Why the legal and tax advisory market is placing less value on information retrieval and document production – and more on professional judgment, evidence analysis, accountability and strategy

Ivars Mileika  |  September 2026

The trajectories of the legal and tax advisory professions are beginning to converge. In both fields, artificial intelligence is first taking over not responsibility for the final decision, but the work that has traditionally consumed much of a professional’s time: information retrieval, document comparison, initial analysis, summaries and first drafts.

This matters because the business model of professional services has historically relied to a significant extent on the client lacking either access to information, the time to process it, or the expertise to structure it. AI is reducing that information asymmetry. Increasingly, clients can reach a sufficiently good initial result on their own.

The key question is therefore no longer whether AI will replace lawyers or tax advisers. A far more practical question is this: which parts of a professional’s work will clients still be willing to pay a full advisory fee for in a few years’ time?

1. AI is no longer merely a tool. It is becoming professional-services infrastructure

Professional-services research in 2026 shows that AI adoption has moved beyond the experimental stage. In Thomson Reuters’ global survey, 74% of professionals said they use AI several times a week, while 44% use it several times a day. The study covered 1,816 professionals across legal, tax, audit, accounting, compliance and other professional-services fields in 62 countries.^[1]

A separate 2026 study on professional services found that organisation-wide use of generative AI had nearly doubled in one year, from 22% to 40%. More than 80% of current users employ AI at least weekly, and more than 90% expect it to become a central part of workflows within five years.^[2]

For law firms, this means legal research, document review, summaries, memoranda and contract drafting. In tax and accounting, common use cases already include tax research, document analysis, accounting tasks, preparation of tax advice and even tax-return preparation.^[3]

2. Tax advisers are under pressure from both sides

The position of tax advisers is distinctive because technology is reshaping the market from two directions at once.

On one side, the client is becoming technologically stronger. A company’s CFO or accounting function can use AI to test the VAT treatment of a transaction, analyse transfer-pricing documentation, compare tax rules across jurisdictions or prepare a first draft of arguments in a dispute with the tax authority.

On the other side, the tax authority itself is becoming technologically stronger. In the long term, this may prove even more important for the tax advisory profession than the spread of ChatGPT or other publicly available AI tools.

OECD data show that in 2023, 69% of surveyed tax administrations were already using AI and a further 24% were implementing it for future use. By comparison, only 9% reported using AI in 2016. The OECD describes applications in analytics, improved case selection, taxpayer services and the automation of high-volume repetitive activities.^[4]

This changes the balance of power. Tax administrations increasingly have not only statutory authority and access to returns, but also the ability to compare vast datasets automatically, identify anomalies and select risks faster than is possible through manual review.

3. Standardised tax compliance work will lose pricing power first

Within the tax advisory market, the work most exposed to automation is work where outputs can be derived from structured data and a relatively clear rule set: preparing returns, reconciling data, conducting standard checks, assembling documentation and producing an initial analysis of routine tax questions.

Thomson Reuters’ 2026 data show that among tax and accounting professionals using AI, 69% use it for tax research, 57% for document summarisation, 55% for document review, and 53% for accounting, tax-advisory and tax-return preparation tasks.^[3] A separate 2026 study found that 81% of tax and audit firm professionals were already using AI regularly in their day-to-day work.^[5]

This does not mean that a tax return or an advisory opinion is produced without human involvement. It means something else: the time required to reach the first 70-80% of the result is falling sharply. In a market where fees are still often justified by hours worked, that becomes a business-model problem.

There is little economic rationale for a client to pay for five hours of junior professional time if a properly governed AI solution can perform the same data assembly and initial analysis materially faster. Yet the client may be willing to pay more for a professional who can determine whether the AI-generated answer is defensible in the specific transaction, how the tax authority is likely to interpret it, and what the consequences will be if the underlying assumption proves wrong.

4. Europe is moving from a world of returns to a world of data

The impact of AI on the tax profession cannot be assessed separately from the much broader digitalisation of tax administration. The European Union’s VAT in the Digital Age (ViDA) reform provides for the gradual development of cross-border B2B e-invoicing and digital reporting, while also simplifying VAT registration for cross-border transactions.^[6]

The economic logic is straightforward: the more transaction data a tax authority receives in structured and comparable form, the less it needs to wait for a traditional return before identifying an inconsistency. Tax control can increasingly take place while data are flowing, not only after the end of the tax period.

For tax advisers, this shifts the centre of gravity of the profession. Historically, a significant part of the value lay in preparing information correctly for submission to the administration. Increasingly, the higher-value task will be to ensure that the company’s systems contain the correct tax logic, data classification and control mechanisms from the outset.

5. Latvia: the State Revenue Service is moving towards a data-driven and automated analytics model

In Latvia, this transition is already practical. The State Revenue Service (VID) is modernising its risk-analysis infrastructure, using taxpayer segmentation and working increasingly with structured data. Since May 2026, the Electronic Declaration System (EDS) can automatically use received e-invoice data to populate part of the VAT reporting process. It is a modest but highly visible step from document submission towards data-flow administration.^[7]

VID has also tested artificial intelligence in preparing responses to taxpayer enquiries, while AI solutions for image and risk-data analysis are being developed in the customs area. These projects should not be overstated – publicly available information does not support the conclusion that AI already makes tax decisions in Latvia. The institutional direction is nevertheless clear: more data, more automated comparison and earlier risk identification.^[8]

For a tax adviser, this has one very practical implication: it is no longer enough to react to a VID question. The ability to anticipate what questions a client’s data may generate is becoming increasingly valuable.

6. In tax disputes, AI’s greatest value may lie in evidence analysis

A tax dispute is fundamentally different from day-to-day advisory work. Finding the correct statutory provision is often not enough. The decisive issue is what can actually be proved: what happened in the transaction, what the parties knew at the relevant time, whether the documents are internally consistent, and whether explanations given later correspond with contemporaneous invoices, contracts, accounting entries, bank payments, emails and other evidence.

Here, AI’s potential extends far beyond producing a document summary. A well-configured system can analyse thousands of documents at the same time, build a transaction chronology, connect people, amounts and dates, locate multiple sources relating to the same fact, and flag points at which those sources conflict. In tax litigation, this can change the methodology of case preparation itself.

AI can create a kind of “contradiction map”: the contract describes one transaction model, the invoice uses another description, the accounting records apply a different code, the bank payment narrative suggests yet another economic substance, while a witness describes the events differently three years later. In a large case, identifying such connections may take a person days; AI can surface them as verifiable signals within minutes.

The same applies to witness evidence. AI can compare a witness’s written statement with earlier email correspondence, meeting minutes, an interview or a hearing transcript and flag changes in dates, amounts, persons or sequence of events. But there is an important boundary: AI can identify a textual inconsistency; it should not automatically conclude that a witness is lying. The difference may be explained by memory, translation, the wording of a question or context. Assessing credibility remains a professional task and, in court, a judicial function.

For that reason, the most valuable AI output in a tax dispute may not be a finished legal opinion but an evidence matrix: fact – supporting source – conflicting source – legal significance – required verification. This approach allows the lawyer to understand more quickly not only what is in the case file, but also what is missing and where the opposing side is likely to attack.

In case-law analysis, AI’s advantage is its ability to search not only by keywords but also by factual pattern and structure of reasoning. It can compare dozens of judgments, extract the applicable legal rule, burden of proof, judicial criteria, distinguishing facts and the arguments that proved decisive in each case. In tax practice, this allows a much more precise question than “is there a similar judgment?”: in factually comparable cases, when did the court accept the taxpayer’s account and when did it reject it?

Yet case-law interpretation is also one of the most dangerous areas for AI use. UK tax tribunal decisions already provide concrete warnings. In Harber v HMRC, the taxpayer submitted nine AI-generated authorities that did not exist at all.^[14] In Zzaman v HMRC, the problem was more subtle: several cited cases were real, but they did not support the legal propositions for which they were relied upon. The tribunal stressed that responsibility for the accuracy of AI-assisted work remains with the human user.^[13]

There is also a positive example. In Evans & Ors v HMRC, a Tax Chamber judge expressly stated that a secure AI tool had been used to summarise voluminous documents, but treated those summaries only as first drafts, personally checked their accuracy and did not use AI for legal research.^[12] In my view, this captures a professionally safe model with unusual clarity: AI performs the large-scale comparison; the human retains control over the legal conclusion.

The EU AI Act reinforces this boundary at regulatory level. AI systems used by judicial authorities to research and interpret facts and law, or to apply the law to a concrete set of facts, are classified as high-risk systems; the recitals also emphasise that AI may support judicial decision-making, but the final decision must remain under human control.^[11] This does not make every AI tool used by a lawyer a “high-risk” system, but it illustrates how sensitive automated interpretation of facts and law is.

A new competitive advantage in tax disputes follows from this: not AI that drafts the longest pleading, but an AI workflow in which every material conclusion is traceable to a verifiable document, a specific paragraph of a judgment and a human professional assessment.

7. AI can prepare an answer. It cannot assume professional responsibility

In February 2026, CFE Tax Advisers Europe published a dedicated Charter of Tax Advisers’ Rights and Obligations in an AI-Influenced Tax-Advisory Environment. It highlights principles including professional autonomy, AI competence, transparency, data protection, professional integrity, objectivity and knowledge of the applicable regulatory framework.^[9]

One of its central principles is particularly important: AI may assist with analysis, but professional responsibility for the advice remains with the adviser. The adviser must be able to verify, challenge and, where necessary, reject the AI output. A similar principle of human responsibility and careful verification is emphasised in the 2025 AI guidance issued for the judiciary of England and Wales.^[15]

This is a powerful market argument in favour of high-quality professionals. A client does not need a human merely to rewrite an AI answer in more polished legal language. The client needs someone prepared to say: I have verified this position, I understand its risks, I recommend a particular course of action, and I am prepared to defend that conclusion professionally.

8. What will clients continue to pay for?

Information itself will become cheaper. Initial analysis will become cheaper as well. But that does not necessarily mean that the professional-services market will become smaller. It may simply become far more demanding.

AI can analyse thousands of case documents and identify dozens of inconsistencies. A strong tax disputes lawyer understands which inconsistency genuinely undermines the transaction narrative and which is merely a technical detail.

AI can compare what a witness says today with documents created at the time of the transaction. The professional must decide whether the inconsistency affects the reliability of the evidence and how it should be used procedurally.

AI can map hundreds of judgments in minutes. A strong lawyer distinguishes the true ratio of a decision from incidental observations, understands the procedural context and spots the factual difference that makes an apparently similar precedent unsuitable for the case at hand.

The professional’s value will therefore concentrate in four areas: judgment, accountability, strategy and trust. Technical expertise will not disappear – on the contrary, more of it will be required to verify AI output and defend a position. But expertise on its own will increasingly cease to be the final product.

9. The hourly-rate problem will become increasingly visible

AI creates an uncomfortable economic question for professional firms. If a task that once took ten hours now takes two, should the client still pay for ten?

In a study of law-firm clients, 71% of in-house legal professionals said they expect external law firms’ commercial models to change as AI use increases, while only 28% of law firms had changed their pricing structures.^[10] Sooner or later, the same pressure will reach the tax advisory market.

Under an hourly-rate model, technological efficiency creates a paradox: a firm invests in technology to complete work faster, but faster work produces a smaller bill. The market will therefore move towards fixed fees, subscription models, pricing based on outcome and risk value, or hybrid arrangements.

Clients will not want to pay for the time saved by AI. They will pay for the better result AI helps the professional deliver.

10. The profession’s greatest long-term risk may not be the number of jobs

There is another issue that particularly affects law and tax advisory firms. Junior work has historically been more than a low-cost resource. It has also been the profession’s training mechanism.

If a young lawyer no longer personally reviews hundreds of contracts, conducts tax research and drafts first versions of opinions, the question arises: how will that lawyer acquire, ten years from now, the judgment we currently expect from a partner or head of tax practice? AI can make a firm more productive in the short term while simultaneously narrowing the “training ground” on which professional experience is built.

Firms will therefore need to design a new professional-development model deliberately: less mechanical document production, more verification of outputs, argumentation, understanding of the client’s business, data analysis and decision-making.

11. What does this mean in practice for a legal and tax advisory firm?

Buying an AI licence is not enough. A firm’s competitiveness will depend on whether technology is embedded in professional methodology and quality control. In practice, this means six priorities:

• Automate low-value work, not professionalism. AI should be used for research, data assembly, comparison and first drafts, while the final conclusion must be professionally verified.
• Build a proprietary professional knowledge base. The competitive advantage will not be access to the same public model everyone else uses, but the firm’s own precedents, document templates, argumentation and sector-specific client experience.
• Build an evidence-analysis methodology for tax disputes. AI should be used for chronology, contradiction mapping, comparison of documents and witness evidence, and case-law mapping, but every material conclusion must be traceable to an original source and human verification.
• Move tax practice from correcting returns to preventive data control. Firms should be able to test clients’ VAT, corporate income tax, payroll tax and other tax risks before they are identified by VID analytics.
• Implement AI governance and demonstrable quality control. It must be clear which tools may process client data, who verifies the output and how human professional oversight is documented.
• Change the pricing conversation with clients. The invoice should increasingly reflect the complexity of the problem, the risk assumed, professional responsibility and value created for the client – and less the time spent producing a document.

12. Conclusion: the profession will not disappear, but part of its traditional product will

The debate over whether AI will “replace lawyers” or “replace tax advisers” is too simplistic. These professions consist of very different kinds of work, and their susceptibility to automation is not the same.

The work that will lose price and demand fastest is work that is repetitive, standardisable and based primarily on information retrieval or structured data processing. In tax disputes, however, AI may actually increase the capacity of a strong professional: it can build a factual chronology faster, review a large evidential record, identify contradictions and compare case law systematically. The ultimate value, however, lies in turning that information into a defensible theory of the case.

For tax advisers, this transformation is particularly sharp because AI is entering not only the adviser’s office but also the tax authority. OECD trends, European tax digitalisation and VID’s own development trajectory point in the same direction: tax administration is becoming more data-intensive, automated and proactive.

The best tax adviser or tax disputes lawyer of the future will therefore not be the person who finds a statutory provision fastest. It will be the professional who understands law, data, evidence and the client’s business at the same time – and who can substantiate, verify and professionally defend a conclusion.

Author’s note

This article represents the author’s analytical view of publicly available international and Latvian developments in professional services and tax administration. It does not reproduce the structure or sequence of reasoning of any single source; external data and institutional facts are identified by references. The conclusions and forecasts are the author’s own assessment.

Sources and notes

[1] Thomson Reuters Institute, Future of Professionals Report 2026. Global study: 1,816 professionals, 62 countries. https://www.thomsonreuters.com/en/institute/future-of-professionals-2026/report

[2] Thomson Reuters Institute, 2026 AI in Professional Services Report. https://www.thomsonreuters.com/en/reports/2026-ai-in-professional-services-report

[3] Thomson Reuters, 2026 AI in Professional Services Report – GenAI use cases by industry (Legal; Tax & Accounting). https://tax.thomsonreuters.com/content/dam/ewp-m/documents/thomsonreuters/en/pdf/reports/2026-ai-in-professional-services-report.pdf

[4] OECD, Tax Administration 2025 – The rise of artificial intelligence. https://www.oecd.org/en/publications/tax-administration-2025_cc015ce8-en.html

[5] Thomson Reuters Institute, Future of Professionals 2026 – Tax and Accounting Report. https://www.thomsonreuters.com/en/institute/future-of-professionals-2026/report-tax-and-accounting

[6] European Commission, VAT in the Digital Age (ViDA) – VAT digitalisation and implementation. https://taxation-customs.ec.europa.eu/taxation/vat/vat-digital-age-vida_en

[7] Latvian State Revenue Service (VID), public information on the modernisation of analytical solutions, taxpayer segmentation and the use of e-invoice data in VAT reporting. https://www.vid.gov.lv/lv/projekts/atveselosanas-fonda-reformu-pasakumi-un-investiciju-projekti ; https://www.vid.gov.lv/lv/jaunums/jaunas-eds-funkcijas-paplasina-e-rekinu-datu-izmantosanu-un-vienkarso-pievienotas-vertibas-nodokla-deklaresanu

[8] Latvian State Revenue Service (VID), public information on the AI pilot project for processing taxpayer enquiries and the development of AI functionality in customs solutions. https://www.vid.gov.lv/lv/iepirkums/pilotprojekts-maksliga-intelekta-izmantosana-vispareju-atbilzu-sagatavosanas-procesa-uz-nodoklu-maksataju-jautajumiem ; https://www.vid.gov.lv/lv/iepirkums/baxe-risinajuma-pilnveidosana-ieviesot-maksliga-intelekta-funkcionalitati

[9] CFE Tax Advisers Europe, Charter of Tax Advisers’ Rights and Obligations in an AI-Influenced Tax-Advisory Environment, 25 February 2026. https://taxadviserseurope.org/project/cfe-charter-of-tax-advisers-rights-and-obligations-in-an-ai-influenced-tax-advisory-environment/

[10] Thomson Reuters Institute, Future of Professionals 2026 – Legal Report (client expectations and pricing models). https://www.thomsonreuters.com/en/institute/future-of-professionals-2026/report-legal

[11] Regulation (EU) 2024/1689 (EU AI Act), recital 61 and Annex III – AI systems used in the administration of justice. https://eur-lex.europa.eu/eli/reg/2024/1689/oj

[12] Evans & Ors v Revenue and Customs [2025] UKFTT 1112 (TC), especially paras 45-48 – judicial use of AI for document summarisation with human verification, without using AI for legal research. https://www.bailii.org/uk/cases/UKFTT/TC/2025/TC09638.html

[13] Zzaman v Revenue and Customs [2025] UKFTT 539 (TC), especially paras 15-29 – AI-assisted submissions relied on genuine cases that did not support the legal propositions advanced; responsibility for verification remains with the human user. https://www.bailii.org/uk/cases/UKFTT/TC/2025/TC09520.html

[14] Harber v Commissioners for HMRC [2023] UKFTT 1007 (TC) – nine AI-generated, non-existent authorities were submitted in a tax dispute. https://www.bailii.org/uk/cases/UKFTT/TC/2023/TC09010.html

[15] Courts and Tribunals Judiciary, Artificial Intelligence (AI) – Judicial Guidance, October 2025 – human responsibility and the risks of hallucination, bias and confidentiality breaches. https://www.judiciary.uk/guidance-and-resources/artificial-intelligence-ai-judicial-guidance-october-2025/

Sources checked: 4 September 2026.

©INNOVATOR 07.09.2026.