Will AI Replace Your Finance Job? An Honest Assessment and a Practical Plan

Will AI Take Finance Jobs? What Treasury and Risk Professionals Should Do Now

If you work in treasury, risk, or regulatory reporting and you are wondering what AI means for your career, this post gives you a specific, honest answer. Not reassurance, not hype — a clear view of what is already changing, what is not, and three concrete steps to take this quarter.

What is actually at risk

The risk is not that a machine replaces you overnight. The risk is that a colleague who uses AI well gets the promotion, the project, or the headcount you were expecting.

That distinction matters. Job titles in finance are not disappearing at speed. Workflows inside those jobs are. A treasury analyst who spent 40% of their week pulling data, reconciling positions, and drafting commentary for a cash report is now, in some teams, spending far less time on those tasks. The question is what they are doing with the time freed up, and whether they made that visible to anyone who matters.

This post is not going to tell you AI is fine and nothing will change. Some junior, high volume roles in finance will shrink, and being honest about that is more useful than reassurance. But wholesale job elimination across treasury, risk, and regulatory reporting is not the near term story. Structural change to what those roles contain is.

What AI is already doing inside finance teams right now

The gap between what firms announce and what is actually running in production is still wide. But some changes are already embedded.

Generative AI tools are being used to produce first drafts of management commentary, board reports, and regulatory submissions. Analysts paste in variance tables, give the model some context, and get a structured paragraph back. It is not always good. It often needs editing. But a structured first draft that takes seconds rather than thirty minutes is now the starting point, and that used to take real time.

Reconciliation and data gathering workflows are being automated at a faster pace than before, partly through dedicated tools like Alteryx and partly through AI assisted Python scripting. Teams that previously needed a specialist to build a reconciliation pipeline can now get a working prototype from someone with moderate coding ability and a good prompt. If you want to understand where Alteryx fits into that picture, Alteryx Designer for finance teams is worth reading alongside this.

Code generation is also real. Analysts who are not developers are producing working Python scripts for data pulls, basic modelling, and report automation. The code is not always ready for production. But it runs, it does the job, and it is compressing what used to be a handoff to an IT team into something one person handles.

The tasks most exposed in treasury, risk and regulatory reporting

Being specific is more useful than being vague here. The tasks facing the most pressure are:

• Data gathering and aggregation. If your value is primarily in pulling numbers from multiple systems and putting them in one place, that workflow is exposed.
• First draft report writing. Commentary that follows a predictable structure, for example variance analysis against budget, or liquidity reporting narratives, is exactly what generative AI handles reasonably well.
• Narrative regulatory submissions. Standard sections of ILAAP and ICAAP narratives that follow a known format are being drafted faster with AI assistance. The judgement layer on top is still human, but the scaffolding is increasingly machine generated.
• Repetitive variance analysis. Identifying that a metric moved by X because of Y, where Y is mechanically derivable from the data, does not require the same analyst time it once did.
• Basic code generation for reporting. Simple data transformation scripts, formatted outputs, and scheduled report pulls are increasingly within reach of someone who is not a developer using AI tools.

None of these are complete job functions. They are components of jobs. But if several of them sit in your current role, the shape of that role is going to change.

What AI still cannot do in a regulated finance environment

Some of the commentary about AI replacing finance professionals falls apart when you look at what actually matters in banking.

Regulatory judgement under ambiguity. Deciding how to classify an asset for LCR purposes when the product does not fit neatly into the UK LCR rules, or determining whether a stress scenario in an ILAAP is credible and proportionate, requires someone who can read a rule, understand the regulator's intent, and make a defensible call. That person owns the answer to the PRA if challenged. No model does.

Accountability. Under the Senior Managers and Certification Regime, accountability must sit with an identified individual. This extends across Senior Manager Functions and Certification Functions alike. AI can help you think. It cannot be responsible.

Stakeholder management under pressure. When a CFO is questioning a liquidity position two hours before an ALCO, or a regulator asks a pointed follow up question in a supervisory meeting, the skill required is not data processing. It is reading the room, knowing what to say and what not to say, and managing the relationship. No model is close to replacing that.

Interpreting genuinely novel events. AI models are trained on historical patterns. When something genuinely new happens, whether a market event, a structural change to a product, or a new regulatory expectation, the professional who understands the business deeply will outperform the one who is relying on a model to pattern match.

AI coding agents are becoming more capable, but they still require someone who understands the domain to catch the errors. The post on AI coding agents in finance and how to maintain control of delivery covers that control question in detail.

The profile that stays competitive: domain depth plus AI fluency

The finance professional who stays competitive is not the one who is most enthusiastic about AI. It is the one who combines genuine domain knowledge with enough technical fluency to direct, interrogate, and quality check AI outputs.

Domain depth means understanding what the numbers represent, what the regulatory framework requires, and what the business is actually trying to achieve. That context is what lets you spot when an AI output is plausible but wrong. A generated NSFR commentary that uses the right terminology but misrepresents the treatment of a specific liability class will not be caught by someone who is just checking the format.

AI fluency means being able to use these tools productively without outsourcing your judgement to them. It does not mean being a software engineer. It means knowing how to prompt effectively, how to validate outputs, and how to build simple automations that free up your time for the work that requires you specifically.

The combination is the thing. Deep domain knowledge without any engagement with AI leaves you slower than your peers. AI fluency without domain knowledge produces confident nonsense. Neither alone is the answer.

If you want to think about the leadership dimension of this, the piece on servant leadership habits for finance managers is relevant, particularly the section on how managers can support teams through structural change.

Three concrete steps to take this quarter

Not this year. This quarter. Three months is long enough to make meaningful progress and short enough to stay honest with yourself about whether you are actually doing it.

1. Learn one automation or coding tool to a working level

Pick one. Python with pandas for data work, Alteryx for workflow automation, or a carefully structured use of a generative AI tool for report drafting. The goal is not certification. The goal is producing something real with it that you could show someone. Practical, finance focused Python courses are available at the Academy if that is the direction you choose.

If your team uses a cloud data platform, understanding where that fits also matters. The post comparing Snowflake vs Databricks for financial services teams gives a practitioner level view of the landscape.

2. Audit your own role for exposed tasks

Write down every recurring task you do and honestly ask which of them follow a predictable pattern. Anything that follows a pattern is, in principle, automatable. This is not a comfortable exercise. But knowing which parts of your role are exposed lets you make a deliberate choice about where to build new skill rather than finding out reactively.

3. Make AI assisted output visible to your manager

If you use AI to draft a piece of commentary faster and then spend the time you saved improving the analysis underneath it, say so. Not in a performative way. Just make clear that you are working differently and that the quality is there. This is partly about career management and partly about helping your team understand what is actually possible. Both matter.

Frequently asked questions

Will AI replace treasury analysts?

Not in the near term, but it will change what the role contains. The tasks most at risk are repetitive and pattern driven: data aggregation, first draft commentary, and templated variance analysis. The tasks that remain firmly human are regulatory judgement, accountability, and stakeholder management under pressure. The analysts best placed are those who can do both: understand the domain and use the tools.

What finance jobs are most at risk from AI?

Roles where the majority of the work is data gathering, formatting, and producing structured narrative from predictable inputs face the most pressure. Junior reporting roles, certain reconciliation functions, and templated commentary work are the areas changing fastest. Senior roles that involve regulatory judgement, relationship management, and decision making under ambiguity are much less exposed.

Do I need to learn to code to stay relevant in finance?

Not necessarily, but some level of technical fluency helps. Being able to use AI tools effectively, understand what a Python script is doing, or build a basic automation without relying on IT is increasingly a differentiator. You do not need to be a developer. You need to be able to direct and quality check the tools that are becoming standard in the workflow.

The takeaway

The finance professionals who will find the next three to five years difficult are not the ones who engaged with AI and found it complicated. They are the ones who waited to see what happened.

AI will not eliminate most finance roles in the near term. But it will reshape what those roles contain, and firms will notice who is adapting. The professional who understands the domain deeply, can use tools that were not available two years ago, and can explain and defend the outputs they produce is not threatened by this shift. They are well positioned for it.

The mindset shift required is not enthusiasm for technology. It is recognising that your competitive position is not static, and that building new capability now is a form of career risk management. That framing is more honest than either the hype or the dismissal, and it is the one that produces useful action.