Where AI eases the workload in a tax advisory firm

What applies to client data

Many of these tasks involve client data. They need a tool approved by the firm that meets the requirements of § 62a StBerG and the GDPR. For more, see AI in tax advisory firms: what is permitted with client data.

Until a tool has been approved for client data, start with tasks that have no client reference, or work with placeholders. As an easy way in, the Federal Chamber of Tax Advisers suggests clearly defined pilot cases, such as internal draft texts without client data.

Which tasks pay off first

Use these questions to sort your firm's tasks before you start.

• If a task runs the same way for many engagements, then it goes to the top. A good template can be reused there again and again.
• If the result can be checked at a glance, such as a reminder email, then the task is a good place to start.
• If the task needs client data and no tool has yet been approved for it, then put it on hold or solve it with placeholders.
• If a mistake would be costly, for example with deadlines or an objection, then use AI only for preparation or as a cross-check, and check thoroughly.
• If a task comes up rarely, then a template adds little. Stick with your current approach there.

The Federal Chamber of Tax Advisers suggests rating each idea by benefit and effort and setting success criteria in advance. Typical measures are time per case, error rate and client satisfaction, for example through faster responses. Before you start, record how long a case takes today. Otherwise there is nothing to compare later.

A simple list with five columns is enough for sorting: task, phase in the firm's year, how often it comes up, whether client data is needed and who checks the result. Whatever ends up at the top is your candidate to start with.

Year end

Many tasks pile up between December and February. Here AI mainly helps with letters and lists that look similar across many engagements.

• requests for missing receipts for all clients in one go
• checklists on payroll changes for each client
• answering team questions from the firm's documents

Worked example: requesting missing receipts. Before, staff write a separate email for each client and word it afresh every time. After, they note in keywords for each client which documents are missing and by when they are needed. The AI turns these notes into drafts in the firm's tone. Staff check each draft, add the personal details and send it.

A sample prompt for this:

“Write a friendly reminder to [SALUTATION]. The following documents are still missing for the income tax return: [DOCUMENTS]. Please send them by [DATE]. Tone: warm, brief. Mention only the documents listed and add nothing.”

Before sending, check three things: Are the name and salutation correct? Is the list complete? Does the deadline match your calendar?

Financial statements and tax returns

From March to July, a lot depends on documents being complete and the figures being consistent.

• checking missing documents against the request list
• flagging contradictions between client information and the draft
• summarising the financial statements clearly on one page

For the check, the firm's request list and the list of documents received go into an approved tool. The AI outputs what is missing and what has been supplied twice. Staff check this against the file and use the result for the follow-up request. The same applies to the summary of the financial statements: the AI writes one page in plain language, and the qualified tax adviser checks every figure against the statements before the page goes on the table at the client meeting.

Tax assessment notices and correspondence

Letters from the tax office arrive all year round. Here AI mainly helps with explaining and preparing.

• drafts that explain tax assessment notices to clients in plain language
• preparing draft objections with the disputed points and citations, with every citation checked
• listing the deadlines from each letter as a cross-check against the deadline register, not as a replacement

AI answers can cite sources or rulings that do not exist in that form or do not apply. The Federal Chamber of Tax Advisers therefore recommends looking up cited sections in the statute and at least checking calculations for plausibility. The greater the impact, the more thorough the check. This sequence works for a draft objection:

• have the AI list the disputed points and compare them with your own review
• open and read every cited source in the original
• recalculate amounts yourself
• check the deadline and the responsible authority separately in the deadline register
• have the qualified tax adviser approve the draft

You will find a sample prompt for explaining a tax assessment notice in plain language in AI in tax advisory firms: what is permitted with client data.

Planning and the annual client meeting

From October to December, annual client meetings and new legislation are on the agenda.

• summarising the client's year on one page for the annual meeting
• checking circulars against the legislative sources
• drafting minutes with tasks and responsibilities from your notes

Worked example: the minutes after the annual meeting. Before, the qualified tax adviser writes up the minutes in the evening from their notes. After, the keywords go into an approved tool. The AI structures them, and the adviser corrects and approves the result. A sample prompt for this:

“Organise the following notes from a client meeting into three sections: outcomes, tasks with responsible person and date, open questions. Add nothing that is not in the notes. Mark unclear points with [?].”

The [?] marker shows where the notes have gaps and directs the review there.

Next steps

• For each of the four phases, write down the tasks your team repeats most often.
• Sort the list using the decision guide above and choose two or three tasks.
• Record how long a case takes today and how often follow-up questions come in.
• Clarify which tool is approved for this, or start by solving the tasks with placeholders.
• Define the review steps for each task and who gives approval.
• Evaluate after a few weeks and decide which task to add next.

The AI workshop for firm teams takes half a day. In it, your team prepares emails, documents and letters with AI, practising with made-up or anonymised examples from everyday firm work. You leave with email templates, checklists, a draft set of rules for AI use and a certificate of attendance for each person. You will find all formats under AI for tax advisory firms.

Which task comes up most often in your firm?

Frequently asked questions

Sources

1. BStBK, FAQ on AI (January 2026) (in German)
2. § 62a StBerG, German Tax Advisory Act (in German)
3. GDPR, Article 28, EUR-Lex

Updated: 1 October 2026. This article is not legal advice.