· 8 min

AI automation: where it pays off in 2026 (and where it still does not)

The subject moved on from chatbots to processes: models that read, classify, fill in and execute. In practice, AI automation only returns money on processes with volume, a clear rule and a reversible error. This article covers the four places where it pays off in Brazil today, the four where it does not yet, and the cost-of-error maths almost nobody does before signing.

Read article →

· 9 min

RAG with Docling: making AI answer questions about your own documents

Asking a model about your company documents only works if the documents were read properly first. In practice, most wrong answers in RAG projects do not come from the model: they come from extraction that turned a table into alphabet soup. This article explains what RAG is, why Docling changed the reading step, how to anonymise before sending, and how to measure whether the result is any good.

Read article →

· 8 min

Querying the database in plain language: what works in NL to SQL

After documents, the natural next question is the other half of the company: the structured data already sitting in the ERP, in finance and in the CRM. Translating a plain-language question into SQL works better than expected on tidy databases and worse than promised on real ones. This article shows what separates the two cases, why the semantic layer matters more than the model, and how to open this up without opening the database.

Read article →

· 9 min

Brazil tax reform 2026: does your system already issue IBS and CBS? What changed, what split payment brings and how to check

Since 3 August 2026, SEFAZ rejects NF-e and NFC-e from regular-regime companies without the IBS and CBS fields. In September, Simples Nacional companies choose between the standard and the hybrid model. In 2027, split payment begins. This article sums up what changed, what still changes and gives a checklist to know whether your ERP, e-commerce or invoicing system is ready.

Read article →

· 8 min

AI on legacy systems: what works in Brazil in 2026

The AI application that returns the most money in Brazilian companies is not the chatbot: it is using language models to understand, document and test old systems in PHP, Java, .NET and COBOL. This article separates what works today from what is still a promise, and explains how JBKR Web uses AI in legacy modernization, always with human review.

Read article →