AI is changing banking. The skills that matter most are proving harder to automate

An analysis of more than 1,500 credit job vacancies at the world's largest banks suggests artificial intelligence is becoming part of financial decision making, but judgement, communication and risk management remain firmly in demand

For all the predictions that artificial intelligence will transform the banking workforce, the recruitment pages of the world's largest financial institutions tell a rather more measured story.

Banks certainly want people who understand technology. They increasingly want employees comfortable with data, programming and AI supported analysis. But they have not stopped looking for the qualities upon which banking has traditionally depended: judgement, an understanding of risk and the ability to explain and defend a decision.

New research from the Global Institute of Credit Professionals has examined more than 1,500 live credit related vacancies at the world's 50 largest banks across eight major financial markets. Its conclusion is that the modern credit professional is becoming more technologically capable, rather than simply being displaced by technology.

That distinction is important.

Much of the discussion surrounding AI and employment has concentrated on which jobs machines might eventually perform. In banking, where enormous quantities of information must be processed and increasingly sophisticated models can assist with analysis, the potential for automation is obvious.

Credit, however, presents a particular challenge. Assessing whether money should be lent, on what terms and at what level of risk is not merely an exercise in processing information. Decisions have consequences, require accountability and frequently depend upon circumstances that cannot be reduced neatly to a single data point.

The research suggests banks recognise this.

The rise of the blended banker

Some 44 per cent of the roles examined remain traditional in structure. But 45 per cent are now described by the research as blended roles, combining established credit responsibilities with tasks supported by artificial intelligence.

These hybrid positions are appearing in areas including underwriting, credit research, data and modelling. AI may increasingly assist the analysis, but responsibility for the resulting decisions remains with people.

This appears less like the disappearance of the traditional credit professional than an expansion of what the job requires.

The skills being sought reflect that change. Banks are looking for knowledge of data platforms and governance, model risk, natural language processing and document analysis alongside adaptability, stakeholder management, communication and what the reportdescribes as a "risk mindset".

In other words, knowing how the technology works is increasingly useful. Knowing when to question what it produces may be more valuable still.

Excel is not dead yet

There is another revealing detail buried in the recruitment data.

Despite the excitement surrounding generative AI and increasingly sophisticated financial technology, the two most commonly requested tools in credit vacancies are decidedly familiar: Excel and PowerPoint.

Python comes third.

Almost one in ten vacancies explicitly asks candidates to be proficient in at least two of those three technologies.

It is an interesting snapshot of banking's technological transition. The spreadsheet has not suddenly been swept aside by the algorithm. Instead, programming is being added to an existing professional toolkit.

For younger workers considering careers in financial services, that makes the message rather different from the suggestion that everybody must suddenly become an AI specialist.

Technical literacy matters. But so does the ability to take an analysis, understand its limitations, reach a defensible conclusion and communicate that conclusion to another human being.

Andreas Karaiskos, Executive Director of the Global Institute of Credit Professionals, describes the change as more nuanced than the prevailing discussion around AI sometimes suggests.

"AI is changing how credit work is performed, but it is being integrated alongside the capabilities that have always underpinned good credit decision making."

He argues that the strongest careers will combine established credit fundamentals with the ability to work confidently with emerging technologies.

Judgement becomes more important further up the ladder

The balance also changes as careers progress.

At junior level, banks are placing greater emphasis on data quality, governance and communication. More senior vacancies increasingly demand leadership, commercial judgement and strategic decision making.

What remains consistent across seniority levels is the expectation that technical knowledge should be accompanied by communication, risk management and judgement.

There is a broader lesson here about the development of artificial intelligence in professional services.

Technology is particularly effective at making information easier to process, compare and interrogate. That does not necessarily make the person responsible for the eventual decision redundant. It can instead shift the value of that person's work towards interpretation, challenge and accountability.

The consequences could become more significant as AI systems move further into the credit lifecycle.

Banks may ultimately require fewer people to perform certain repetitive analytical tasks. The research does not attempt to determine whether that will happen. What the recruitment data does show is what major employers want today, and it is not a workforce composed solely of programmers or AI specialists.

They want people capable of operating between the old and new worlds of finance.

That means understanding the model without surrendering judgement to it, using data without confusing quantity with certainty and being able to explain why a decision was reached when somebody inevitably asks.

For an industry built upon deciding whom to trust with money, those distinctly human abilities may prove remarkably difficult to automate.