What the accountancy function looks like in 2030: I was the only person in the room who didn't say 'a bit more efficient'
Published 19 May 2026 · Updated 14 August 2026
What does the accountancy function look like in 2030?
It is a fair question, and the answer most of the profession is giving is the wrong one. The dominant view is that the function in five years looks like the function today, run faster. Same close. Same reports. Same workflows. Just with more AI bolted on the top to take some of the load.
I was in an FP&A peer group recently when the question came up. Everyone in the room described some version of “a bit more efficient.” Reconciliations done quicker. Variance commentary drafted by a model. Reports out a day or two faster. I was the only voice arguing for something different. I left that room with an honest question: am I too pro-AI, or is the profession collectively underestimating what is coming?
This is the field note behind the fuller finance function of 2030 framework. That article sets out the operating model and skills. This one is about why experienced finance people can look at the same evidence and reach very different conclusions.
This post is the answer to that question.
What “a bit more efficient” actually means
The incremental view is not stupid. It is the dominant one, and there are sensible reasons for that.
It assumes the architecture of finance work stays roughly where it is. The general ledger still works the way it works. The month-end close still happens because data still accumulates in batches. Management accounts still get produced by humans, with AI helping at the edges. FP&A still spends most of its time pulling the numbers together and the rest of it explaining what they mean.
In that picture, AI shows up as a productivity layer. Faster reconciliations. AI-drafted commentary. Automated variance flags. Copilot in the spreadsheet. Real gains, but additive ones. The function is the same shape. It is just running on better tools.
I understand why this is where most of the profession sits. Most finance professionals’ working experience of AI to date has been exactly that: a model that helps with Excel and a model that drafts narrative. If your direct experience is “useful assistant,” your forecast for 2030 is “more useful assistant.” That is rational extrapolation from the inputs.
It is also, I think, wrong.
What changes when the inputs change
The reason 2030 is not “a bit more efficient” is that the inputs to the function are changing, not just the processing speed inside it.
Cleansed, governed data at source. Continuous accumulation rather than periodic batch. Agents that act on that data, not just summarise it. None of these are speculative. They are already legible in the businesses that have done the foundation work. The data quality post covers why this layer is the precondition for everything else.
Once those inputs move, the shape of the work moves with them.
The monthly close stops making sense as a discrete event. It exists because financial data has historically been accumulated in batches and processed in cycles. When the accumulation is continuous and the processing is automated, the close stops being a thing the team does and starts being a state the system is in. Most days. Most of the time.
FP&A stops being report production and becomes decision orchestration. When the numbers are already there, on demand, the question is no longer “can we get them” but “what decision needs them and what is the answer.” That is a different job. It uses different skills. It rewards different people.
Audit reshapes around continuous attestation. If the data is governed at source and the processing is observable, the annual statutory audit becomes a check on a system that has already been checking itself. The shape of the audit relationship changes. So does the role inside it.
The team gets smaller, more senior, and spends its time on judgment, not collation. Not because anyone is being replaced for the sake of it, but because the collation layer is the bit that the inputs above have eliminated. I covered the longer, balanced version of this in The Future of the Finance Function. That post lays out what changes and what stays human. This one is the more opinionated companion.
The shorthand: the function in 2030 is not faster. It is differently shaped.
Why the room saw it differently
I have thought about this a lot since the meeting, and I think there are three honest reasons the consensus in that room landed on “a bit more efficient.”
The first is what I covered above. Most professionals’ lived experience of AI in finance to date is the productivity-layer version. Linear extrapolation from that experience gives you a linear answer. You see a faster version of what you already do.
The second is that the structural shift requires data-layer work that most organisations have not started. If you have not seen what cleansed data at source enables, you cannot picture what runs on top of it. The image you can form is bounded by the image you have already seen. Most finance functions are still running on data that needs to be wrestled into shape before it can be used. From inside that experience, “continuous, governed, agent-ready data” sounds like marketing copy. It is not. But it requires sitting inside a function that has done the work to see what it does.
The third is harder to say and worth saying. There is a professional cost to being wrong loud. Saying “the close goes away” in a peer group is riskier than saying “we will get faster.” Caution sounds more credible. It is not necessarily more correct. The most measured forecast is not automatically the most accurate one. It is just the one that is hardest to be embarrassed by later.
Am I too pro-AI?
This is the question I went home with, and the one I want to answer properly.
Let me steel-man the case that I am wrong.
AI in finance has overpromised before. The data foundations are harder than the demos suggest. Regulatory and audit frameworks move slowly, and the function’s shape is partly defined by those frameworks. The legacy ERP estate inside most businesses is not going anywhere by 2030. Junior accountants are still being hired and still need to be trained on something. The why junior accountants are not being replaced piece makes that case in full. All of this is true.
None of it changes the structural argument. What it changes is the timeline, and the distribution across businesses.
By 2030, the median business will probably still have a recognisable close, a recognisable FP&A team, and a recognisable audit. The radical picture will not be the average picture. The radical picture will be the picture in the best-prepared organisations, and the gap between those organisations and the rest is what will define the next five years of the profession.
So no, I do not think I am too pro-AI. I think the profession is collectively under-prepared for the version of the technology that is already legible, never mind the version that is coming. The error I am most worried about is the cautious one, not the enthusiastic one.
What both sides of that room should do this year
Here is the part that matters more than who is right.
If the incremental view turns out to be correct, the cost of preparing for the radical version is a year of data quality work, process documentation, and governance design. That work pays back regardless. None of it is wasted on a finance function that ends up “a bit more efficient” instead of structurally redesigned.
If the radical view turns out to be correct, the cost of not preparing is being structurally outclassed by the businesses that did. A finance function that arrives at 2030 still running a manual close because nobody invested in the data layer is not five percent behind. It is a generation behind.
The asymmetry is not balanced. The downside of being too forward is small and recoverable. The downside of being too cautious is large and hard to recover from. That is the actual decision in front of every finance leader this year.
The specific work, in order: invest in data quality at source, document the process as it actually runs rather than as the diagram says it runs, build the governance for agentic deployment before the agents arrive (see the agentic AI piece for the detail on what that means in practice), and develop the team’s judgment and AI literacy in parallel with the technology, not after it.
Where I land
I do not need the peer group in that room to agree with me. I do need the businesses I work with to be ready, because the function in 2030 is not a faster version of today’s function. It is a different one.
The work to be ready for it does not start in 2029. It starts now. The organisations that begin it this year are not being too pro-AI. They are being early to a structural change the rest of the profession has not priced in yet.
That is a more comfortable place to be than the alternative.
Maebh Collins is a Fellow Chartered Accountant (FCA, ICAEW) with Big 4 training and twenty years of operational experience as a founder and senior finance leader. She writes about AI in finance transformation from the inside out.