We build complete AI systems your team will actually use, with data flows
you can trace. EU infrastructure, a signed data processing agreement, and
support until the system lands in your office instead of sitting on a hard drive.
For regulated law firms, tax advisory practices and insurance brokers, data
protection is the technical starting point. The same build carries in any
business, whatever the industry. You get the running system with a data
processing agreement and logging, set up and signed off.
Other providers ship an interface. antomata ships the complete, usable
system: the screen your staff actually touch, the data processing
agreement that is signed, and the support through the first 60 to 90 days, until
your team uses it daily.
02/ Solutions
Our AI systems
Four building blocks, one system underneath: built once, configured for each
client. The blocks stack; every stage contains the previous one, and every
price includes setup and onboarding. What stays the same: EU infrastructure,
a data processing agreement, and human approval before any outgoing document.
Block 01 · Reports
Your numbers live in scans, attachments and forms.
Scanned receipts, email attachments, forms: someone gathers them, types them
over and builds the same reports from them every month. The Reports block
reads that input, structures it to a fixed schema and generates the recurring
reports automatically. You read and decide; a review screen shows every
result before it counts.
Drafts, replies, summaries: your team types them by hand.
Client letters, standard replies and summaries are written from scratch every
day, in Outlook, DATEV or WhatsApp. The AI Tools take over that routine work
right inside the programs you already use: the system lays out the finished
draft, and a person approves before anything leaves the office.
Your company's knowledge sits in folders and in heads.
Whoever needs an answer digs through filing systems and old emails, or asks
the one colleague who knows. The Knowledge Chatbot answers in plain language
from your own documents, with the source attached. The knowledge stays in
the firm, even when someone leaves. It runs on EU servers with a signed data
processing agreement.
Incoming emails, documents and deadline triggers all need reading, sorting
and answering before the real work can start. The Agent System runs those
multi-step processes on its own, with an approval queue and a full audit
trail. Your team reviews instead of typing.
We do not deliver a tool or a report. We deliver a running system: built once,
configured for your office, handed over with every contract a regulated firm
needs.
The shared core
Built once. Tested. Reused.
Beneath each of the four building blocks runs the same core: EU infrastructure, an
approval queue with human review before any outgoing document, full logging,
a data protection architecture. We do not rebuild that base every time. That is
what makes the build fast and the result robust.
· Human approval before anything is sent
· Complete logging of the workflows
· EU region, client data kept separate
The adapter per office
Shaped to your workflows.
What changes from client to client is the adapter: a prompt, a schema, a rule,
a connection to Outlook, DATEV or your ERP. That is how the system adapts to
a tax practice, a law firm, a brokerage or a mid-market business without
losing the tested base.
· Your language, your templates, your rules
· Connected to the tools you already use
· Extensible with further building blocks on the same core
What gets handed over
DPA under GDPR Art. 28
The signed data processing agreement that governs how your data is handled.
A SaaS sign-up does not replace it.
Work contract
A fixed deliverable at a fixed price. No hourly overrun, no rolling license
you can never escape.
Acceptance protocol
At handover we test and record that the system does what was agreed. Only
then is the stage complete.
Infrastructure
The default is an EU region with a signed data processing agreement. For
especially sensitive mandates, such as professional-secrecy holders, we work
with cautious infrastructure (self-hosted at Hetzner in Germany, open-weight
models as the careful default) and settle the exact setup in the discovery
call.
Pricing
From the AI Audit to ongoing operation.
The entry point is an AI Audit for €500, credited in full toward the
build. From there the system grows in fixed stages: Reports €2,000, with
AI Tools €4,000, with the Knowledge Chatbot €6,500, all four blocks on
request. Every stage contains the previous one; every price includes setup
and onboarding. For ongoing operation there is a monthly model; the pricing
page has the full picture.
Four segments. Based in Leipzig. Clients nationwide and abroad.
Segment 01
Tax advisory firms
5–15 staff
The practice lives in DATEV. Where this usually starts: ChatGPT-Plus running
as shadow IT, client documents uploaded to Notion AI, no clarity on what EU
AI Act Art. 4 asks of them.
Focus areas
DATEV DPA · Client data agreement · Shadow-IT inventory
Segment 02
Law firms
5–20 staff
Professional secrecy under BRAO § 43e plus the GDPR double layer. Where this
usually starts: client communication through ChatGPT, research via Claude.ai,
no signed DPA with US providers.
Focus areas
§ 43e inventory · Sub-processor list · DPA gaps
Segment 03
Insurance brokers
3–10 staff
VVG and IDD obligations. Where this usually starts: advisory documentation
running through ChatGPT-Plus, claims handled in Notion AI, application data
ending up in US cloud by accident.
Focus areas
VVG advice documentation · BaFin evidence · DPA
Segment 04
Mid-market businesses
10–50 staff
Whatever the industry. Where this usually starts: quotes are written by hand,
supplier emails pile up, order confirmations get retyped, and instead of
DATEV the central system here is an ERP.
Running solo or with a small team? The entry point is the same: the €500 AI
Audit, credited in full toward the build.
05/ Stack
Tool recommendations with a clear line: what we use and what we don't.
Recommended
· Self-hosted n8n at Hetzner in Germany
· Claude via AWS Bedrock eu-central-1
· Make.com on an EU instance
· Notion with AI features disabled
· Plausible (EU) for web analytics
Not in client systems
· Notion AI features (US, no DPA for AI)
· Firebase (Google US, Schrems-II risk)
· US-hosted LLM gateways
· Direct Anthropic API without a signed DPA
· Google Analytics 4 (cookie banner required)
As of 2026-05. This list is extended with every audit.
06/ Limits
What we do not deliver.
Legal Services Act · § 5 RDG
We provide no legal subsumption and no legal advice within the meaning of
§ 5 RDG. What we deliver is a technical assessment, not legal advice.
The legal assessment is made by your own legal department or the firm you
instruct.
EU AI Act · Annex III
We decline mandates whose use case falls under EU AI Act Annex III (high-risk),
for example CV-screening automation, employee monitoring, behavioural scoring or
credit scoring, as long as no conformity assessment procedure is in place. That
is the line between deployer and provider under Art. 25
EU AI Act. We stay deployer.
Guarantees
We give no compliance guarantee and no regulatory success guarantee. What we
deliver is a professionally grounded technical assessment with references from
EU AI Act, GDPR, RDG, BRAO and SGB. The assessment in a dispute is made by a
court or a supervisory authority.
07/ Discovery call
Discovery calls open now
We take on a limited number of new projects per quarter. The entry point is the AI
Audit (€500), credited in full toward the build. You will find every
price on the pricing page.
We reply within 24 hours with two suggested slots for a 30-minute discovery call.
If your use case falls under Annex III, we say so in the first sentence, with the
note that an external legal assessment is needed for it.