What boards, CROs and people leaders must do to keep AI powerful, safe and profoundly human.
AI in financial services is no longer a lab experiment. Models now sit inside credit decisions, investment workflows, surveillance, HR and client engagement. The question is shifting from “can we use AI?” to “how do we lead when a growing share of work is done by digital teammates?”. Leadership and governance have become the real edge.
Regulators and standard-setters are moving quickly. The CFA Institute’s 2025 chapter on Ethical AI in Finance calls out four pillars for responsible AI, fairness, transparency, accountability and privacy, and urges firms to embed them into real systems rather than treat them as slogans. The American Bankers Association’s updated framework for AI regulation in finance echoes this, arguing for principles-based oversight across the full lifecycle, from data collection to model retirement, with clear roles for risk, compliance and cyber functions. In parallel, round ups of enacted AI rules highlight how fairness, explainability and accountability are now hard requirements, not “best practice”.
For boards, this means AI can no longer be delegated entirely to the CIO. It demands cross functional literacy. Directors need to understand where AI is in the value chain, what could go wrong and which decisions remain irreducibly human. CFA Institute’s latest work on explainable AI makes the point clearly, if you cannot explain model behaviour to diverse stakeholders, from supervisors to retail clients, you have a trust problem, regardless of technical accuracy.
At the same time, leadership research is reframing what “good” looks like in a hybrid human AI environment. A 2024–25 literature review on human AI collaboration proposes leadership as a mediating force between human and machine intelligence, shaping how tasks, authority and responsibility are divided. Other studies show that emotional intelligence materially influences trust in AI systems and collaboration quality in knowledge intensive industries. This is not just academic. Recent commentary on leadership trends stresses that strategic AI skills, human AI collaboration and emotional intelligence now rise together, not in opposition.
In practice, leading the algorithmic workforce involves at least three shifts.
First, leaders must treat AI systems as part of the team. McKinsey’s 2025 State of AI survey shows that only a minority of organisations have moved from pilots to scaled value, and that “high performers” are far more likely to redesign processes around AI and clarify when humans step in, rather than simply sprinkling tools on top of existing workflows. This demands a domain level view, for example redesigning the entire KYC journey or credit process, not just adding a chatbot at the front.
Second, leaders need to become translators between ethics, regulation and technology. The ABA’s latest guidance on AI legislation calls for avoiding a patchwork of rules while still strengthening penalties for AI enabled fraud. National governments are also asserting clearer AI policy frameworks. In this landscape, the most effective CROs, CIOs and business heads will be those who can frame governance choices in commercial language, for example articulating how explainability and model risk controls support product approval, distribution and capital efficiency rather than blocking innovation.
Third, leadership becomes more human, not less. As AI takes on routine analysis and process work, the differentiator for managers is their ability to coach, empathise and create psychological safety. Studies on leadership in human AI collaboration argue for “augmented intelligence” models where machines handle pattern recognition and humans own meaning, values and direction. It is no accident that senior technology leaders like Satya Nadella are publicly calling empathy a “workplace superpower” in the age of AI.
Finally, there is the question of scale. Many banks still have dozens of promising AI proofs of concept that never make it into production because workflows are brittle and accountability is fuzzy. Drawing on recent playbooks from McKinsey and others, a pragmatic path forward is to: pick a high value journey; map tasks and failure modes; redesign around explicit human decision points; instrument outcomes such as cycle time, loss events and customer impact; and build feedback loops so models improve in the flow of work.
The thread through all of this is simple. AI does not remove the need for leadership or governance, it raises the bar. The institutions that will win in an algorithmic era are those where boards set clear guardrails, executives design genuinely human plus machine operating models, and managers are equipped to lead teams that include both people and intelligent systems. Done well, that combination can produce operating models that are more efficient and innovative, yet also more transparent, trustworthy and human centred.
