A Chief AI Officer built from an engineering and lab pedigree
Bank of Ireland has named Prag Sharma as its first Chief Artificial Intelligence Officer, effective October 2026, and the profile it chose says a lot about how the bank intends to run AI. Sharma is an engineer by training, holding a bachelor's in electronic engineering from University College Dublin, a PhD in computer vision and pattern recognition, and later an MBA from the same institution. He spent roughly a decade at Citi, most recently as director of emerging technologies for the Citi Institute's Future of Finance, and earlier as global head of AI and a senior figure in Citi Innovation Labs. This is a builder with a research spine who has shipped production systems at global scale.
That background matters because the Chief AI Officer title has become a catch-all, sometimes given to a communications lead, sometimes to a data executive with a new business card. Bank of Ireland credited Sharma with building Citi's AI Centre of Excellence, which the bank says delivered end-to-end AI products across 96 countries. Whatever discount we apply to corporate phrasing, the claim points to someone who has moved AI from proof of concept into regulated production, across jurisdictions, under real controls. For a retail and commercial bank operating under European supervision, that specific experience is worth more than a general innovation reputation. It suggests the role is designed to deliver, and to survive an audit.
The reporting line is the real story
The reporting structure is where we read intent. Bank of Ireland says Sharma will lead AI delivery and enablement across the enterprise, working directly with Digital and Data, Security, and IT leadership, and will shape emerging technology strategy alongside group strategy. That wiring puts AI next to security and data governance from day one rather than in an isolated lab that hands finished toys to the business. When the person accountable for AI outcomes sits in the same conversations as the people accountable for risk, models tend to ship with controls attached. When those functions are separated, you get demos that never clear compliance.
For CxOs designing a similar role, this is the transferable lesson. The value of a Chief AI Officer is decided by the reporting line and the decision rights, long before any model is trained. If the role owns enablement but not risk, it becomes a marketing function. If it owns risk but not delivery, it becomes a brake. Bank of Ireland appears to have given Sharma both, embedding him across the functions that either accelerate or block AI in a regulated firm. We would want to see the budget authority behind the title, because accountability without spending power is a familiar way to set an executive up to fail.
Why banks are racing to create this seat in 2026
Sharma joins a visible wave of banks creating this seat in 2026. The pattern is consistent across the sector: institutions that spent 2024 and 2025 experimenting with generative AI are now appointing a single accountable owner to industrialize it. The driver is partly competitive, as boards worry about falling behind on productivity, and partly defensive, as supervisors sharpen expectations around model risk, data lineage, and consumer protection. A named executive gives regulators a person to hold responsible and gives the board a clear line of accountability. That is a governance instinct as much as a technology one.
The risk in the trend is theater. Appointing a Chief AI Officer can substitute for the harder work of fixing data foundations, cleaning up access controls, and retraining staff. A title does not resolve fragmented data or legacy core systems, and it does not conjure the engineers needed to maintain models in production. Bank of Ireland has tied the role to a strategy running through 2028, which at least implies a multi-year mandate and the patience that AI at scale requires. The test will be whether Sharma controls a real delivery organization or merely convenes one. We will judge the appointment by what ships in 2027, not by the announcement.
'Safely accelerate' is the phrase that matters
Chief Operating Officer Ciaran Coyle said Sharma 'brings vision, leadership and expertise required to safely accelerate our AI transformation.' We flag the word safely because it is doing real work in that sentence. In a bank, speed without control is how you end up explaining a biased lending model to a regulator, and control without speed is how you cede ground to nimbler competitors. Coyle put both values in a single phrase, which reads as a deliberate framing of the mandate. The honest question is whether the organization behind the title can actually hold that tension.
Every technology leader outside banking should borrow that framing when pitching an AI hire to their own board. The reason to fund a senior AI owner is the ability to move quickly while keeping failure modes contained, an argument that survives a downturn where budgets for pure novelty do not. If you are preparing a case for a Chief AI Officer, anchor it to the specific risks your business runs, model governance, data privacy, and customer trust, and to the specific speed you are losing without coordinated ownership. Concrete stakes travel further with a finance committee than ambition does.
The talent signal: poaching from Citi's Future of Finance
There is a talent signal here worth naming. Sharma left a role at the center of Citi's forward strategy, its Future of Finance think tank, to run AI at a smaller national champion. That direction of travel tells us the Chief AI Officer title now carries enough scope and status to pull senior people out of the largest institutions. For banks and enterprises below the mega-cap tier, that is encouraging, because it means the role can be a genuine draw rather than a consolation prize. The scope of the mandate, not the size of the logo, is increasingly what moves this tier of executive.
It also raises the cost of doing nothing. If your competitors are creating well-scoped AI leadership roles and winning the people who built AI programs at global banks, a vacant or watered-down version of the role puts you at a hiring disadvantage that compounds. The market for executives who have shipped regulated AI at scale is thin, and it is being drained quickly. We would rather define this role precisely and fill it deliberately now than reverse-engineer it in eighteen months after watching rivals pull ahead. Talent scarcity turns hesitation into a lasting gap.
What we would pin down before creating the role
If we were on Bank of Ireland's board, three questions would decide whether this appointment pays off. First, what is Sharma's budget and headcount, and can he redirect existing IT and data spending toward AI priorities. Second, who owns the final call when an AI use case is commercially attractive but sits in a regulatory grey zone, the Chief AI Officer, the risk function, or a joint forum. Third, what are the measurable outcomes for 2027, in cost, revenue, or risk reduction, that this role is expected to deliver. Clear answers separate a real operating role from an expensive symbol.
For readers running their own AI agendas, the appointment is a prompt to audit your own structure. Ask whether anyone in your organization actually owns AI outcomes end to end, or whether responsibility is smeared across a data team, a security team, and a scattering of product managers. Bank of Ireland has placed a clear bet that a single, technically credible, well-connected owner beats diffuse accountability. We think that bet is correct for most enterprises now attempting AI in production, provided the mandate attached to the title is real.
