.webp)
When client management is tightly scheduled, tax law changes are coming thick and fast, and the flood of documents never seems to end, it’s natural to wonder: couldn't AI take a lot of this off your plate? The answer is a clear yes—but only where it is professionally compliant, runs in accordance with data protection regulations, and actually fits into your firm's existing workflows.
This article shows which AI tools really work in practice, where the Federal Chamber of Tax Advisors (BStBK) draws clear lines, and how to reconcile their use with confidentiality obligations, the EU AI Act, and DATEV environments.
Professional law and AI: What the BStBK allows and where the line is drawn
In February 2026, the Federal Chamber of Tax Advisors officially took a stance for the first time: they published a comprehensive 23-page FAQ catalog on the use of AI in tax firms, covering strategy, tool selection, governance, and law. The core message can be summarized in one sentence: AI is an amplifier, not a replacement. It is intended to give tax advisors more time for actual consulting, not to take over the consulting itself.
That’s more than just a platitude. Professional responsibility remains entirely with the tax advisor—every professionally relevant AI response must be verified before it reaches a client. This "human-in-the-loop" requirement is non-negotiable.
The most important professional legal framework
Tax consulting is a reserved task—AI can assist, but it cannot provide independent advice. In practice, this means:
- Confidentiality remains mandatory: Client data may only be entered into systems where data processing agreements and third-country transfers are properly regulated.
- Competence requirement under the EU AI Act: Employees must be sufficiently trained to professionally evaluate AI systems and identify risks. The more extensive obligations for so-called high-risk systems only take effect as of August 2026—this currently only affects most firms on the periphery, but it should be kept on the radar.
- Documentation requirement: Which tools are approved for which use cases should be recorded internally.
The German Association of Tax Advisors provides concrete guidance with its model AI usage policy published in April 2026. It is intended as a non-binding template that every firm must adapt to its own size, IT landscape, and risk profile—but it follows a clear logic: AI may only be used if the tool is approved, the use case is defined, and the requirements of the respective risk level are met. Freely accessible chatbots like the free version of ChatGPT remain particularly tricky: without additional contractual and data protection agreements, they are simply too risky for client data.
AI tools that actually make a difference in tax firms in 2026
Four categories have proven their worth in firms so far. What they all have in common is that they integrate into existing workflows, provide clear contractual foundations for data processing, and deliver results that can be professionally verified.
DATEV-integrated AI features
For firms already working with DATEV , these are the most logical starting point—they are already embedded in familiar processes and benefit from the cooperative's data protection and data processing agreement infrastructure.
Typical areas of application:
- Document classification and automated pre-accounting: One of the biggest productivity boosters. AI recognizes document types, suggests account assignments, and significantly reduces manual review steps.
- Client communication: Pre-drafted responses to standard questions that are still reviewed internally.
- Year-end preparation: AI suggests plausible entries and identifies anomalies in the bookkeeping.
The advantage: Data protection and professional regulations are already largely covered by the DATEV infrastructure. The obligation for professional verification, of course, remains.
Specialist publisher AI for tax law research
Systems from Haufe, NWB, Deubner, juris or Stollfuß excel at tax law research. They use professionally curated databases, link directly to sources, and provide much more precise answers than general chatbots.
Especially valuable:
- Source-based research: Instead of vague summaries, the AI provides references to current court rulings, Ministry of Finance circulars, and expert commentary.
- Expert-curated data: The training data comes from verified professional publications, not the open internet.
- Client-independent use: Research AI typically works with general queries, so sensitive client data is not included.
For complex edge cases and recent legislative changes, these systems are significantly more reliable than generic LLMs.
Microsoft Copilot for internal law firm processes
Microsoft Copilot is primarily suited for internal workflows: drafting emails, meeting minutes, document templates, and knowledge management. A properly established corporate and data protection framework is a prerequisite.
Useful use cases:
- Email templates and client correspondence: Copilot suggests phrasing that is then reviewed internally.
- Meeting summaries: Minutes are automatically created and structured.
- Internal knowledge base: Quick access to past cases, similar mandates, and best practices.
The same rule applies here: whatever Copilot suggests for client correspondence should never be sent out without review.
Caution with public chatbots
Public chatbots like ChatGPT Free or Gemini should be used with caution. Without an appropriate contract and data protection framework, client data simply doesn't belong there - if you use them anyway, stick to general questions and make sure to consistently anonymize sensitive information. They can still be useful for internal brainstorming or drafting assistance, as long as it's clear that an expert needs to review the final output.
“The key is not to ban AI in the firm, but to set clear guardrails. Confidential client data belongs in closed firm systems like DATEV or Azure—never in open web prompts.”
— Jens Bohse, Co-Founder of bakedwith
Proven use cases in tax consulting
The biggest impact is found where there is a lot of routine work and where research usually eats up a lot of time. This is where using AI really pays off:
Document processing and classification: AI recognizes document types, suggests account assignments, and significantly reduces manual review time per document. This pays off especially with high document volumes.
Tax research: Quick access to the latest rulings, BMF letters, and commentary literature – without spending hours scouring databases.
Financial statement preparation: AI suggests plausible entries, identifies anomalies, and prepares annual financial statements. The professional review remains with the tax advisor.
Client communication: Pre-drafted responses to standard questions take the pressure off the team and speed up processing. Important: Always double-check before sending.
What tax advisors should keep in mind when introducing AI
If you want to use AI productively in your firm, you should take a structured approach. The most important steps:
Define clear use cases
Not every process benefits from AI to the same extent. It makes sense to start with clearly defined use cases: document processing, research, or client communication. Pilot projects help you gain initial experience and realistically assess the potential.
Set up data protection and data processing agreements properly
A standard data processing agreement is often not enough if professional confidentiality and potential third-country transfers aren't clearly addressed. Especially with cloud-based tools, you should verify where data is processed and which subcontractors are involved.
Train your team
The EU AI Act requires a duty of competence: Employees must be sufficiently trained to professionally evaluate AI systems and identify risks. Training sessions should be documented and refreshed regularly.
Consistently implement a human-in-the-loop approach
Every professionally relevant AI response must be reviewed before it is sent to a client. This rule is non-negotiable—both from a professional and a technical standpoint.
Bottom line: AI is a productivity tool, not a replacement
AI has arrived in tax consulting in 2026—it's professionally permitted, practically useful, and has a noticeable impact on daily firm life. This assumes that data protection, confidentiality obligations, and final professional responsibility remain exactly where they belong. The tools that work best are those closely integrated with firm software or verified professional databases and have a clearly defined use case. The most pragmatic way to start: pick one use case, learn from it, and go from there.
FAQ: Frequently asked questions about AI for tax advisors
Which AI is best for tax advisors?
That depends on the use case. For document processing and firm workflows, DATEV-integrated systems are best; for tax research, AI from professional publishers like Haufe, NWB, or Stollfuß scores points; and for internal processes like email and document work, Microsoft Copilot is a great choice.
Will tax advisors be replaced by AI?
No. In its FAQ catalog, the BStBK explicitly classifies AI as a tool and a sparring partner, not a replacement. Professional responsibility, confidentiality obligations, and human oversight remain strictly with the tax advisor. AI handles routine tasks and speeds up research—but the actual consulting remains a human endeavor.
Which AI should I use for tax returns?
Ideally, AI functions integrated into DATEV that handle document classification, pre-accounting, and plausibility checks. Public chatbots are risky for client data due to professional regulations and should only be used for general questions at most.
What software do tax advisors use?
Most German firms work with DATEV. This is often supplemented by industry-specific tools for research (Haufe, NWB, juris) and office organization (Microsoft 365, Copilot)—the right combination depends on the size of the firm, the client base, and internal processes.
How do I integrate AI into existing firm workflows?
The best approach is to start with a clearly defined pilot project—such as document processing or research—train your staff, and document the results. Be sure to check data protection, data processing agreements, and professional regulations beforehand. Once the pilot phase is successful, you can automate further processes step by step.
Posted by:
Jens is the co-founder of bakedwith, a boutique consultancy firm specialising in smart automation and AI. He helps medium-sized companies and corporations to optimise processes, reduce manual work and achieve growth through efficient workflows. Previously, he was Growth Lead at OMR and worked as a freelancer, helping numerous companies optimise their CRM and automation systems. Jens is passionate about combining growth and efficiency to enable teams to focus on what matters most.
