8 September 2026
Data Sharing, Automation and AI | Report
The financial sector is no longer just sharing data, it's starting to act on it. Here's where that shift stands in 2026.
For most of banking's history, a single institution owned the customer, the transaction, the data and the margin. That integrated model is coming apart. Data sharing, process automation and AI-based decision systems are now in production across the sector, and the direction of travel is clear: from open banking, to open finance, to open data.
A maturing market
The most advanced ecosystem is the United Kingdom, where a centralised model pushed the largest banks to open their data. By December 2025 it had reached 16.5 million user connections (up 36% year over year) and processed 351 million payments across the year (up 57%). By July 2026, the cumulative total had passed 1 billion payments and 100 billion API calls since launch.
Europe tells a different story: no single harmonised metric, multiple API standards, and interoperability held together in large part by private aggregators built on a public obligation to share data. Across the EU and UK, 547 authorised third-party providers were active by mid-2026, a figure that has held steady. The number of players has stopped growing, but transaction volumes keep rising. This is consolidation, not stagnation.
Pressure on the traditional model
The paradox is striking. In 2024, global banking generated roughly $1.2 trillion in profit, the highest of any sector, yet it still carries the lowest price-to-book ratio. Record profits today sit alongside a deep discount on tomorrow.
Challengers are capitalising on the split. Revolut closed 2025 with about $6 billion in revenue (up 46%) and a 38% pre-tax margin; Nu Holdings reported $16.3 billion in revenue and a 33% return on equity. Meanwhile, a quieter risk looms: if AI agents make it easy to move even a small share of the $23 trillion sitting in near-zero-interest accounts, global banking profits could fall by as much as $170 billion over the next decade.
AI: broad adoption, limited depth
Adoption is widespread but shallow. Of organisations surveyed, 81% already use AI in some form, yet 41% remain in pilot, 26% are scaling, and only 14% have reached the point where AI is reshaping their operating model.
Where it works, the value is real. JPMorgan identified roughly 450 AI use cases and opened its internal platform to over 200,000 employees. Klarna's AI agent handled the workload of 853 full-time staff and cut average resolution time from 11 minutes to under 2. McKinsey estimates generative AI could create $200–340 billion in annual value for the sector. The main obstacle to scaling isn't technical, it's explainability: a model that denies credit and can't say why is unusable where the rationale is a legal requirement.
Four regimes, one operation
A single credit decision built on open finance data and assessed by an AI model can fall under four regulatory layers at once, PSD3/PSR, the future FIDA framework, the EU AI Act and DORA. The payments overhaul reached final texts in April 2026, with full application expected around 2028, while high-risk AI obligations have been pushed to December 2027.
The strategic question for every institution is no longer whether to move, but how fast, under which governance model, and which layers of the value chain to compete in.
Want the full picture, market data, adoption maturity and the complete regulatory map? Read the complete report below.