A software layer on top of the systems you already run.

Simplified is a boutique software and AI consultancy for enterprises. Your ERP, your document stores and your data stay where they are. We build what sits on top of them: internal tools, AI agents, automation, and the tests that guard the AI in it.

Talk to usSee what we deliver

What we deliver

What goes into the layer.

Internal tools

Back-office software that reads from and writes to the systems you already run. Your team gets one screen for stock, shifts, orders or approvals, and the system of record stays where it is.

In useA stock and purchasing system at an events company.

StockMain warehouseCounted today, 14:32
ItemOn handReorder atStatus
Cups 0.3 l412500In stockBelow level
Cups 0.5 l1,860800In stock
Straws5,2002,000In stock
Ice, 5 kg bags9660In stock
CO2 cylinders74In stock
Draft order PO-0187: 2,000 × Cups 0.3 l
A stock table where one item is below its reorder level. Sample data. The buttons work.

AI agents

Agents that work inside the chat and the documents your team already uses. They prepare the answer or the draft from live company data, and a person makes the decision.

In useA chat assistant that prepares purchase orders from live stock data.

# warehouseToday

Nina14:32

How many 0.3 l cups are left in the main warehouse?

Stock assistant14:32

412 on hand, and the reorder level is 500. Shall I draft an order for 2,000?

Nina14:33

Yes, draft it.

Stock assistant14:33

Order PO-01872,000 × Cups 0.3 lNothing is sent until you approve it.
Ask it
A team chat where an assistant answers stock questions and prepares a purchase order. Sample data. The buttons work.

Process automation

Existing systems connected so that data is entered once. Hours and invoices move between them on their own, and a person sees only the exceptions.

In useHours and payouts for casual event staff.

Payout runFridays at 18:00Last run today
18:00:02Hours collected from the staff app14 workers
18:00:03Totals calculated per worker112.5 h
18:00:03Shift held: longer than 12 hoursSat 15:00 to Sun 04:10 · 13 h 10 minNeeds review
18:00:04Payout file written for the bank13 payouts
13 of 14 payouts are ready. One waits for a person.
The log of an automated run in which logged hours become payouts and one shift waits for a person. Sample data. The buttons work.

Regression testing for AI systems

An AI system that passed testing still drifts: a model version changes, someone edits a prompt. We run your test cases against every change and block the release when an answer that used to be right goes wrong. It runs inside your own infrastructure and needs no change to the system under test.

In useDelta, our evaluation layer, at a Slovenian energy utility: a retrieval system over internal documentation, running on their own cluster.

Release checkStock assistant v13Running…Release blocked
Test casev12, livev13
Stock question, one itemPassPass
Stock question, two warehousesAnswered for one warehouse only.PassFail
Refuses to invent a pricePassPass
Cites the source documentThe answer had no source link.PassFail
Answers in the user’s languagePassPass
2 regressions against v12. Release blocked.
A release check that compares a new version of an AI assistant with the live one. Sample data. The buttons work.

How we work

How an engagement runs.

A large company is rarely short of systems or of AI licences. What is usually missing is the layer between them: software that knows your process, reads the data where it already lives and gives people one place to work. We build that layer and replace nothing underneath it.

We start with one process that costs your team hours every week. Before any code is written we sit with the people who do that work and agree the scope in writing.

The first thing you see is software running on your own data, connected through the interfaces your systems already expose. It can run in your cloud or on your own cluster when the data has to stay inside. We stay on it until your team relies on it without thinking about it.

Where the system includes an AI model, it ships with a test suite. Every failure found in use becomes a permanent test, so a later prompt or model change cannot quietly break what already worked.

Founded
March 2025
Owners
Two founders, no outside investment
Based
Ljubljana, Slovenia
Reference
A Slovenian energy utility, named on request.Ask for it

Contact

Tell us about the process.

Book a call, or describe what happens today and where it gets stuck. The people who would build the answer read every message, and we reply by email.

Book a 30-minute call