Find project knowledge
"What did we promise on maintenance back in 2019?" The agent searches the released project files and answers with source and date.
AI agents & software · Vienna
Bespoke AI agents.
Searching documents, preparing reports, moving data between systems and more. We build agents for work like this, integrate everything into your IT and support it in operation.
A good year in customer operation
Satoru runs across several of our projects. At Sentinel the customer works with it directly: checking systems, fixing faults, asking back only at the agreed limits.
See Sentinel →Starting point
The knowledge is there. Nobody finds it without digging.
Information sits in documents, email, SharePoint, network drives, line-of-business systems and in people's heads.
-
Searching the same thing twice
The same information gets looked up again and again, in several versions, with no clear validity.
-
Copying by hand
Data moves between systems manually. Every break costs time and introduces errors.
-
Knowledge tied to people
Project experience sits with individuals. A move or a retirement takes it along.
You already own the knowledge. You just pay again every time to dig it back out.
What we build
An agent hands over a result, not just an answer.
It works with the documents and systems you release to it, runs through several steps and hands over something you can carry on with.
Compare documents
Two contract versions, changed clauses, missing details. The agent assembles the differences instead of someone reading page by page.
Prepare reports
Notes, a voice message and photos from the field become a draft report. Back at the desk you complete it and sign it off.
Fix faults
The agent runs recurring checks, narrows down the cause when something breaks and fixes it. It only asks back where you require sign-off.
Possible applications, from routine work through to tasks it carries out, such as sending email or issuing invoices. Scope, data access and limits are settled case by case.
You draw the limits: which documents the agent uses, what it handles itself and where it asks back. Inside those limits it works straight through.
We connect what is already in place: Microsoft 365, SharePoint, email, network drives, databases or your own application.
Evidence
What is already running in production.
Sentinel web portal
Video data from capture through processing to the end customer's web portal, end to end from one team. Three years in operation and still growing. Our agent Satoru checks the running systems and reports relevant faults.
Python
Flask
FFmpeg
View details →
Zedl
Photograph a receipt, get structured accounting data back, export to DATEV, BMD, CSV or Excel.
View details → Website ↗Nodl
Spoken notes become structured documents, not just transcripts.
View details → Website ↗ÖBB PoC
Synthetic training data for detecting missing, broken and misaligned rail fastening clips.
View details →Voice-paced teleprompter
A web teleprompter with live speech recognition that follows the speaker's pace.
View details → Website ↗Satoru works for us too: development, testing and recurring office work such as incoming invoices.
Working together
How we start.
We begin with one clearly bounded area. You see early whether the solution helps in daily work.
We look at your existing documents together and pick the ones that suit the first step. A tidy data landscape is not a precondition.
Phase 1 · PoC
Phase 2 · Further development
01
First conversation
Thirty minutes. Your situation, and whether we are the right people for it.
02
Workshop
Processes, data, access and goals. After that, scope and fixed price are set.
03
PoC
A small, usable part with real cases. Target: first use after about a month.
04
Further development
What proves itself, we extend. We support operation to the scope we agree.
Whether an agent helps is measured on three things: handling time including rework, correctness of results, effort in daily use.
Pricing
Phase 1 · PoC
EUR 2,500–15,000
Fixed price after the workshop. Scope and schedule are agreed before we start.
Phase 2 · Further development
As needed
A block of hours on call, or fixed hours per month. Separate from the PoC.
Testimonials
What our customers say
“Thank goodness there are clever people like you who work out the perfect solution for problems like this!”
Lawrence
Customer, Zedl
“You are the best!”
Customer, Sentinel, to our agent Satoru
“Brilliant! Lads, this is brilliant! BRILLIANT!”
Customer, Sentinel
“Are those your images now, or real ones? I cannot tell the difference any more.”
ÖBB
“When I scanned the first receipts I was simply amazed at how the Zedl app makes tedious bookkeeping and time sinks disappear.”
Magdalena B.
Customer, Zedl
Team
Straight to the people who build your solution.
Two developers in Vienna, jointly responsible for delivery.
Christian Eichberger
Processes and architecture
Christian works with your people to find where the effort actually sits, then turns that into a solution that can be built. Focus: AI applications, data platforms and technical architecture.
Education and personal work experience
- Degrees
- Ing., MSc., BSc.
- Education
- FH Technikum Wien, FH Sankt Pölten
- Personal work experience
- Adidas, XXXLutz, ÖBB, Paiqo
Sebastian Beyer
Development and operations
Sebastian builds the applications and data flows that have to keep running afterwards. Focus: robust software, maintainable systems, dependable operation.
Education and personal work experience
- Degrees
- Ing., MSc., BSc.
- Education
- FH Technikum Wien, HTL Hollabrunn
- Personal work experience
- Profactor, Genera, Emotion3D, BD
Contact
Where does it get stuck for you today?
Describe one routine that eats too much time: searching, copying, assembling. We will tell you straight whether an agent is worth it.
A free, no-obligation first conversation of 30 minutes.
Response time: under 24 hours
Blog
Notes and thinking
You Don't Need to Read Every Line of AI-Generated Code
Sebastian Beyer of ex-nihilo argues that reading every line of AI-generated code is not the only quality bar, and often …
Read all articles →Why AI Coding Fails at Taste
Sebastian Beyer of ex-nihilo explains why AI coding produces slop in high-taste codebases. When the real test of quality…
Read all articles →AI Can Code When Verification Is Cheap
Sebastian Beyer of ex-nihilo argues that the debate over whether AI can write good code misses the point. AI produces sl…
Read all articles →