AI in tax and legal: What works, and what does not (yet)

Where AI genuinely delivers

Let’s start with what works. AI excels where tasks are repetitive, data-rich and rule-based. One important attention point: human review remains important, and even critical when it comes to numerical outputs.

Contract and document review is one of the clearer wins. AI tools can scan thousands of pages of agreements, flag relevant clauses and surface anomalies in a fraction of the time a human team would need. In legal due diligence and tax treaty analysis, this translates into real value and meaningful efficiency gains, while also reducing the risk of human error.

Tax data extraction and reconciliations are two more strong use cases. Pulling figures from invoices, ERP systems or financial statements for a company’s VAT returns, CIT computations or transfer pricing files? AI handles this well, especially when the data is structured and consistently formatted.

Management reporting and dashboarding leverage the consolidation and visualisation capabilities of AI. The creation of dashboards can now happen in a matter of minutes, allowing a clear focus on what truly matters and diving deeper on these insights where required.

AI is also proving its value in research assistance. Large language models can rapidly summarize case law, regulations or rulings. They can also support the creation of dashboards and presentations, turning raw analysis into ready-to-review drafts. This allows advisors to work from a solid first draft, rather than a blank page. The key word here: first draft.

Where the human remains in the loop

Here’s the honest part. AI does not replace professional judgment, and in Tax and Legal, judgment is at the core.

Complex tax structuring and planning requires an understanding of intent, business context and the regulatory landscape (including its grey zones). AI models trained on past data will not anticipate how a tax authorities could interpret a novel structure tomorrow.

Litigation strategy and advocacy also remain human territory. Persuasion, influencing, reading a rom and building arguments from sparse facts – all these require creativity and context that AI cannot reliably replicate just yet.

Finally, perhaps the most critical, accountability. When an AI-generated position turns out to be wrong, someone still has to sign off on it. The professional responsibility remains, whether or not a machine suggested the answer.

What matters going forward

The most effective Tax and legal professionals are not asking “Will AI replace me?”, they are asking “What can AI take off my plate, so I can focus on what requires my knowledge and experience?”.

That shift is where the real opportunity lies. AI is a powerful co-pilot. But for now, we are still the pilots.

At Deloitte, that’s exactly the conversation we’re having, and we’d love to bring you into it.