AI for tax firms and auditors

Your clients.
Your knowledge.
Your agent.

AI for document review, assessment checks and firm knowledge – from an on-demand assistant to an agent with its own area of responsibility. Running in your own firm or in a German data center. Not a single piece of client data leaves the EU.

Three use cases that save time from day one

Document review & account coding

Receipts and invoices in – coding proposals, anomalies and ready-to-send client queries out.

Less rework, faster month-end close

Assessment review & objections

Tax assessment and return in – deviations, the objection deadline and a reasoned draft objection out.

No deviation overlooked, no deadline missed

Firm knowledge on demand

“How did we handle this for client X in 2023?” – an answer with its source from the file, correspondence and guidelines.

The firm's experience available to every member of staff

0

client data outside the EU – technically ruled out

2

operating models: in your own server room or hosted in Germany

8

weeks from first conversation to an evaluated pilot

Why sovereign – and why for a tax firm this is not an option but a professional duty

Professional secrecy under § 203 StGB (German Criminal Code) – client data must never end up with third parties who can access it. On-premise makes the question moot.

Client data is never used to train third-party models – technically ruled out, not just contractually promised.

GDPR, NIS2, CLOUD Act – processing exclusively within the EU; subcontractors subject to US access are contractually excluded.

GoBD compliance – every query and answer is logged and attributable to a person; we deliver the required process documentation with it.

The platform – two operating models, one software

On-premise · NN-Forge appliance

A pre-configured inference server in your own server room – no internet connection needed in operation.

For firms with their own IT · productive in 4–6 weeks

Hosted EU · data center in Germany

Operated by Neuronetz; connected to DATEV, your practice software or DMS with permission-aware access.

For a fast start and multiple locations · productive in 1–2 weeks

OpenAI-compatible interface · open-weight models only · switch from hosted to on-premise at any time, with no data migration.

From assistant to agent – automation in three stages

STAGE 1 · ASSISTANT

The human asks, the AI answers

Document review, assessment checks, firm knowledge – on demand, with sources.

STAGE 2 · AUTOMATION

The AI prepares the work, the human signs off

Inbox sorted, deadlines (objections, filings) monitored, client replies prepared – ready for sign-off in the morning.

STAGE 3 · AGENT

Its own role, its own mailbox

Runs deadline control and first drafts for a defined group of clients; the advisor reviews spot checks and exceptions.

Each stage builds on the previous one. You set the pace – trust is built on results, not promises.

A pilot in 8 weeks – no investment decision required

WEEK 1

Scoping

Half a day: use case, client group, 3–5 users.

WEEKS 2–3

Setup

Platform, connection to your document store, permissions, onboarding.

WEEKS 4–7

Operation

Daily use in real work, weekly check-ins.

WEEK 8

Evaluation

Time saved, quality, adoption – roll out or stop.

You provide: one contact person from the business side and the documents of one client group. We deliver: platform, setup, support and a final report. We discuss terms in the first conversation.

Let's talk about your firm.

Stephan Kasdorf · Managing Director · Neuronetz UG (haftungsbeschränkt)

question@neuronetz.ai · 089 93960222

Exposé as PDF