We automate the flow of documents, invoices, approvals and reports, and the data moving between your systems. AI goes only where content has to be read and understood.
Each step of the process gets the cheapest tool that does it predictably: a rule, code, AI or a person.
Documents and messages land in the right place on their own: read, classified and attached to the right case instead of being retyped by hand.
Purchase and sales invoices, matching payments from bank statements, and settlements. Amounts are always computed by code, never by the model.
Requests, approvals and hand-overs between people follow clear rules, with deadlines and a record of who approved what.
Reports that someone assembles by hand from several files today are generated automatically from the source data.
Integrations move data between the company’s software without copying and retyping.
Rules and code cost less than AI and always give the same result. So AI only gets the steps that rules cannot handle.
Routing a document by sender, an approval deadline, an amount threshold, moving data to another program: steps like these are done by a rule or code.
AI gets documents in changing formats and content that has to be understood: invoices from different suppliers, letters and customer messages.
Totals, settlements and payment matching are computed by deterministic code, not the model. The same data always gives the same result.
A low-confidence reading or a case outside the rules goes to someone on the team, together with what has already been read. Every decision leaves a record.
Documents arrive by upload, e-mail and bank statement import. AI reads, classifies and books them in the mode the client chooses: the automatic mode the client switches on, or approval with one click. A deterministic engine computes taxes, VAT and ZUS, and low-confidence cases go to a person.
Explore QkwitBrokik settles rent and utilities with invoices, and property managers import bank statements instead of retyping them. A setup assistant drafts property and tenant records but saves nothing without approval.
Explore BrokikWe sit with the people who do the work and write the process down step by step: where documents come from, who checks them and where the data goes.
We assign each step a rule, code, AI or a person, and pick one process to start with.
We automate it end to end on your documents and check how much manual work disappears before we add the next one.
We keep the automation running, adjust the rules when documents or regulations change, and add further processes.
Example: a purchase invoice arrives by e-mail. A rule routes it to a department by sender, AI reads the line items and amounts, code checks the totals and matches the payment from the bank statement, and an invoice with a mismatched amount goes to someone in finance for review.
The ones that repeat often and follow clear rules: the purchase invoice flow, routing incoming mail, a weekly report built from the same files. We start with one such process and automate it end to end before adding the next. We work inside the programs you already use, without replacing them.
We agree the price with each company separately, for a specific scope: which processes we automate and how many programs have to be connected. We learn the scope in the first conversation or in an audit, and that step comes with no commitment.
The time depends on the process and your documents: how many kinds there are, how many exceptions, and the state of the data in your programs. We learn that while we write the process down with you, so we don’t give a timeline before then.
We process your data in the European Union. When the project ends, you get everything: the code and the data.
No. First we check whether a rule or code can do the step: routing a document by sender, approval deadlines, moving data between programs. AI comes in where a document in a changing format has to be read or a message has to be understood.
Deterministic code, never the AI model. AI can read an amount from a document, but code computes and checks the totals, payment matching and settlements, so the same data always gives the same result. That is how Qkwit works: an engine computes taxes, VAT and ZUS, not the model.
It goes to someone on the team, together with what has already been read and what is missing. We design confidence thresholds and the exception queue from the start, and every decision leaves a record: who, when and on what basis.
Process automation
Every back office is full of repetitive work, but not every process has the same potential for AI automation. We ranked candidates by where rules are clear and volume is high — not by whatever is currently annoying the team most.
Process automation
AI rollouts rarely fail because of the model. They fail on integration — data sitting in three systems that were never built to talk to each other. That's the most common reason a promising project stalls for months before anyone sees a real result.
Process automation
The first time a company gives AI access to its data, the security question always comes up — and it should. This isn't about being afraid of it; it's about setting a few concrete rules before AI gets access to anything.
