Skip to content
Bernhard GötzendorferAI automation that runs on its own · Vienna

Everything you end up with

I automate workflows with AI: as autonomous as possible, checking with you as often as needed.

AI automation

A workflow that runs on its own

by handruns aloneasks you

An agent takes over the work someone in your business does by hand every week.

You end up with a workflow that only asks you when it matters.

How it works, and an example

It only has to ask about money, anything that goes out, and anything that cannot be undone. You get the source code and the tests.

The language model
reads what comes in, decides the next step and carries it out.
The code
does the maths, checks, and only stops where you need to decide.

Example: StatAgent works through 24 months on its own and holds two of them back.

If the workflow needs its own interface, I build the web app for it. I hand it over to your IT, with a guide. Where that isn't possible, it runs on the infrastructure you already have, and we agree together who looks after it.

AI phone assistant and voice agent

A voice agent that listens and gets it done

CallBooked„word for word

An agent handles your calls by phone or in the browser and hands over to you.

You end up with a voice agent that picks up within the agreed scope when you can't.

How it works, and an example

It says it is an AI, works out what the caller needs, books appointments and callbacks, and sends you a record with word-for-word quotes after every conversation.

The language model
holds the conversation, asks follow-up questions and sums up.
The code
books the appointments, checks every quote against the transcript and hands over to a person when one is needed.

Example: WalkAITalkie turns speech into finished text, on your Mac.

Document processing with AI

Documents and requests, read and sorted with a paper trail

RequestsReceiptsForms

Emails, requests, receipts, forms and contracts are read, sorted and put into the right system.

You end up with a workflow from the inbox to the right system.

How it works, and an example

Fields are extracted, numbers double-checked, every document filed under the right case. What the checks cannot settle lands with you.

The language model
reads, sorts, and backs every decision with the passage it came from.
The code
double-checks every number and every quote and writes the result into the right system.

Example: BuchhaltGenie reads receipts. The Pflegegeld-Prüfer maps every sentence to a rule. BewerbungsGenie checks every sentence against the profile.

A technical assessment of where AI fits your work

AICodeCodeAICode

You'll know what a language model takes on and where code has to do the maths.

You end up with one page with a yes or no per task, tested on your own cases.

How it works, and an example

Plus a comparison of AI models on your own cases. You also see where plain code is the better choice.

The language model
gets the tasks where a suggestion is enough.
The code
gets everything that has to be right.

Example: I compared 7 image-reading AI models on real receipts before BuchhaltGenie got one.

Running AI agents safely

An environment where your agents work within fixed limits

freeyour OK

Your agents work on their own, overnight too, because the limits are in the code.

You end up with a project that sets out what agents may do and what needs your go-ahead.

How it works, and an example

With automated checks, a fixed workflow per session and a rule file. The few open decisions wait in one place. What proves itself in one session becomes a rule for the next.

The language model
plans, builds and checks in rounds, with no one watching.
The code
sets the limits and blocks what is not allowed.

Example: my own agent environment with session-orchestrator and lead-session-orchestrator.

A team that builds with agents on its own

built themselves

Your people leave with something they built themselves, not with slides.

You end up with one result per person from their own daily work, and a cheat sheet.

How it works, and an example

Built with an agent, plus a working environment that is already set up. The cheat sheet shows the workflow.

The language model
works alongside on the day, on your people's real tasks.
The code
stays with you: the workflow, the rules and everything built that day.

Example: the courses by Agentic Builders.

Workshops for your team

On site, on your own laptop, with your own use case.

  • Your first workflow that an agent handles for you

    Who it's for: teams without their own developers, such as office, sales and admin.

    Afterwards you can hand a recurring task from your daily work to an agent and decide where it acts alone and where it asks you.

  • From one session to checked code

    Who it's for: development teams already working with a coding agent.

    Afterwards you can start a session with a plan, split the work across parallel agents and have every round checked.

  • Agents on a leash: rules, checks and approvals for your project

    Who it's for: heads of IT, tech leads and anyone who wants to allow agents without giving up control.

    Afterwards you can write down what an agent may do freely, what needs your go-ahead and what must never happen, and enforce it in the project.

What do you want to end up with?