Automation · live demo
Try it right here: pick a monthly job, watch it run step by step, then download the result as a real Excel file (.xlsx).
The real tool, not a simulation
The “Invoices ↔ payments” job above, as an app I built: drop in two spreadsheets and get the matches, the gaps and the differences, plus an XLSX report — offline.
129 automated checks
How it works
01 Reader
It starts from what your systems already export — Excel, CSV, bank statements, supplier files. Your team keeps its tools.
02 Rules
Written down with you in the audit, in plain language. You can read every rule — no black box.
03 Checker
Anything missing or unusual is flagged for a named person to decide. Nothing is quietly “fixed”.
04 Writer
The file you already use, with the same columns, ready before you need it. Every run leaves a trail you can check.
The demo runs all four steps as plain rules. In your version, a step that has to read free text — emails, PDFs, supplier notes — can be an AI agent: a small program that uses an AI model for that one step. Everything it produces still passes the Check step before it reaches the file.
Pricing
Every project starts with a short conversation and, where useful, an audit of your files or terminals. You get a written, fixed quote before any work starts — no open-ended billing.
These are not trend pieces. Each one has a date attached, and the date is not mine to move. The shape of the job is almost always the same: pull data out of systems that were never designed to talk, chase the suppliers who did not answer, and produce a filing that survives an auditor reading it backwards. Chasing suppliers for documents and reconciling what they send is the core of it — work I did in procurement, customs and import paperwork included.
Definitive regime since 1 January 2026; authorised declarants file an annual CBAM declaration — the first one covers 2026. The engineering problem is not the form — it is collecting embedded-emissions data from dozens of suppliers in a dozen formats, and chasing the ones who go quiet. Ingest → match → flag the gaps → chase automatically → draft the annual declaration.
Applies from 30 December 2026 for large and medium operators (30 June 2027 for micro and small enterprises). Geolocation and legality evidence per supplier, a risk assessment that holds up, and a due diligence statement. Mostly supply-chain paperwork, a little software — which is the right mix for me.
Since 1 January 2025, invoices issued to consumers (B2C) must be sent through RO e-Factura (Government Emergency Ordinance 69/2024). This is not another invoicing tool — the market is full of those. It is the reconciliation layer: what your billing system issued, versus what actually reached the state portal, versus the difference, versus the days left.
High-risk obligations for stand-alone (Annex III) systems — including AI used to monitor and evaluate workers — apply from 2 December 2027, after Regulation (EU) 2026/1744 moved the date. If what I build for you uses AI, the log trail, the human-oversight record and the who-approved-what evidence come with it. Not sold separately — it is a by-product of building it properly.
Applies from 28 June 2025 and covers e-commerce among other services. My website builds include a WCAG 2.1 AA audit; for businesses in scope, that audit stops being a nice-to-have.
Dates and scope above link to official sources. For these five regulations I do not quote penalty amounts: the figures circulating in commercial blog posts did not survive checking, and I am not going to sell you on a number I cannot stand behind. Where this site does quote a fine — on the LumAIstra Pontaj page — it comes straight from the statute, with the source next to it. If a penalty figure matters to your decision, we verify it together against the legislation before it goes anywhere near a proposal.
Recurring work where a person is the connector between two systems. The three I am asked about most: the report rebuilt from four or five exports before every meeting; the supplier who sits in your system four times under four spellings, so nobody trusts the spend figures; and supplier documents chased and reconciled for a filing such as CBAM or EUDR.
Describe yours in two sentences and I will tell you honestly whether it fits — or who fits better.
Five working days on one recurring workflow: I map the sources, the steps and the exceptions, and come back with a build plan and a fixed quote. The fee is credited in full against the build.
If personal data is involved, we sign a short data processing agreement before I look at anything.
Honest answer: I have no hands-on experience with those systems, and I won’t pretend otherwise.
What I bring is the data side: I work from the exports and master data any ERP produces, reconcile them against the files people actually trust, and build the checks around them. If you need a certified consultant or a pre-built connector for your ERP, I’m not your vendor.
Both. I use AI where language is the problem — reading, sorting, drafting — and plain rules where rules are enough. LumAIstra Pontaj, for example, is rules only: no AI scoring of people.
The data is fictional and generated in your browser; every step in the demo is a plain rule, and no AI model runs. The .xlsx and .csv files are created on your device — nothing is uploaded or stored. You can also try the timesheet or the supplier-quotes job with your own CSV (each has a template): only your browser reads it.
Prefer email? [email protected] · WhatsApp +40 752 494 070