Lenexus wants to map everything that could take your business down

Some ideas feel obvious in hindsight, but nobody had really executed them well until now. Lenexus, a platform built by MJ Corp, is one of those: map every system, supplier, person and AI a company depends on into a single living graph, so you can finally answer the question most organizations struggle with: if something breaks, how far does it spread, and what does it cost? I spent time in their demo. Here's what I found. The problem they're solving In most companies, the information you'd need already exists, just scattered. The ERP knows the suppliers. IT service management knows the servers. HR knows the people. The AI platform knows its own models and agents. But when a supplier goes down, a key employee leaves, or a cloud service gets compromised, none of these tools alone can tell you what actually breaks and what it's going to cost. It's a problem that sounds simple until you try to solve it properly. What the demo actually delivers What sets Lenexus apart from typical risk management software is how much of it is real and working, not just mocked up: - A full, navigable dependency graph with hundreds of typed nodes (systems, suppliers, people, business processes) and explicit relationships like depends on, runs on, supplied by, knows. - An operational digital twin that scores every element by criticality and gives you a live health picture of the whole organization. - Attack and failure simulation that traces a realistic compromise path, calculates the expected financial impact (average and worst case), and ranks countermeasures by how much damage each one prevents. - A financial decision simulator where you adjust levers like pricing, headcount or costs and see the projected impact on the income statement, paired with a plain-language analysis that flags risks and gives a nuanced recommendation instead of a flat yes or no. - A document intelligence module, which is arguably the most impressive piece. You paste in free-form text, an incident note or a snippet of documentation, and it extracts entities and relationships, links them to known failure points in the graph, and suggests actions. Nothing gets written to the graph without human approval, which avoids the usual failure mode of AI quietly inventing connections. I tested that last module with a fictional internal document written as loose, unstructured prose. It correctly picked out around fifteen entities and twenty relationships, with a level of nuance well beyond simple keyword matching. The most interesting bet: bring your own model One of the more promising things about Lenexus isn't visible on screen, it's architectural. The AI layer that interprets, summarizes and explains results is separated from the deterministic engine that actually produces the numbers. The risk scores, the impact figures, the way things propagate through the graph, all of that stays fixed and verifiable. What changes is the quality of the explanations and recommendations wrapped around them, depending on the language model plugged in behind the scenes. In practice, that means the platform is built to get better on its own as AI models improve, without needing to rearchitect anything. The stronger the model you plug in, the sharper the analysis gets. It's a smart bet on where the field is heading rather than locking the product to one AI vendor. What's still unproven To be fair, Lenexus is still a demo product, not yet a production platform with publicly confirmed paying customers. A few open questions remain: - How does it hold up against messy, incomplete, real-world data instead of carefully curated demo datasets? - What's the actual go-to-market plan and timeline? Worth addressing directly: there's a rumor going around about an imminent Google acquisition. I couldn't find any verifiable source backing that up. Treat it as unconfirmed for now. Bottom line Judged purely on the product, Lenexus is executing at a level well above what most startup demo pages deliver. The combination of dependency graphing, financial simulation and AI-driven document extraction is coherent, well thought out, and, rare for a demo, actually seems to work rather than just look good in a screenshot. The deterministic-engine-plus-swappable-AI-model architecture is a particularly smart call. It's built to age well. Whether the go-to-market execution matches the product quality is still an open question. But for a company still in demo mode, it's hard not to be impressed. Top comments (0)