From their base in South Africa at CohesionX, Michiel du Toit (Chief Innovation Officer) and his team build AI that actually works in business. The GenAI conversation is full of buzzwords — agents, copilots, and “LLM ops” — but most teams still hit the same wall: the data they need is scattered across spreadsheets, databases, PDFs, and Sharepoint. “That’s why we built CXQL: a language that lets teams ask better questions of all their data and get usable answers back.” he says. 

CXQL feels familiar to anyone who’s used SQL, but it’s designed for modern AI‑first workflows. It can read files, connect to data sources, and apply AI prompts to summarise, classify, and structure information. 

CXQL feels familiar to anyone who’s used SQL, but it’s designed for modern AI‑first workflows. It can read files, connect to data sources, and apply AI prompts to summarise, classify, and structure information. Instead of stitching together scripts and prompts, you use one query to turn messy inputs into a clean, decision‑ready output. 

Here’s a simple, real‑world example: Imagine a merchandising team receiving hundreds of suppliers’ documents every month. With CXQL, they can point it at a folder of PDFs, ask for one‑sentence summaries, and flag anything that fails an allergen policy. The result? A neat table of products that need review, without weeks of manual reading. 

The result? A neat table of products that need review, without weeks of manual reading. 

Another scenario: a payroll or HR team needs to reconcile exports from multiple systems. CXQL can pull the CSVs, detect missing fields, and ask the model whether a column looks like a “benefit” or “deduction” field. It then produces an exception list for the payroll team to resolve. Fast, auditable, and repeatable. 

Legal teams can use the same approach to sift through agreements and extract termination clauses. 

Legal teams can use the same approach to sift through agreements and extract termination clauses. Operations teams can combine sensor readings, field notes, and order data to spot anomalies across regions. Risk teams can enrich due‑diligence datasets with web‑search context and quickly surface anything that looks suspicious. These are exactly the kind of workflows CXQL was made for: high‑volume, high‑impact, and time‑sensitive. 

“The practical magic is that CXQL turns AI output into rows and columns that people can act on.” 

“The practical magic is that CXQL turns AI output into rows and columns that people can act on. Instead of a paragraph of text buried in an email, you get a clear list: which items passed, which need attention, and why. That makes it easier for business teams to review, approve, and move forward without chasing down a data engineer every time they need a new answer. 

Behind the scenes, CXQL also supports data profiling, recurring schedules, and clean mapping into Postgres schemas. But the key is the experience: you get AI‑augmented insight without building a new data pipeline for every question. 

CXQL is how CohesionX delivers “invisible augmentation”: AI that fits into existing workflows and quietly lifts the quality of every decision. If you’re ready to put AI at the center of your data stack, we’d love to show you what CXQL can do. 

Visit cohesionx.co.za/contact/ to discover CXQL magic for yourself!