Apps, RPA and AI agents from one team. We add AI where it solves a problem — elsewhere, good code is enough. Since 2017.
Three things you would normally buy from three vendors. We deliver them separately, or as one.
An RPA bot where clicking instead of a human is enough. An LLM agent where the work needs to understand a document, an email, a context.
Dashboards, B2B portals, mobile apps for field teams. Full-stack, built to still work in five years.
Deploying AI now comes with rules. We get the systems, records and oversight into a state that survives an audit.
Not portfolio samples. Software we develop, host and sell under our own name — and use ourselves.
How many people can still run a critical process without AI — shift by shift, with every number traceable and a forecast of when capacity drops below a safe line.
A realtime learning management system for companies that train their own people — and need to know who actually completed what.
Deploying AI without knowing how much human capability is left behind it has two consequences. You feel one in operations, the other under inspection.
Automation does not remove people all at once — it removes the ones who can still do the work unaided. Headcount looks unchanged while real coverage falls, often to a single person per shift. You find out during the outage.
Article 26(2) requires the people assigned oversight to have the competence, training and authority to actually exercise it. A nominal assignment does not satisfy it. Article 14(4) additionally names automation bias as a risk to be managed. The evidence has to show the person could intervene — not that they are on a list.
Regulation (EU) 2026/1744 deferred the high-risk obligations to 2 December 2027 for standalone Annex III systems and 2 August 2028 for AI embedded in regulated products. That is a deferral, not a repeal — and measurements cannot be produced retroactively.
Read this as a deadline, not a threat. AI Act penalties attach to obligations that only apply to high-risk systems from December 2027 and August 2028 — until then it is time to prepare, not an enforceable fine. NIS2 and GDPR duties apply today. For SMEs the lower of the two figures applies. The tool produces measurement, evidence and a checklist.
Eight quantities for every capability in scope. No language model computes any of them — the calculation is deterministic and every figure decomposes back to its inputs.
Six weeks from first scoping to the executive readout. We assume no particular MES, CMMS or type of AI — we map what you already have.
Effort on your side: roughly 5–7 person-days, mostly from whoever owns the records — not necessarily IT. We do not need names, an employee_id is enough. We do not take performance appraisals, disciplinary records, or the content of AI conversations.
Today it is an assessment you re-run whenever the inputs change. Connectors are what turn it into continuous measurement — that is the next phase, not today's state.
Manual entry stays in every phase. Some plants will never allow a connector onto their network, and they still need to be able to run the assessment. Integrations are a separate commercial and security conversation — network access, a data processing agreement, the works council.
We built it because most LMS platforms either cost like an ERP or cannot tell you whether a learner actually watched the video. It runs in production at elearning.batacon.sk.
Course translation through a language model runs today, over an OpenAI-compatible API — so it works equally with AWS Bedrock or a model on your own hardware. The rest we add where it earns its place against your training volume.
In production today: course translation. The other three are custom work on top of the existing system — we will estimate them once we know your volume and languages.
An LMS earns nothing if people need another password for it and HR retypes the roster by hand.
The answer is the same whether we build an app, a bot, or the whole stack.
Thirty minutes on a call beats three months of internal debate. You do not need a finished spec or an approved budget — just a process that is not working.
info@batacon.eu +421 908 996 704