By Christian Furness, published 6 February 2026
Artificial intelligence has arrived in audit and tax with a crack of thunder. Data that once took weeks to analyse can now be processed in minutes. Entire populations of transactions can be scanned, compared, and flagged with remarkable speed. For a profession long burdened by volume, repetition, and compliance, AI feels less like a tool and more like a force of nature.
And it is great. Truly great.
But lightning, for all its power, is not something you hand the reins to and walk away from. In accounting, audit and tax, AI must be ridden - not worshipped, not feared, and certainly not left to run unattended. Humans are still essential, not as a sentimental attachment to tradition, but because the profession itself demands judgment, accountability, and scepticism that technology cannot replicate.
What AI Is Doing Brilliantly
Let us start with what AI gets right. In practice and industry alike, AI excels at the heavy lifting accountants have long endured.
It can extract data from invoices and contracts, reconcile accounts at scale, identify anomalies in journal entries, and aggregate information across systems that were never designed to speak to one another. In tax, it can assist with return preparation, data validation, and the mechanical application of rules. In audit, it enables population-level testing and more targeted risk assessment.
Used well, AI frees professionals from drudgery and creates space for higher-value work. It is faster, more consistent, and far less prone to fatigue than even the most diligent junior associate. Ignoring these benefits would be both impractical and irresponsible.
But recognising AI’s strengths does not require pretending it has no weaknesses.
Where the Hype Outruns Reality
The problem arises when efficiency is mistaken for understanding.
AI does not comprehend economic substance. It does not understand why a transaction occurred, only that it resembles others it has seen before. It cannot grasp management intent, commercial pressure, or the subtle incentives that sit behind accounting decisions. In audit, it may flag what is unusual but fraud and error are often depressingly routine. In tax, it may generate confident-sounding advice that collapses under jurisdictional nuance or a single overlooked anti-avoidance rule.
This is where misapplication creeps in.
Too often, anomaly detection is treated as audit coverage rather than a starting point. AI-drafted tax analysis is accepted with insufficient scepticism. Judgments on materiality, going concern, and valuation assumptions are quietly nudged towards standardisation because the model prefers neatness over reality.
The danger is not that AI will get everything wrong. The danger is that it will get enough right to lull professionals into false comfort.
The Limits That Matter
There are also limits that no amount of model refinement will remove.
Accountancy, audit and tax are accountability professions. Someone must sign the opinion. Someone must defend the position to regulators, courts, or revenue authorities. Someone must explain not just what was concluded, but why.
AI cannot be cross-examined. It cannot assume legal liability. It cannot exercise professional scepticism or challenge a persuasive but incomplete management narrative. Nor can it reliably explain its own reasoning in a way that satisfies regulators who demand transparency and defensibility.
Data quality compounds the problem. Accounting data is rarely clean, complete, or neutral. It reflects human systems, human decisions, and human biases. Feeding this into AI without rigorous oversight risks amplifying errors rather than correcting them.
These are not temporary obstacles. They are structural features of the profession.
Why Humans Still Matter
This is where the human accountant comes back into focus, not as a processor of information, but as an interpreter of reality.
Judgment is not an inconvenience to be automated away. It is the point. Professional scepticism, contextual understanding, ethical responsibility, and the ability to say “this does not make sense” are not optional extras. They are the foundation of trust in accounting, audit and tax.
Ironically, as AI takes over more mechanical work, human skills become more, not less - important. The ability to evaluate evidence, challenge assumptions, explain conclusions clearly, and take ownership of decisions will define professional value in the years ahead.
AI may illuminate patterns at lightning speed, but humans decide where to steer.
Riding the Lightning
The future of accounting, audit and tax is not a contest between humans and machines. It is a partnership - albeit an asymmetric one. AI supplies speed, scale, and pattern recognition. Humans supply judgment, accountability, and meaning.
Those who succeed will be the professionals who learn to ride the lightning: to harness AI’s power without surrendering responsibility, to question its outputs rather than defer to them, and to remember that consistency is not the same as correctness.
AI is here to stay, and that is a good thing. But in a profession built on trust, it is still people, not algorithms, who must hold the reins.
