AI knowledge base for tax advisory firms: the practical guide

An AI knowledge base is most useful to a tax advisory firm when it starts with the firm’s internal knowledge: work guidelines, checklists, templates and documented special cases become findable by asking a question in plain language – with a source pointing to the original document. Client data and accounting stay in the practice management software; the knowledge base answers the question of how the firm works. For professionals bound by secrecy, the architecture is what counts: data held in the EU, no use of the documents for third-party AI training, clear access rights.

In brief

• The best way into AI for tax advisers is the firm’s internal knowledge – standards, checklists, instructions – and not the client file.
• Firms carry enormous process knowledge in a few heads; when an experienced employee resigns or retires, it is suddenly missing.
• An AI knowledge base answers questions from your own documents and always names the source document, so that every answer can be verified.
• Professional secrecy and data protection require data held in the EU, no third-party training and role rights – that is the minimum bar for any AI in a firm.
• The practice management software remains the leading system for accounting, deadlines and client files; the knowledge base adds the how.

Why is internal firm knowledge the best starting point for AI?

Because that is where the bottleneck is. Professional research is well catered for in tax advisory firms – specialist databases, commentaries, seminars. What is hardly documented, by contrast, is how your own firm works. According to which checklist is an annual financial statement completed? How is a new payroll client set up? Which standard text applies to the request for a deadline extension? These answers are in the heads of a few experienced employees and in scattered Word files.

This becomes expensive in two places. First, during onboarding: new trainee tax advisers and clerks tie up the experienced colleagues’ time for months, because every process question is answered verbally. Second, when knowledge holders leave: when the long-standing payroll accountant leaves the firm, the countless special cases she has solved over the years go with her – from partial retirement to cross-border employment.

A third point speaks in favour of this way in: internal firm knowledge contains no client data. The firm can gain experience with AI search without sensitive data being affected, and decide the trickier questions closer to clients later on a secure basis.

What knowledge does a tax advisory firm have?

Knowledge area	Typical content	Everyday benefit
Firm standards and work guidelines	Procedures for financial statements, payroll, handling of receipts	Consistent quality across all employees
Checklists	Annual financial statements, tax returns, client onboarding	Nothing is forgotten, even under deadline pressure
Templates and standard text blocks	Letters to authorities, client letters, fee agreements	Faster and more consistent correspondence
Software and portal instructions	Practice management software, FinanzOnline (the Austrian tax authority’s online portal), public authority portals	Fewer questions about routine tasks
Documented special cases	Solved special topics from payroll and financial statements	Special cases only have to be worked through once
Onboarding documents	Induction plans, internal responsibilities, firm A–Z	New employees get to work faster

The common thread: it is about instructions, standards and documented experience – not about receipts. A receipt belongs in the client’s accounts; the work instruction on how the firm handles this type of receipt belongs in the knowledge base.

What questions does the firm’s team ask in everyday work?

The benefit becomes visible in the questions that today are called across the office or emailed to the same colleagues again and again:

1. ‘Which checklist applies this year to the annual financial statements of a GmbH (limited company), and what has changed compared with last year?’
2. ‘How do we set up a new payroll client – from the power of attorney to the registration?’
3. ‘Is there an internal work instruction on handling company cars in payroll?’
4. ‘Where is our standard text for a request for a deadline extension, and who approves it?’

The AI answers such questions from the firm’s documents and refers to the source; one click opens the original. This is crucial in a tax context in particular: an AI can be wrong, and an answer is only reliable once the underlying guideline can be checked. The knowledge base therefore replaces neither the specialist database nor professional judgement – it makes the firm’s own know-how findable. How this works technically and which role and rights concepts are possible is shown in the junova product overview.

What do professional secrecy and data protection mean for the choice?

As professionals bound by secrecy, tax advisers are subject to a strict duty of confidentiality – and this sets the framework for every AI tool in the firm. From a firm’s point of view, three criteria are the minimum bar:

First, data storage: the documents should be processed and stored on servers in the EU, with a data processing agreement (DPA) as the basis. Second, the training exclusion: the firm’s documents must not be used to train third-party AI models – this should be guaranteed by contract. Third, access control: role rights must reflect that not every employee may see every piece of content, from fee calculations to management matters.

The contrast with a free chatbot is clear: when employees type questions including context into public AI tools for lack of an internal solution, information flows outwards uncontrolled. A firm’s own knowledge base with clear data storage is also an answer to this shadow IT problem: it gives the team the convenience of AI search within a framework the firm controls.

How does the practice management software remain the leading system?

The division of labour is the same as between ERP and knowledge base in a commercial business: the practice management software knows what has been done – the knowledge base knows how it is done. Accounting, deadline management, document filing per client, time recording and fee notes stay in the practice management software; it is built for this and remains so.

The knowledge base comes in alongside it: with the processes, standards and experience that are not in any client file. It does not answer ‘What was booked for client Meier?’, but ‘How do we handle this situation as standard here, and where is that written down?’

How does a firm realistically introduce the knowledge base?

A start in four steps has proven effective: first, the firm collects the team’s actual everyday questions for two weeks. Then structure and access rights are defined – separated by specialist area, management level and administration. Next, the starting material is loaded: checklists, guidelines, templates, software instructions, all without any connection to clients. Finally, two or three employees test the answers against the source documents before the whole firm gets access.

The cross-industry basis for this sequence of steps is described in the guide AI knowledge base in your company. One special feature applies to firms here: the clients’ works council is not an issue, but the firm’s own governance is – who approves content, who keeps it up to date, who decides on later stages of expansion.

The concrete next step: take a look at how the introduction works step by step – from the initial meeting through the rights structure to the test with your own firm standards. A firm that has its checklists and guidelines together usually sees within a few weeks whether AI search really eases the workload in everyday work.