.webp)
The conversation about AI and jobs in finance is being framed the wrong way around. Here is what I have actually seen across fifteen years of working with finance teams, and why headcount is the wrong number to be watching.
TL;DR
AI is changing what finance teams do, not how many of them are needed. In the teams I have worked with as a fractional CFO, nobody has become less busy after introducing AI. The expectations, the timelines, and the depth of analysis have all moved up. The right question is not how many people AI will replace. It is what your team can now achieve that they couldn't before. Selling AI-native software on headcount savings misses the point entirely, and reduces a structural opportunity to a cost-cutting exercise.
The question being asked is the wrong one
There is a lot of noise about AI replacing finance jobs. For certain narrow tasks, that is genuinely happening. For finance roles as a whole, I don't think it is, and I don't think it should be.
After fifteen years working with SME finance teams as a fractional CFO, the pattern I have seen with AI is consistent. The teams that have started using it well have not got smaller. They have got more ambitious. The reports they produce are richer. The analysis is faster. The decisions land sooner. The people doing the work have shifted from processing to thinking, which is the work they trained for in the first place.
The goalposts have moved. We shouldn't be asking what jobs AI will eliminate. We should be asking what your team can now achieve that they couldn't before.
Lots of people have advised me that the best way to sell Finzu is to quantify hours saved and attach a headcount cost. I disagree. That is the old way of thinking. The right way to think about modern AI-native software is what it lets your existing team do that they couldn't do before.
Why the headcount-savings pitch is a trap
Software sold on headcount savings has a short half-life. The buyer eventually works out that the savings either didn't materialise (because the work expanded to fill the time) or did materialise but cost them a team member with institutional knowledge they now miss. Neither outcome creates a happy customer.
More importantly, headcount-savings framing pushes finance back into the cost-centre corner it has been trying to climb out of for years. Finance teams that focus on transaction processing get measured on cost per transaction. Finance teams that focus on commercial impact get measured on the decisions they enable. The first category gets cut in a downturn. The second category gets resourced.
AI is the first tool that genuinely shifts finance from the first category into the second. Pitching it as a way to do the cost-centre work more cheaply is the wrong sale.
Where AI genuinely earns its place in finance
Finance is the least automated business function despite possibly having the most opportunity for automation. That gap is what makes the next few years interesting.
In the teams I have worked with, AI is delivering the most value in the places where finance has always been stretched thinnest:
- Pulling structured insight out of unstructured documents (contracts, invoices, supplier statements) at a speed humans simply cannot match
- Spotting patterns in transactional data that a person scrolling through a spreadsheet would never see
- Drafting first-pass commentary on management accounts so the FD spends time refining rather than starting from scratch
- Surfacing the questions that a board pack should answer before anyone has asked them
None of this is replacing the finance team. All of it is making the finance team more useful to the rest of the business.
But AI shouldn't be the answer for everything
It is worth being honest about where AI doesn't belong. Not every workflow needs AI. Some need straight automation, which is cheaper, safer, and faster. Some need a human, which is unfashionable but true.
There are real costs attached to AI: usage fees that scale with volume, energy consumption that scales with usage, and the more dangerous cost of believing an output that is confidently wrong. Language models are probabilistic by design. For anything where the answer must be reproducible (arithmetic, reconciliation, tax) AI on its own is the wrong tool. Deterministic code does that work, and AI sits around it to interpret and communicate.
We have thought carefully about where to use AI in Finzu, and where to invest in well-structured automated workflows instead. The principle is straightforward: where there is no margin for error, 1 + 1 must always give you 2.
What this looks like in practice
The teams that get this right are reorganising around the work AI is good at, not shrinking around the work it can do. That means finance staff spending less time on data entry and reconciliation, and more time on commercial analysis, scenario planning, board-level reporting, and the early-warning work that prevents the problems that show up six months later.
The teams that get this wrong are using AI to maintain the same outputs with fewer people. They will be flat-footed in eighteen months when their competitors are operating at twice the analytical depth on the same headcount.
Junior finance roles still matter, perhaps more than before. The training pathway from junior to senior happens through the volume work AI is now doing, which means the way we develop the next generation of finance professionals has to change too. That is a real problem the profession will need to solve. It is not a reason to skip the AI transition.
Why this matters for UK SMEs right now
This isn't a theoretical conversation. The scale of the financial-management problem in UK SMEs is staggering, and AI in finance is one of the few interventions that could materially shift it.
The Small Business Commissioner reports that 82% of SME failures are due to poor financial management. The Federation of Small Businesses puts the cost of late payments to SMEs at £22,000 a year on average, and counts 38 businesses failing every day because of late payments alone. These are not problems caused by a shortage of finance professionals. They are problems caused by finance professionals being too stretched on the processing side of the work to do the proactive side.
AI in finance, used properly, is a way to free that capacity. The opportunity isn't to remove finance staff from UK SMEs. It is to redeploy them to the work they should have been doing all along.
The bottom line
If you are evaluating an AI-native finance platform, the question to ask the vendor is not how many hours their software will save you. It is what your team will be able to do with the time it gives back. If the answer is just "the same work, with fewer of us", you have found a cost-cutting tool. If the answer is "work that has been on the wishlist for three years and never got to the top of the pile", you have found something that will actually matter.
Finance is one of the last business functions to get this transition. It will also be one of the most transformed by it. The teams that lead the change will be more valuable, not smaller.
Frequently asked questions
Will AI replace accountants and finance professionals?
Not in general, no. AI is replacing certain narrow tasks within finance (data entry, basic categorisation, first-draft document review) and is dramatically expanding what finance teams can do beyond those tasks. The teams that adapt move from transaction processing into commercial analysis and decision support. The role changes. The need for finance expertise does not go away, and arguably increases, because someone has to validate what the AI produces and translate it into business action.
What finance tasks is AI actually good at right now?
Extracting structured data from unstructured documents (invoices, contracts, supplier statements), categorising transactions at scale, drafting first-pass commentary on management accounts, surfacing anomalies in transactional data, and answering ad-hoc questions about financial performance in plain language. Anything where the answer must be deterministic and reproducible (the actual maths, the actual reconciliation, the actual tax filing) is better handled by code with AI sitting around it, not by AI on its own.
How should SME founders think about AI in their finance function?
Start by asking what your finance team can't currently get to, not what they currently do that could be cheaper. The biggest wins from AI in finance come from unlocking work that previously didn't happen at all: weekly cash forecasts, supplier behaviour analysis, scenario modelling, proactive board reporting. If you frame the question as cost-cutting, you will under-invest and get a small return. If you frame it as capability expansion, you will get a finance function that materially helps the business grow.
Does AI in accounting work for small businesses, or only large ones?
It works for both, but the value distribution is different. Large finance teams get the most from AI's ability to handle volume and identify patterns across millions of transactions. Small businesses get the most from AI's ability to give them, for the first time, the kind of financial intelligence that only larger businesses could afford to produce. For an SME without a full-time FD, AI in accounting is the closest thing to having one in the background.
What is the biggest risk of using AI in a finance function?
Trusting outputs that look confident but are wrong. Language models are designed to produce plausible answers, not correct ones. In finance, where decisions are made on the numbers, that is dangerous unless the AI is surrounded by deterministic checks and human oversight. The fix is architectural: the calculations themselves should not run through the AI, only the interpretation and communication of them. Anyone selling you a finance AI without being able to explain how it handles this distinction is selling you risk.
