Why AI governance for financial services leaders needs judgement

What is AI governance for financial services leaders? 

AI governance is the set of policies, roles, and review processes that a financial institution uses to oversee how artificial intelligence systems are built, deployed, and monitored. In practice, this covers who signs off on a credit-scoring model, who checks an AI-generated client recommendation before it goes out, and who is accountable when an automated decision is challenged by a regulator or a customer. 

For leaders in banks, asset managers, and Big 4 firms, governance is fundamentally a leadership responsibility, because the leader who approves an AI-assisted decision remains accountable for its outcome, regardless of how the system reached its conclusion.

What does AI governance mean for financial services leaders?

AI governance in financial services means defining who is accountable for an AI system's outputs and how that accountability is exercised at each stage of the system's use. This differs from AI governance in other sectors because financial decisions, particularly credit, insurance pricing, and investment advice, carry direct legal and regulatory consequences for the people affected by them.

AI governance versus AI risk management

Governance and risk management are related but distinct. Risk management is the technical layer, covering model validation, data quality, and monitoring for drift or bias. Governance sits above this layer and answers a different question: who has the authority to approve, pause, or override a model, and on what basis. A firm can have strong technical risk controls and still have weak governance if no individual is genuinely equipped to exercise that authority when it matters.

What does the EU AI Act require of leaders?

The EU AI Act requires financial services leaders to maintain human oversight of high-risk AI systems, meaning a named person must be able to understand, interpret, and if necessary override a system's output. Financial institutions should treat this as a current and moving requirement, since the compliance timeline for standalone high-risk systems has already been revised once by EU regulators.

Which financial services uses count as high risk

Credit scoring and insurance underwriting are the clearest examples of high-risk AI use under the Act, because they directly affect a person's access to financial products. KPMG's analysis of the EU AI Act notes that such systems must be designed to allow proper human understanding of their outputs, and that supervisors, including the European Central Bank, expect banks to demonstrate they have genuinely reviewed a model's recommendation rather than followed it automatically.

What human oversight requires from named reviewers

The regulation does not treat a signature as sufficient. It expects the named reviewer to interpret the system's reasoning well enough to challenge it. This is a meaningfully higher bar than most institutions have built into their current sign-off processes, and it is the gap that leadership development is positioned to close.

Why do AI governance frameworks still fail?

A governance framework typically fails at one specific point: the person responsible for oversight lacks the standing or the developed judgement to exercise it under pressure. Board-level data supports this. McKinsey's research on board oversight found that more than 88 percent of organisations use AI in at least one business function, yet only 39 percent of Fortune 100 companies disclosed any form of board oversight of AI as of 2024. The same research found that 66 percent of global board directors report limited to no knowledge or experience with AI.

The accountability gap boards face

A governance policy that names an accountable individual does not automatically give that individual the confidence or the analytical habits to push back on a system's recommendation, particularly under time pressure or when the system is usually right. This is the accountability gap that shows up in board discussions long before it shows up in an audit finding.

Is AI governance a leadership skill or a policy?

A governance policy defines who is responsible for an AI system, but the judgement of that named person determines whether the policy actually works. This makes AI governance function as a leadership skill as much as a written document, and The Henka Institute's coaching-led approach treats that judgement as something to be developed deliberately, not assumed.

This distinction sits at the centre of The Henka Model, which describes a leadership philosophy built on suspending judgement, recognising cognitive bias, and staying genuinely curious rather than directive. Within the Five HenkaQs, HeadQ addresses exactly this capability, the discipline of pausing before accepting a conclusion, and of recognising when a leader's own heuristics are doing the deciding rather than a considered review.

	Compliance-led governance	Leadership-led governance
Focus	Documented policy and control objectives	Developed judgement of the named reviewer
Ownership	Risk and compliance functions	Senior leaders accountable for the outcome
Output	Signed-off approval or audit trail	A challenge to the model where one is warranted
Risk if missing	Regulatory exposure and audit findings	Deference to AI outputs without genuine review

How coaching builds the judgement AI oversight assumes

The Henka Institute's Leader as Coach programme is built around this same skill set, teaching leaders to ask better questions rather than accept the first plausible answer, whether that answer comes from a colleague or a model.

The HeadQ habit of suspending judgement

HeadQ is specifically about creating the mental space to co-create a path forward with others rather than defaulting to the fastest available conclusion, which is the exact behaviour EU regulators are asking named reviewers to demonstrate.

How does AI governance affect client trust?

Client trust depends on AI governance because clients in financial services, particularly in private banking and wealth management, are trusting a relationship built on judgement as much as on any single recommendation. When a client learns that part of their advice was AI-assisted, their confidence rests on believing a competent, curious human reviewed it.

This is where HeartQ, the Henka Model's principle of empathy and connection, becomes commercially relevant rather than a coaching concept alone. Trusted-adviser relationships in financial services are built over years through consistent, attentive judgement, and clients notice quickly when that judgement is replaced by process. The Henka Institute's leadership development for financial services treats this relationship depth as a measurable outcome of coaching.

What should financial services leaders do now?

Financial services leaders preparing for the EU AI Act's revised deadlines should treat the named human reviewer role as a leadership development priority as well as a compliance appointment. A reviewer who has not built the habit of questioning confident, data-backed conclusions will struggle to meet the standard regulators are describing, regardless of how detailed the underlying policy is. Building coaching skills across a leadership population, or drawing on a trusted adviser to senior leadership teams, gives that reviewer the standing and the practised judgement the role actually requires. This is the practical shape of AI governance for financial services leaders in the years ahead.

Frequently asked questions

What is AI governance in financial services?

AI governance in financial services is the set of policies and accountability structures that determine who oversees an AI system's decisions, from approval through to ongoing monitoring. It applies most directly to high-impact uses such as credit scoring, fraud detection, and investment advice. Its purpose is to keep a clear line of human accountability over outcomes that affect customers.

Who is responsible for AI governance in a bank?

Responsibility typically sits with a combination of the board, senior risk and compliance leaders, and named individual reviewers assigned to specific high-risk systems. McKinsey's board research found that formal board-level ownership of this responsibility remains rare, even as AI use has become widespread. The gap between who is technically accountable and who is genuinely equipped to act on that accountability is where most governance failures begin.

What does the EU AI Act require for AI used in financial services?

The EU AI Act classifies systems such as credit scoring and insurance pricing as high risk, requiring human oversight, documentation, and the ability for a named reviewer to interpret and challenge the system's output. KPMG's coverage of the Act sets out the supervisory detail for financial institutions specifically, including the expectations European regulators have signalled for banks.

What is human oversight in AI governance?

Human oversight means a named individual has the genuine capacity to understand an AI system's reasoning and override it when needed, rather than simply approving its output. Regulators, including the European Central Bank, have said explicitly that they expect this to function as a genuine check on the system. Building that capacity is a leadership development task as much as a compliance one.

How do you build an AI governance framework?

A workable framework combines documented policy, technical risk controls, and leaders who have developed the judgement to act on their oversight responsibilities. The ICF's work on AI in coaching reflects a similar principle from the coaching profession itself, treating human judgement as a capability that needs deliberate development alongside any new technology.