Tax Doesn’t Have an AI Problem. It Has a Data Problem.

Perhaps it will. But after years of speaking with Heads of Tax, Tax Directors and tax professionals across the UK, Europe and North America, I’ve become convinced that most tax functions don’t have an AI problem. They have a data problem.

Tax is increasingly expected to produce sophisticated outputs from data it had little say in designing, capturing or governing. That’s not a new challenge. It’s one many tax leaders have been living with for years – fragmented systems, inconsistent processes, multiple ERP environments, conflicting data sources, and reporting obligations that keep growing. AI is now being layered on top of all of it, as though it were the answer to problems that predate it entirely.

AI has real potential. It’s just not the immediate priority

To be clear, AI has genuine value within tax. Automating repetitive processes, reviewing large volumes of information, flagging anomalies, supporting compliance – all of that is compelling, and plenty of organisations are already exploring it.

But AI is only as good as the data behind it. For many tax functions, that data is still fragmented, inconsistent, or hard to access. Before AI can transform tax, tax teams need control of their data – that’s the actual next step. AI doesn’t solve data problems. It amplifies them. Feed it inconsistent or incomplete information, and it will simply deliver answers faster, not better.

A different battle

When I speak with tax leaders, very few say their biggest issue is a lack of AI capability. Far more are dealing with data spread across multiple systems, inconsistent reporting between jurisdictions, manual adjustments, legacy processes, and information requests arriving faster and in greater volume than ever. None of this is new. AI hasn’t created these problems – it’s simply exposed them.

EY recently found that 80% of respondents believe their data isn’t ready for AI adoption, which says a lot about how far many tax and finance functions still have to go before AI can deliver anything meaningful. Most tax leaders don’t need convincing of that. They’ve spent years dealing with the fallout of finance-led transformation programmes, inconsistent data capture, and acquisitions that brought in new systems faster than anyone could rationalise them. The problem was never recognising the issue – it’s been getting the wider business to prioritise fixing it.

When reporting obligations increase, through Pillar Two, tax governance requirements or growing scrutiny from tax authorities, the expectation is usually that tax will just find a way. Often that means spreadsheets, workarounds, and very talented people making do.

The ERP myth

One assumption I hear constantly: “We’re implementing a new ERP, so our data problems will be solved.” Tax professionals know it’s rarely that simple.

Take a multinational running SAP in Germany, Oracle in the US, different local finance systems across Europe, and legacy platforms inherited through acquisition. The data exists, but is it captured consistently? Can it be trusted? Can it be compared across jurisdictions? Those are very different questions from “do we have a new system.”

I’ve seen organisations spend millions on ERP transformation only to find tax teams still relying on spreadsheets and manual reconciliations, because legal entities, reporting structures and local requirements were never properly aligned during implementation. Technology doesn’t fix a governance problem – as we’ve written about previously, data quality is consistently the make-or-break factor in AI and transformation success.

KPMG’s research backs this up – many tax functions still spend significant time just making their data usable, with quality, accessibility and format issues remaining major obstacles. Before asking whether AI can automate tax, the better question is whether you trust the data you’d be feeding it.

Pillar Two has exposed the problem

If one development has laid this bare, it’s Pillar Two. Most tax professionals understand the rules well enough. The real challenge has been obtaining the information needed to apply them – pulling together data from multiple systems, jurisdictions and processes, much of it never captured with tax reporting in mind. It isn’t a legislation problem. It’s a question of whether the underlying information can support what’s being asked of it. If humans are already struggling with that, AI won’t solve it overnight.

There’s no single dashboard

Another common assumption is that there must be one global solution – usually well-intentioned. Finance wants consistency, leadership wants visibility, the CFO wants a dashboard. But tax operates in a world of local complexity.

A UK team, a Luxembourg team and a US team may all need different information, at different times, for different regulatory purposes. Building a dashboard is the easy part. Standardising the data behind it is significantly harder, and AI doesn’t remove that complexity; it just surfaces it faster.

What AI is really exposing

I think AI is exposing the cumulative cost of decisions made years ago – around ERP implementation, data ownership, process design, systems architecture and accountability. Most of those decisions were made with finance, operations or technology priorities in mind, which is understandable.

Tax often became an afterthought, not deliberately, but because it came late in the process. The result is that many tax functions are now trying to meet increasingly complex reporting requirements using data structures that were never designed for tax in the first place. AI didn’t create that weakness. It’s just made it harder to ignore.

This is where organisations need to be careful. CFOs are under real pressure to drive efficiency and adopt new technology, and AI naturally becomes part of that conversation. But AI can make a governance problem look like a technology problem. A dashboard is only as good as the information behind it. A model is only as good as the assumptions feeding it. Tax leaders are being asked for more insight and more strategic input, often using data and systems they don’t own and had little say in when they were built.

The organisations making the most progress aren’t necessarily the ones talking loudest about AI – they’re the ones fixing the fundamentals: data governance, process consistency, clearer ownership of tax data, and closer collaboration between Tax, Finance and Technology. Deloitte’s research points the same way, with leading tax functions continuing to prioritise data and operating model improvements as the real foundation for AI adoption. Getting the basics right first is, frankly, just sensible.

What this means for tax professionals

One of the more interesting shifts I’ve seen in the recruitment market is that clients are no longer looking solely for technical tax expertise. Five years ago, being technically strong was often enough. Increasingly, clients want people who can bridge tax, finance, data, technology and business operations. Conversations about ERP exposure, data governance and process improvement come up far more often than they used to – not because organisations want fewer specialists, but because they need people who understand how tax fits into the wider business.

That’s also reshaping what it means to be a Head of Tax. Technical capability still matters, but it’s no longer enough on its own. Increasingly, tax leaders are expected to influence technology investment, challenge finance processes, improve data governance and support commercial decision-making – the same shift we’ve seen play out in what successful finance transformation actually depends on. Many of the best Heads of Tax I work with spend surprisingly little time talking about tax itself – they’re talking about strategy, risk, operations and systems. They’re becoming business leaders who happen to specialise in tax, and I expect that trend to accelerate.

The question worth asking

I sometimes think we’re framing the AI conversation the wrong way round. The question tends to be “how will AI transform tax?” A more useful one might be “what problems are we actually trying to solve?”

Most Heads of Tax aren’t losing sleep over whether AI can draft a memo. They’re worried about rising compliance demands, resource constraints, data quality and future-proofing their function. The most forward-thinking leaders I speak to aren’t asking how to replace people with AI – they’re asking how to free their people from low-value work so they can spend more time on judgement, risk and business partnering. That’s a very different conversation.

Final thought

A lot of AI discussions treat tax as primarily a process. It isn’t. Compliance can be automated, calculations can be automated, and data analysis can be accelerated, but the best tax professionals I’ve worked with were never valuable because they processed information faster. They were valuable because they understood what the information meant, recognised risk, challenged assumptions, and knew when something looked technically correct but was commercially wrong.

Tax has never really been about producing an answer. It’s about knowing whether the answer makes sense.

The organisations that succeed with AI won’t be the ones that adopt it fastest – they’ll be the ones that build the strongest foundations first: strong data, strong technology, and tax professionals who know how to challenge, interpret and influence.

The real question isn’t whether AI can do tax. It’s whether we’re underestimating the people who know when the numbers don’t tell the full story – because the organisations that benefit most from AI may not be the ones with the best AI strategy. They may simply be the ones with the best data strategy.

If you’d like to discuss how this is shaping tax hiring, or talk through a specific challenge in your own function, I’d be happy to continue the conversation. Get in touch at [email protected]