The conversations at Financial Forum Innovations this year made one thing unmistakably clear: the banking sector is no longer debating whether to adopt AI - it is grappling with something far more consequential. How do you build a new financial infrastructure without undermining the trust that the entire system depends on?
The most forward-looking voices in the room converged on a shared framework: AI is a powerful but ultimately supporting technology. The real transformation happening beneath the surface is structural - a fundamental rewiring of how money moves, how assets are held, and how trust is established in a digital economy.
From automation to infrastructure
Banks have long used machine learning for fraud detection, credit scoring, and risk management. What’s different now is the emergence of large language models that reshape how institutions interact with customers, documents, and regulation. The technology cycle that once spanned several years has compressed to months - and that pace demands a different posture.
The real challenge for banks is not to be the fastest to deploy AI, but to be the institution clients trust most with AI acting on their behalf. The vision of autonomous agents negotiating offers, executing payments, and managing assets is closer than it seems - and it raises a question the forum addressed directly: when AI makes a consequential financial decision, who bears responsibility?
This is not merely a legal question. It is an architectural one. Institutions that deploy AI without meaningful audit trails, explainability layers, and human override mechanisms are not just exposed to regulatory risk - they are building on foundations their own management cannot fully defend. AI is changing how leadership thinks, not just how operations run. That shift demands governance frameworks, not just technology roadmaps.
The identity layer
One theme that ran through multiple sessions deserves particular emphasis: digital identity. As AI agents increasingly act on behalf of clients - querying offers, initiating transactions, interfacing with third parties - the question of who authorized this action becomes critical.
This is the third pillar of European financial sovereignty, and arguably the least mature. The digital euro is anticipated. Instant payments are operational. But the state’s ability to digitally verify its own citizens - and banks’ readiness to serve fully digital clients - remains a work in progress. Until that infrastructure is reliable and interoperable, the promise of autonomous finance carries an identity-shaped hole at its centre. Solving it is a prerequisite, not a follow-on.
The sovereignty question
Running through the entire forum was a thread about European financial sovereignty. Bulgaria’s entry into the eurozone brings this into sharp focus: the question is no longer about membership in the European financial architecture, but about how to use that membership strategically.
The digital euro, DLT-based tokenization of assets, and instant payments are not separate initiatives - they are different expressions of the same transformation. Getting this right means maintaining the primacy of central bank money as the ultimate anchor of trust, even as private digital money, tokenized deposits, and stablecoins proliferate. Innovation at the edges, stability at the core.
For enterprise technology providers, this framing matters enormously. Sovereign infrastructure is not just a political concept - it is a technical requirement. Data residency, computational governance, and the ability to audit AI systems are becoming baseline expectations for any institution operating in regulated European markets.
The quantum dimension
One underappreciated theme raised at the forum concerned post-quantum cryptography. The security assumptions underlying today’s financial systems - asymmetric encryption, digital signatures, secure key exchange - are on a trajectory toward obsolescence as quantum computing matures. Institutions that begin preparing now will have a significant structural advantage; those that wait will face a disruptive transition under pressure.
What the new stack demands
These discussions point toward a clear set of requirements for the next phase of financial services. Infrastructure must be sovereign by design. It must be AI-native without being AI-opaque - explainability, human oversight, and audit trails built in from the start, not retrofitted. It needs a credible digital identity layer anchoring the trust chain from citizen to agent to institution. And it must be interoperable across digital money, tokenized assets, and distributed ledgers.
Sirma’s perspective
This is precisely the domain where Sirma operates. With deep roots in financial technology - spanning core banking interfaces, regulatory reporting, and compliance automation - and a sovereign AI platform purpose-built for enterprise environments, Sirma brings a combination that is increasingly rare: genuine fintech domain expertise paired with the ability to deploy AI on-premise, within a client’s own infrastructure, without any data leaving the organization’s control.
Solutions on the Sirma AI Enteprise platform address the regulatory complexity that participants highlighted throughout the forum - the sheer volume of rules, formats, and reporting obligations that already burden compliance teams, and that will only grow as digital asset regulation matures. Moreover, the Sirma.AI Enterprise platform directly answers the accountability challenge raised on stage: when AI is deployed within your own infrastructure, with full audit capability and governance controls, the question of who bears responsibility has a clear and defensible answer.
As the financial sector moves from isolated AI experiments toward integrated, production-grade deployments, the institutions that will lead are those that can navigate the intersection of technology ambition and regulatory reality - where sovereignty, identity, and accountability are not obstacles to innovation, but the conditions that make it sustainable.
That is the conversation Sirma is built to have.
Visit sirma.ai to explore our platform further.
