All notes

Is AI transforming management accounting, or just speeding it up?

Management accounting4 min read

A ledger grid on the left feeding into a network of connected nodes that ends in a question mark

Most writing on AI in accounting falls into one of two camps: it will replace us, or it will free us to be strategic. I wanted to know what the research actually supports, so I wrote a critical review of the academic literature on AI in management accounting.

My conclusion: AI is conditionally valuable. It improves efficiency fairly reliably. Whether it changes what management accountants do depends on things that have little to do with the technology.

The case for transformation

The optimistic argument, made most fully by Zhang et al. (2025), is that AI widens the scope of management accounting. It can bring in large volumes of unstructured and non-financial data, and support predictive, real-time decisions instead of backward-looking budget and variance reports. Forecasting tools such as Workday Adaptive Planning are the usual examples.

That would be a real shift: away from reporting what happened and towards advising on what to do.

Why it is more gradual than that

Other researchers are less convinced. Appelbaum et al. (2017) argue that analytics improves existing practice rather than redefining the function. Alruwaili and Mgammal (2025) find that infrastructure, process misalignment and adoption problems limit the effect, so change is incremental.

It is also worth noting that AI in budgeting and planning is at an earlier stage than AI in audit and financial reporting. A lot of the confidence is ahead of the evidence.

Efficiency is not the same as strategic value

The main reason organisations adopt AI is efficiency: automating routine work, cutting delays. That part is well supported.

But freeing up time does not automatically produce better decisions. The research is consistent that strategic value appears only when the organisation changes around the tool: management commitment, investment in training, and AI output being built into planning rather than sitting beside it. Without that, the result is the same work done faster.

More data can mean more error

This was the point I found most relevant to my own work. It is tempting to assume more data gives better answers. Richins et al. (2017) argue the opposite: without strong governance and human oversight, more data can increase errors rather than reduce them.

Two problems come up repeatedly:

  • Data quality. Poor inputs produce confident but wrong outputs.
  • Opacity. If an accountant cannot see how a figure was produced, they cannot exercise professional judgement over it. Moll and Yigitbasioglu (2019) warn that advanced analytics can hide the key assumptions.

Anyone who has reconciled two systems that should agree and do not will recognise this. The output is only as good as the data behind it, and someone has to understand that data well enough to challenge it.

Skills matter, but not only technical ones

There is broad agreement that accountants need to understand AI well enough to question it. But one study of the public sector in Kuwait (Al Wael et al., 2023) found that organisational culture, regulatory support and whether people see the tools as useful mattered more for adoption than technical skill alone.

There is also a quieter risk. Chai et al. (2025) suggest that heavy reliance on AI can erode professional capability over time, even when the complex, high-risk decisions stay with humans.

What the models miss

Much of the research uses two frameworks to explain adoption: Technology-Organisation-Environment (is the organisation ready, is there external pressure?) and the Technology Acceptance Model (do people find it useful and easy?). They are helpful, but studies that combine them still do not capture internal politics, ethics or governance. Those are often what decides whether a system gets used properly.

Where I landed

AI reshapes the management accountant's role rather than replacing it. Routine bookkeeping and reporting are automated; judgement, strategy and governance stay human. That matches ACCA's position that AI changes responsibilities rather than removing the professional.

For accountants, the practical implication is to build the skills to interrogate AI output, not just to operate the tool. For organisations, the benefits follow investment in data quality, training and oversight. They do not arrive with the software.

What I took from it

Writing this made me more careful about my own use of AI tools. I use them to help build Excel formulas and to speed up analysis. The research is a good argument for two rules: check the output against something that can be verified independently, and do not report a number you could not explain.

Based on an essay I wrote during my BA (Hons) Accounting and Finance at the University of Greenwich. The full reference list is available on request.

Written by Shashikala Mothe. Views are my own; examples use public or synthetic data only.

Back to all notes