As financial institutions push artificial intelligence deeper into everyday operations, the harder question is not what the technology can do but how firms need to change to work with it. Maximilian Groth, chief executive and co-founder of the confidential-computing and data-collaboration company Decentriq, argues that the first thing to break is the approval chain, and that the firms getting the most from AI are rarely the ones deploying the most of it.
The debate about AI in financial services has largely been a debate about capability: what models can do, how fast they are improving, and which task they will take on next. A quieter question is whether banks and fintech firms are built to use any of it well. From 2 August 2026, organisations fall fully under the EU AI Act’s Article 50 transparency obligations, and UK regulators have started to map how far autonomous, agentic systems could reach into retail finance. Both point at the same issue: accountability for decisions a machine helped make.
Groth’s vantage point is a particular one. Decentriq sells confidential computing and data collaboration, the infrastructure that lets two organisations combine data without either seeing the other’s raw records, so his view of AI readiness is shaped by whether firms will let their data be used at all rather than by whether a given model performs. The company has a commercial interest in the argument, and the case studies below are its own. The underlying point, that readiness is about operating models rather than tools, is one worth testing on its merits.
Asked which part of a traditional financial services organisation fails first when AI systems start generating and acting on information across teams, Groth points to the review step that most processes are built around.
“The first thing to break is the approval chain, specifically the assumption that a human reviews output before it moves,” he says. “AI systems don't wait for that cadence. They generate at a pace and volume that makes case-by-case human review either impossible or purely theatrical, a rubber stamp on something nobody actually read. Day to day, this looks like backlogs of unreviewed recommendations, or reviewers who start approving in bulk because the volume has outstripped their capacity to actually check anything.”
If the review point is the weak link, governance is the word most firms reach for as the fix,and it usually means a committee. Groth argues that a committee is the wrong shape for the problem.
“Governance-as-committee fails because a committee is built to make decisions periodically, not continuously, ” he says. “What works instead is decision rights defined by category rather than by instance: before the system runs, someone specifies which classes of output can be acted on automatically, which require review, and which require escalation, revisited on a schedule rather than argued about case by case.”
On the question of who is answerable when an AI system gets a decision wrong, he rejects the idea that the technology dilutes responsibility.
“If an institution deploys a system and defines the boundaries of what it's allowed to decide, the institution is accountable for the outcome within those boundaries, the same as it would be for a junior analyst acting within their mandate,” Groth says. “The mistake is treating the AI’s involvement as diluting accountability when it actually should sharpen it, because now the boundaries are written down instead of implicit in someone’s judgment.”
Groth says the confidential-computing business shows him something a vendor selling models or agents would not see: whether an organisation is willing to let its data be used in the first place. He cites a collaboration Decentriq supported between a global wealth manager and a large publisher, aimed at reaching high-net-worth prospects.
“The wealth manager was legally not allowed to hand over raw customer data to a publisher under any circumstances, and the publisher had the same constraint in reverse,” he says. “Neither side’s blocker was the targeting model. It was that the deal could not exist at all until there was a way to combine signals without either party seeing the other’s raw data. The stall isn’t really about the technology, and it isn’t really about the AI either. It’s a signal that the organisation hasn’t resolved its own internal comfort with data use, and no model, however capable, fixes that.”
That leads to what Groth offers as a practical test of readiness. Since almost any firm can now deploy something, he argues, the amount of technology in use tells you little. What tells you more is where a project gets stuck.
“A better signal is friction location: watch where a specific AI initiative gets stuck, and check whether it’s a technical blocker or a decision-rights blocker,” he says. “If it stalls because a model needs retraining or more compute, that's a normal technology problem with a normal fix. If it stalls because nobody can say who signs off, or because two departments both assume the other owns the outcome, that's the real test failing.”
He contrasts that with a Swiss bank Decentriq worked with that had settled how its first- party data would and would not be used before any campaign began. According to Decentriq’s account, the campaign that followed recorded a 129 per cent increase in click- through rate and a 44 per cent reduction in cost per page view, figures the company reports and which have not been independently verified. “None of that came from a better model,” Groth says. “It came from the bank already knowing what it would and wouldn’t allow before the technology was ever switched on. Firms that are ready can usually point to exactly who owns a given AI-driven decision before the project starts. Firms that think they are ready can point to the technology stack and not much else.”
Asked for the most common mistake he sees, and the honest cost of it, Groth describes AI readiness being treated as a purchase.
“The most common mistake is treating AI readiness as a procurement decision: buying the capability and assuming the operating model will sort itself out once people see the results,” he says. “It doesn’t sort itself out. It usually surfaces as a slow, expensive stall: a system that’s technically live but practically idle, generating a return somewhere well south of what was modelled in the business case, while the licence, integration and maintenance costs keep running regardless of whether anyone’s acting on the output.”
“The honest cost is that nobody notices this kind of failure until much later,” he adds. “A cancelled project gets a post-mortem. A stalled one just keeps drawing down cost every quarter with no one accountable for closing it out, because on paper it’s live and technically a success.”
The timing gives the argument an edge. The EU AI Act’s transparency requirements take full effect on 2 August 2026, and the delayed obligations for higher-risk systems mean firms have longer to prepare rather than less to do. Whether the extra time is used to settle who owns an AI-assisted decision, or simply to buy more capability, is the distinction Groth is drawing.
The post AI Readiness is an Organisational Problem, not a Technology One, Argues Decentriq’s Maximilian Groth appeared first on The Fintech Times.
