Automation without the black box
Processes run themselves. You see every step.
Cordix Intelligence is a software platform that connects your systems and takes over the repetitive work between them. AI prepares the data; a rule you approved makes the decision — and for every case you can say what happened and why.
Your systems stay as they areYour rules decideEvery step can be verified
For companies whose data is spread across several systems.
The sequence of steps over time
- Inputsdocuments and emails
- DataAI reads and extracts
- Decisionthe rules you approved
- Outputstask for a person
The decision with its reason stated
Task for a person
ReasonA mismatch means a person verifies it.
The extracted values, each linked to its source
- AI reads and extractsPDF
- Direct connectionERP
- Input from a personrequests, extra details
Every value has a source and a confidence score; the uncertain goes to a person.
Manual work is what holds your systems together today. We’ll take it over.
ERP, inventory, databases, CRM, leases, contracts, email, spreadsheets. In most companies these systems don’t talk to each other — and so people sit between them, carrying data from one to the other, comparing it and verifying it. Day after day, the same work — nothing in it to work out, only something to get wrong.
Cordix Intelligence connects to the systems you already have and takes that work over. It moves data between them and verifies it, decides according to the rules you approved, and writes the result where it belongs. Whatever does not fit the rules goes to your people — together with the reason why.
And because a full record is kept of every case, the answer to why something happened is not hunted down across emails. It is opened.
Nothing is migrated and nothing is replaced along the way. The platform builds on top of what already works in your company.
Data, decision, execution. And a record beneath it all.
Inputs
documents and emails
PDFs, attachments, scans
systems and databases
ERP · inventory · CRM · web
people
requests, extra details
Data
AI reads and extracts
direct connection
input from a person
every value has a source and a confidence score; the uncertain goes to a person
Decision
the rules you approved
readable as a table · same input, same result
every decision with its reason stated
Outputs
written to a system
task for a person
report · alert
Record
a full trail of every case: inputs · rule · result · approval
Data. Straight from your systems; from documents, by reading them; and from people, who enter it by hand. Every value carries a link to where it was found — and when the platform is not certain, it says so and hands the case to a person. Whatever can be recalculated is recalculated. A mismatch means a person verifies it.
Decision. In most processes the outcome is decided solely by a rule you approved — written down as a table the process owner can read and change without a programmer. The same input always leads to the same result, and every decision comes with its reason stated. Where it makes sense, and where you have agreed to it in advance, an AI estimate can enter the decision too — and then it has exactly as much authority as you give it, while an irreversible step still waits for your approval.
Execution. The result is written into your system so that nothing is done twice. Steps that cannot be undone go ahead only after an authorised person approves them. And when the approval deadline passes, the case is rejected — not quietly approved.
Record. A full trail remains of every case: which values came in and from where, which rule decided and why, what was carried out and who approved it. Every decision can be traced back and explained. And the rules can be changed with confidence — you replay a new version against past cases and see, before it goes live, what would have been decided differently.
Everything in one place. We do not want decision-making left scattered across scripts and individual scenarios — nor do we want AI’s judgement taking it over unthinkingly. Here, decision-making always sits in one place: in rules that can be read, changed and traced with their change history, instead of being spread across dozens of scenarios where it gradually becomes unclear what decided what. And the record does not appear afterwards: it is part of every run, not something written up at the end. That is why this kind of automation can be trusted with a process where you need to know not only that it ran, but why it ran the way it did. And every process can run with AI switched off entirely — a person supplies the data instead, and the rest works exactly the same.
Any process with rules. Across the whole company.
The precise scope depends on how you operate. These are the areas where the platform applies most often:
Document processing.
Orders, delivery notes, leases, contracts, forms. From a PDF or an inbox into the right system — with the values extracted and verified.
Matching and checks.
Order against delivery, stock levels against the inventory records, the same data in two systems. The comparison runs itself; the differences go to people.
Approvals.
Purchase requisitions, orders, pricing exceptions. Checks, limits, approval, entry.
Recurring reports.
Monthly and quarterly, from several sources, including consolidation across a group of companies.
Deadlines and limits.
Contractual dates, stock levels, agreed terms. A warning and an escalation before anything is missed.
Answers from your own data.
A question in plain language, an answer with figures straight from the database — including where they came from.
We take on the processes where the benefit to your company is greatest. Where it would be small, we will not recommend it.
Four steps. The first proof comes from your own data.
Step 1
Assessment.
Thirty minutes on your systems and processes: where the volume is and what an error costs. By the end you know whether this makes sense for you — including a price range.
Step 2
Dry run.
We put the chosen process on your own data as a dry run: it evaluates everything and executes nothing. That is where we test how it behaves and work through the errors and the exceptions, before anything is written into your systems.
Step 3
Full deployment.
We connect all the systems and processes and set the final rules, limits and approvers.
Step 4
Operation.
The platform runs; we keep watch. Every further process is easier than the one before it — it connects to the memory the platform has built up about your systems along the way: connections already in place, data mappings already verified, checks already configured. None of your existing processes changes as a result — not without your approval. The next one simply attaches to them faster.
Five rules. They apply to every process.
Your data under your control.
The platform runs in the European Union or entirely inside your own environment; the architecture also supports running with no outbound network access.
Your company controls access.
Sign-in and roles are taken from your company’s own sign-on system — revoke someone there and they lose access here as well. Everyone sees only what they need for their work.
Irreversible steps only after approval.
A rule built into the platform itself — not even an administrator can change it.
Every decision can be backed by evidence.
The full record of a case opens in one click, and you can rely on it in a review or a dispute.
Documentation on request.
Architecture, data flows, data protection, resilience testing, exit plan — prepared for your specialist teams.
A machine can’t be trusted. It can be verified.
You trust a colleague: they have a reputation, professional experience and qualifications behind them. A machine has to be verifiable. We built the platform so that verifying it is straightforward — every step has a record and every rule can be read.
Most processes are built so that the outcome is decided solely by a rule you approved, and in them AI has one job: it reads, it extracts, it prepares the material. In some processes it is worth handing a share of the decision to an AI estimate as well — but where, and whether at all, is decided by you, not by us. With us, AI has exactly as much authority as you give it: its estimate enters the decision with a stated confidence score, your rule sets both the weight and the threshold, and an irreversible step waits for a person’s approval regardless of what AI proposed. And if you would rather not use AI at all, we will set the processes up without it.
We believe automation should relieve people of routine, not of responsibility. Anyone who approves something in a company needs time, information and the right to say no — the platform is built so that they have all three.
A platform built around a single question.
What do you put on the table when someone asks why it was decided this way? For years, inside companies, we saw good systems — and between them, people retyping data from one screen to another. And alongside those were automated processes nobody trusted enough to let near anything that genuinely mattered. The reason was not the technology; it was the missing answer to that question. Cordix Intelligence was built around it.
We are the same founders as at Capila, an accounting and CFO firm whose team does the financial work on the client’s behalf. Cordix Intelligence is a software platform: your processes and your data stay with you, and the platform builds automation, decision-making and the record on top of them.
We are based in Bratislava.
What we get asked most often.
Do we have to change our systems for this?
Wouldn’t an ordinary automation tool do?
Where will our data be?
What part does AI play in this?
How do I know the platform decided correctly?
What if it gets something wrong?
Thirty minutes. You’ll know whether it makes sense.
Send us a few sentences about your systems and about the process that costs you the most time or money. We’ll come back to you within two working days — and from your first message you are talking to the person who will be responsible for your deployment.