Banks Pay $3.5 Million for AI Chiefs Built to Fade

76% of surveyed organizations now have a chief AI officer, up from 26% in 2025. Banks spent the first half of 2026 filling that seat with pay that approaches $3.5 million.

HSBC, Commonwealth Bank of Australia and Lloyds Banking Group all created or restaffed the job. The people taking it talk as if the title should work itself out of existence, the way banks never hired a chief Excel officer. The last time lenders invented a C-suite job like this, for digital transformation, many of those seats were later folded back into the CIO line.

Three Banks, Three Different Bets on One Title

HSBC went inside. David Rice, an 18-year lifer who had been chief operating officer for Corporate and Institutional Banking, became the group’s first chief AI officer on April 1, 2026. Georges Elhedery, the group chief executive, tied the hire to faster, more personal service. Rice said AI would take a growing share of HSBC’s plans and that the new brief was to push the transformation agenda. Mario Shamtani, the chief technology officer, had his remit widened at the same time to build the plumbing Rice’s team will need.

Commonwealth Bank of Australia went the other way and poached. Ranil Boteju, who had been group chief data and analytics officer at Lloyds since 2021, was named CBA’s chief AI officer with a start in early 2026. It is his third tour at the Australian lender, after a stint from 2007 to 2010 and 15 years abroad that also took in Standard Chartered and HSBC. At Lloyds he ran a team of more than 2,000 people and more than 50 generative AI projects, a run the bank said helped lift it 12 places on the 2025 Evident AI Index. Matt Comyn, CBA’s chief executive, said Boteju’s job was to scale value from AI for customers and staff. CBA already sat 4th on that Evident ranking.

Lloyds then replaced the man it lost. On April 16, 2026 the group said it had appointed Sameer Gupta as Chief Data and AI Officer, joining in June 2026 from DBS Bank in Singapore, where he was chief analytics officer. He has more than 30 years in financial-services technology. He reports to Ron van Kemenade, group chief operating officer, not to the board. A core part of the brief is governance: responsible, transparent and secure use that meets regulatory and ethical standards, plus fraud tools and an AI financial assistant for customers. DBS’s AI programme had already been written up as a Harvard Business School case.

WHO TOOK THE NEW SEATS

The three hires already disagree on what the job is. Rice is an operator asked to spread tools across a global bank. Boteju is a data executive brought home to scale a programme that is already ranked. Gupta is a builder-plus-governor who reports through operations. Pei Ying Chua, LinkedIn’s head economist for Asia-Pacific, has said the post is there to help a firm make sense of its AI strategy, including what not to do, and that no two strategies match.

The Median Pay Package Is $1.6 Million

Equilar, which mines disclosed top-50 executive pay, put a median pay package of $1.6 million on AI executives in its 2025 Top 50 Survey, published on September 11, 2025. Exact bank salaries are rarely filed. The tails are wide, which is why the banking raid is so loud.

WHAT EQUILAR’S AI PAY BANDS SHOW

• Tenth percentile: $567,235 in total pay.
• Twenty-fifth percentile: around $700,000.
• Median package: $1.6 million.
• Seventy-fifth percentile: about $2.5 million.
• Ninetieth percentile: approaches $3.5 million.

Base salary is the largest median piece, at $439,375, then time-based stock awards at $242,997, cash bonuses at $191,307 and performance awards at $90,186. Stock options sit at zero at the median. Those slices are medians of each part, not a sum, and they do not add back to $1.6 million. Banks that want the scarce names still have to clear the top of that range, which is why rivals are raiding each other instead of waiting on a campus pipeline.

Pay at that level is what you post when AI has left the lab and become part of how the bank runs, not a side experiment. It also sits a long way from the rest of the AI labour market, where younger workers already wary of AI jobs are pulling back from the same tools these lenders are capitalising. In Singapore, LinkedIn has ranked head of AI posts among the fastest-growing job classes, with specialised tech demand still ahead of supply.

Chief Digital Officers Were Called Transformers, Then Absorbed

A decade ago the fashionable hire was a chief digital officer. Consultants billed the person as the transformer in chief, a bridge between a CIO accused of minding the lights and a business that wanted apps. By 2022 the same title was being joked about as the chief disappearing officer, as firms quietly merged it into the CIO, CTO or chief product officer line.

The logic was baked into the job description. Digital transformation was sold as a project with a start and a finish. A CDO who really finished the work made the extra box on the org chart look like leftover scaffolding. Banks used the role heavily because mobile apps and fintech rivals had opened a gap. Once the apps shipped, many boards decided two technology chiefs were one too many.

The new AI brief copies that script almost line for line. Spread the tools. Train the staff. Stand up governance. Then, in the telling, step back so every business head owns the models the way they already own a spreadsheet. David Hardoon has said any chief AI officer should work on the assumption that the role should eventually become unnecessary, and he reached for the missing chief Excel officer and chief email officer as the punchline.

That analogy only holds if bank AI behaves like email. Email did not underwrite loans, flag fraud, or sit in a model inventory a supervisor can ask for. The CDO wave is still the closest rehearsal. It is not a perfect one.

What Hardoon Meant by a Chief Excel Officer

Hardoon is the cautionary tour. He joined Standard Chartered in April 2025 as global head of AI enablement, based in Singapore, reporting to group chief data officer Mohammed Rahim. He had been the Monetary Authority of Singapore’s first chief data officer and, before that, chief data and AI officer at UnionBank of the Philippines. The bank put him inside the chief data office to find places AI could add value and to hold the governance line. Standard Chartered ranked 26th among global banks on the 2025 Evident AI Index.

By March 2026 he was gone, on gardening leave, after less than a year. Standard Chartered declined to comment. In June 2026 he joined Accenture as managing director and head of advanced AI for Southeast Asia, with a brief covering generative AI, agentic AI and responsible AI across industries. The person who said the bank seat should vanish is now selling the capability from a consultancy that gets paid when the next bank decides it needs one.

HOW THE TITLE MOVED IN 18 MONTHS

1. April 2025: Hardoon joins Standard Chartered as global head of AI enablement.
2. December 2025: CBA names Boteju as chief AI officer, effective early 2026, taking him from Lloyds.
3. March 2026: Hardoon exits Standard Chartered after less than a year.
4. April 1, 2026: Rice becomes HSBC’s first chief AI officer.
5. April 16, 2026: Lloyds names Gupta, to join in June 2026 from DBS.
6. May 18, 2026: CBA adds Professor Mary-Anne Williams as its first chief AI scientist, reporting alongside Boteju.
7. June 2026: Hardoon joins Accenture in Singapore; Gupta starts at Lloyds.
8. August 2026: Abu Dhabi Islamic Bank names Pedro Uria-Recio as chief AI officer.

The August hire is the tell that the June wave was not a one-quarter fad. CBA splitting a chief AI officer from a chief AI scientist, and HSBC adding a chief AI officer post inside its legal team, points the other way from Hardoon’s vanishing-title story. The box is multiplying, not shrinking.

The COO Still Signs for the Risk

If the enablement half of the job looks like the old CDO brief, the risk half does not. Gupta’s own appointment letter puts regulatory and ethical oversight in the first page of duties. Rice’s seat was carved out of a COO job. In the United Kingdom, the Senior Managers and Certification Regime already parks technology integrity with SMF24, the chief operations function, and model and conduct risk with SMF4, the chief risk function. A chief AI officer who reports to the COO can coordinate. The accountable executive on the regulator’s list is often still the person who already signed.

That is the hidden stake. Boards get a title they can announce. Supervisors still want a named senior manager when a credit model, a fraud engine or a customer assistant goes wrong. US agencies rewrote their model-risk guidance in April 2026 and left generative and agentic systems outside the formal rulebook, while telling banks to apply their wider risk practices anyway. In Europe, lenders already juggle the AI Act, DORA operational-resilience rules, and older model risk management in the age of AI expectations from the ECB and the UK Prudential Regulation Authority.

WHERE EXPERTS DISAGREE

• Hardoon: Treat the chief AI officer as a temporary enablement post that should work toward its own redundancy, the way no bank staffs a chief Excel officer.
• IBM’s surveyed CEOs: Among firms that already have a chief AI officer, every respondent in the 2026 CEO study expects that role’s influence to increase by 2030.
• Chua at LinkedIn: The job is a strategy translator, including decisions to hold back, and it will not look the same from bank to bank.

THE RULEBOOKS THAT OUTLAST A TITLE

• UK SMCR: AI infrastructure and model risk still map onto existing senior-manager functions, mainly operations and risk, whether or not a CAIO sits on the chart.
• US model risk: The April 2026 rewrite narrowed what counts as a model and parked generative tools outside that text, without parking them outside governance.
• EU stack: The AI Act, DORA and internal-model guides all ask for inventories, oversight and a human who can explain a decision.

A title can disappear. The demand for a person who can answer for a model does not. That is why the CDO comparison, which is real, only covers half the job.

A Small Circuit of Names Keeps the Seats Filled

Watch the CVs, not the press-release adjectives. Boteju’s path runs CBA to HSBC to Standard Chartered to Lloyds and back to CBA. Gupta comes from DBS, the Singapore bank that other lenders treat as a finished AI case study. Hardoon went from a regulator to a bank to a consultancy in 14 months. Rice is the exception, promoted from inside HSBC’s wholesale operations rather than hired off that circuit.

Lloyds felt the cost of the circuit directly. It lost Boteju to CBA and then hired Gupta into a broader chief data and AI officer brief, on top of an earlier group head of AI post. The specialised skills, as the banks themselves keep saying, are scarce enough that the same few dozen people rotate through London, Singapore, Hong Kong and Sydney. Each move resets a programme and takes a year of context with it. Hardoon’s sub-year tenure is the extreme case. It is not the only short one this market will produce.

Appointing a chief AI officer also does not settle what the bank will actually do. Chua’s point about deciding what not to do sits underneath every raid. Some of these jobs will be builders. Some will be governors. Some will be expensive translators for a board that wants a slide. The title is the same word in each announcement.

IBM’s CEOs Expect Influence to Rise Through 2030

The IBM Institute for Business Value, working with Oxford Economics, surveyed 2,000 chief executives across 33 geographies and 21 industries from February to April 2026. The study, released on May 4, 2026, is the source of the 76 percent of surveyed organizations figure. It also undercuts the idea that the new box is already on the way out.

WHAT THE 2026 CEO STUDY FOUND

• Adoption: 76% have a chief AI officer in 2026, up from 26% in 2025.
• Scale: Firms that redesigned the C-suite around an AI-first setup have scaled 10% more AI initiatives than peers.
• Use: Only 25% of staff use AI regularly in the job, even though 86% of CEOs think employees already have the skills to work with it.
• 2030: CEOs expect 48% of operational decisions that can be coded with guardrails to be taken by AI without a human, against 25% of those decisions now.

Among organizations that already have a chief AI officer, every surveyed CEO expects that role’s influence to increase by 2030. At the same time, 85% say every functional leader must become a technology expert in their own domain, 79% are pushing decisions out from the centre, and 83% say AI success depends more on people’s adoption than on the tools. Those lines can both be true. Influence can rise while the work spreads. That is closer to how a risk function behaves than to how a conversion project behaves.

The CEO’s role has always been to lead through disruption. What AI changes is the velocity and consequences of leadership.

Gary Cohn, Vice Chairman, IBM Institute for Business Value CEO study

Between 2026 and 2028, the same CEOs expect 29% of employees to need reskilling for a different role and 53% to need upskilling in the job they already hold. Fifty-nine percent think the chief human resources officer’s influence will rise. Seventy-seven percent say talent and technology leadership are converging. Firms that redesigned five core areas (technology, finance, HR, operations and cross-functional work) were four times more likely to have hit their business objectives. Sixty-four percent are already comfortable using AI-generated input on major strategic calls, and 83% call control over their own models, data and infrastructure essential.

Van Kemenade, announcing Gupta, put the banking version of that in plainer words: the hire has to improve customer experience and still keep the trust the licence sits on. Gupta said used well, AI can get customers help faster, protect them from fraud, and make managing money simpler. Those are production claims, not a transformation programme with an end date.

Hardoon’s Excel line is still the cleanest warning in the file, and the CDO decade shows how fast a fashionable title can be absorbed. The IBM survey, the SMCR map, and the fact that CBA and HSBC are already splitting the job into scientist and legal variants all point to a seat that gets renamed more often than it gets deleted. The banks that just paid packages that approach $3.5 million are buying a coordinator. The person who still signs when a model fails is, in a lot of these structures, the COO they already had.