Kognitos - Write automations in plain English. They run, exactly.

by
Most automation makes you translate your process into brittle scripts or flowcharts. Most AI automation guesses. Kognitos does neither. You write the rule in plain English, and that English IS the program. A symbolic engine runs it exactly as written, every time, with a full audit trail and zero hallucinations. "If an invoice is over $5K, route it to the CFO." Write it. It runs. Built for finance, AP, and back-office ops. Start free.

Add a comment

Replies

Best
Hunter
📌
Hi Product Hunt. We built Kognitos because automation has always forced you to translate. You know your process in plain English, then someone turns it into an RPA bot, a script, or a flowchart that breaks the moment a screen changes. And the new wave of AI automation has a different problem: it guesses, which is fine for a draft and not fine for an invoice. So we made the English the code. You write a rule like "process today's AP invoices, match them to POs, and flag anything over $5,000," and that sentence is the actual program. A symbolic execution engine runs it exactly as written, logs every step, and never improvises. When it hits an exception it does not understand, it asks, learns the answer, and remembers it forever. We call it English as Code. It is deterministic by architecture, so there are zero hallucinations and a complete audit trail your finance and compliance teams can read without a developer. That is why it is already running in production at large finance and operations teams. You can start free today and build your first automation by describing it. I would love your feedback, and I am happy to go deep on the neurosymbolic side if that is your thing. What is the most tedious back-office process you would point this at first?

One thing I'd love to see is a community marketplace where users can share their natural language automations as templates. Would help newcomers get started faster and let power users showcase creative workflows they built.

The error handling actually feels closer to how a teammate would react than a typical bot, which surprised me. Setup was quick and the natural language input didn't require me to learn any weird syntax.