processes, roles and tools then need to support this work through continuous quality monitoring, alerts when deviations occur and systematic error correction.
In theory, data quality is about extracting accurate data from source systems and processing it correctly. In practice, data quality needs to be embedded throughout the entire organisation – from strategy and culture to tools and platforms.
Poor data quality can have far-reaching consequences for organisations and lead to significant financial losses. There are numerous examples of businesses losing between 5 and 30 per cent of their annual revenue due to inadequate data quality. Ensuring high data quality should therefore be a priority for every organisation – not only for day-to-day operations, but also as a critical prerequisite for successful AI implementation.
What is an AI-ready data platform?
An AI-ready data platform handles traditional tasks such as collecting, storing, transforming and making data available, while also incorporating capabilities for data quality, data governance and tools relevant to AI development.
The data platform should also be based on relevant and flexible cloud technologies, making it easy to scale up or down as needed. Examples of services include Snowflake, Databricks and Azure Data Factory.
Security and Compliance in practice
In the financial sector, security and compliance should be continuous disciplines. This includes classifying data according to sensitivity, comprehensive logging and traceability of data and AI activity, strict access controls, compliance with data protection requirements, model governance with explainability and documentation, and encryption of data both at rest and in transit.
MLOps practices – including versioning of data and models, approval processes before production and continuous quality monitoring – make accountability and auditability inherent features of the solution. These practices need to be built in from day one. This makes auditability and accountability part of the solution itself, rather than a last-minute exercise ahead of an audit or regulatory review.
For banks, these mechanisms should reflect the requirements of DORA and NIS2: traceability of data and AI activity, clearly defined roles and responsibilities, incident management and third-party risk management. This also simplifies the documentation required by financial supervisory authorities.
A mature data platform reduces risk
Experience shows that organisations that combine clear governance, data quality targets and platform modernisation are better able to keep AI solutions stable in production.
Valderhaug emphasises the importance of striking the right balance.
- We should balance AI initiatives with platform initiatives. For banks, this often means planning for the modernisation of the underlying platform.
For the banking and finance sector, the conclusion is practical and measurable. A mature data platform reduces risk, accelerates the journey from idea to production and makes it easier to realise value from AI.
Investing in a data platform involves less risk than many organisations might assume, while offering significant potential benefits. The organisation’s data platform should serve as a natural hub for people, technology and processes – with generating value from data as the primary objective.
Once the data platform is in place, developing AI becomes easier, more measurable and audit-ready. That is exactly what the industry needs.
Expert contributions
Read more
Swedbank Pay – Scalable and secure payment services with Kubernetes in Azure
Swedbank Pay, part of Swedbank, is a leading provider of payment solutions with more than 35 years of experience in card acquiring. The company offers a unified platform for card, point-of-sale, and online payments, with a strong focus on security, transparency, and seamless purchasing experiences.