ArlemInsights.
Short notes on the record, the numbers and the agents, written from running these systems for our own companies first.
Your database is only as good as its data entry
Ordered data is not a reporting problem. It is what makes every future decision, hire, tool and AI project cheaper. And it is won or lost at the moment someone types something in.
September 2026 · 6 minRead Ratio · Vigilia Analytics & BIAnalytics and BI without the big-data theater
Most companies do not have big data. They have messy data and three questions they need answered every morning. Here is what analytics and business intelligence actually are at that scale, and when the heavy tools earn their keep.
September 2026 · 6 minRead Legatus AI agentsAn AI agent in the group chat: a Telegram walkthrough
What it looks like when an AI agent joins the team as a member rather than a tool: taking dictation, checking the database, issuing documents, and asking before it writes. A short walkthrough with a tea wholesaler as the example.
September 2026 · 7 minRead Legatus · Tabula GuardrailsPreview, then confirm: letting an AI write to your books without losing sleep
The question every owner asks before an agent touches real data: what if it gets it wrong? The answer is a design rule, not a better model. Model proposes, person confirms, code writes.
September 2026 · 5 minRead Ratio ArchitectureThe model never does the math
A short architectural rule that makes AI adoption boring in the best way: language models decide what to do, code computes the numbers, the database holds the truth. Why the split matters and what it looks like in practice.
September 2026 · 5 minRead