Dario Amodei's January essay is the most important thing a technology CEO has written for financial services leaders this year. Not because it's about banking — it isn't. But because everything it describes is about to land on the desks of every bank CEO, CTO, and board member.

 

​

In January 2026, Anthropic CEO Dario Amodei published a 20,000-word essay titled The Adolescence of Technology. It is a remarkable document — part strategic warning, part policy manifesto, part candid admission from someone building the very technology he warns about.

​

The essay's central metaphor is arresting: humanity has entered a technological adolescence — a phase where we possess extraordinary and rapidly growing capabilities but lack the institutional maturity to wield them safely. Amodei argues that artificial intelligence capable of matching or exceeding top human experts across most intellectual domains could arrive within one to two years. The implications — for labour markets, for economic concentration, for governance, for national security — are, in his telling, both transformative and deeply uncertain.

​

Most commentary on the essay has focused on the headline predictions: the potential elimination of half of all entry-level white-collar jobs within five years, the spectre of autonomous AI systems, the risks of bioweapon misuse. These are important. But for those of us who advise and govern financial institutions, the more consequential reading lies beneath the headlines.

 

Here are the three themes that matter most for banking, wealth management, and asset management — and the questions they should provoke in the boardroom.

 

1. The end of the "systems of record" era

​

For two decades, the technology conversation in banking has revolved around core banking systems — the monolithic platforms that process transactions, hold positions, and generate regulatory reporting. Billions have been spent migrating to, from, and between Temenos, Avaloq, Finnova, Olympic, and their peers. These are systems of record: reliable, auditable, and — by design — slow to change.

​

Amodei's essay describes a world in which AI systems don't merely assist humans with tasks but operate as autonomous agents capable of sustained, complex reasoning across domains. If that trajectory holds — and the pace of progress over the past two years suggests it might — then the strategic centre of gravity in bank technology shifts fundamentally.

​

The value migrates from the system of record to the system of intelligence.

 

What does this mean in practice? The core banking platform doesn't disappear. Transactions still need to settle, positions still need to be held, regulators still need their reports. But the layer where competitive advantage is created — where client insights are generated, where risk is priced dynamically, where compliance is monitored in real time, where relationship managers are guided toward the right conversation with the right client — that layer increasingly lives above the core. It is powered by AI, fed by data, and differentiated not by which vendor you selected in 2019 but by how intelligently you connect information across the enterprise.

 

Banks that spent the last decade optimising their systems of record now face a more uncomfortable question: do they have a system of intelligence? And if not, how quickly can they build one — or partner to acquire one?

​

2. The workforce question is a board-level issue

​

Amodei's prediction on employment deserves serious engagement, not dismissal. His argument is not that AI will simply automate routine tasks — that story is decades old. His argument is that AI is becoming a general substitute for human cognitive labour, and that the transition will be faster than labour markets, education systems, and institutions can absorb.

​

For banks and wealth managers, this is not abstract. Consider the functions most exposed: compliance monitoring, KYC review, client reporting, credit analysis, trade reconciliation, investment research. These are not peripheral activities. They represent a significant share of the cost base — and of the workforce — in any mid-to-large financial institution.

​

The strategic question is not "will AI replace these roles?" but rather "what is the right operating model when AI can perform 60–80% of the cognitive work in these functions?" The answer is neither wholesale automation nor denial. It is a deliberate redesign of how human judgment and AI capability are combined — what some firms are beginning to call "augmented operating models."

​

This redesign is a board-level conversation, not an IT project. It touches headcount planning, compensation structures, talent strategy, risk appetite, and — critically — the culture of the organisation. Boards that delegate this to the technology committee alone will find themselves approving headcount reductions without having shaped the strategic logic behind them.

 

​

3. Concentration risk — in technology, in talent, in thinking

 

One of Amodei's most underappreciated arguments concerns economic concentration. As AI capability consolidates among a small number of companies controlling the foundational models, the data infrastructure, and the compute capacity, the power dynamics of the technology industry shift dramatically. The firms that build and operate these systems become, in Amodei's words, "power brokers" with an influence that extends well beyond their commercial footprint.

​

For financial services, this creates a new category of concentration risk that most boards have not yet grappled with. Today, a mid-sized European wealth manager might rely on a single cloud provider, a single core banking vendor, a single market data supplier, and — increasingly — a single AI model provider. The dependency chain is lengthening and narrowing simultaneously.

 

This matters for operational resilience, obviously. But it also matters for strategic autonomy. If the intelligence layer of your business is powered by a model you don't control, trained on data you can't inspect, and updated on a schedule you don't set, then you have outsourced a meaningful portion of your strategic capability to a third party. The FINMA and ECB frameworks on outsourcing and cloud risk were designed for a world of infrastructure-as-a-service. They are not yet calibrated for a world of intelligence-as-a-service.

 

​

Three questions for the boardroom

​

Amodei's essay is not a roadmap for banking. But it is a provocation that should inform how boards think about technology strategy, workforce planning, and risk governance over the next three to five years. I would suggest three questions that every bank and asset management board should be discussing:

 

First: Where does our competitive intelligence actually reside — and who controls it? Map the critical decision points in your value chain. For each one, ask: is this driven by human judgment, by a proprietary system, or by a third-party AI capability? If the answer is increasingly the latter, the board needs to understand the dependency and the alternatives. This is not a technology audit. It is a strategic sovereignty question.

​

Second: Do we have a workforce transition plan that matches the pace of AI capability growth? Not a vague commitment to "upskilling," but a concrete, function-by-function view of how roles will evolve, what new capabilities are needed, and how the organisation will manage the human dimension of this shift. The firms that navigate this well will attract and retain the best people. The firms that don't will face either uncontrolled cost-cutting or a slow erosion of relevance.

 

Third: Is our governance framework designed for the speed and nature of AI-driven change? Most bank boards meet quarterly. Most risk frameworks are calibrated for incremental change. AI capability is advancing on a cadence that is closer to monthly. Boards need to ask whether their information flows, their committee structures, and their own digital literacy are adequate for overseeing an organisation that is increasingly shaped — and exposed — by AI.

​

Amodei writes that humanity is entering a phase that will "test who we are as a species." That may be true at the civilisational level. At the institutional level, the test is more specific: can financial services firms — historically among the slowest-moving institutions in any economy — adapt at the pace that this technology demands?

​

The essay does not answer that question. But it makes clear that the question can no longer be deferred.

​

 

Silvan Schriber is Managing Director at Alvarez & Marsal and a Board Member at Zuger Kantonalbank. He advises financial institutions on strategy, transformation, and governance.

​