A new hire opens your sprinkler-installation SOP for the first time. By page three they've already got questions. Some your content answers if they keep reading. A few they could probably figure out from the context. One or two will leave them stuck. Which ones is which depends on what's actually on the page, what's left to assumed knowledge, and what's hiding in a different document that the new hire hasn't been pointed to yet.
The same thing happens when Mereon reads the same SOP. It pulls out the procedures, the concepts, the decision rules, and some of what it finds is solid while some has gaps or contradicts what's already in another document on your shelf. Until now all of that lived under the hood, and the only way you'd find out Mereon had been confused about something was when a learner got a wrong answer. We've changed that. The new Ontology view shows you what Mereon learned from your operations, and the issues panel shows you the questions it couldn't answer on its own.
What Mereon learns from your operations
Operational training is not "content" in the soft-skills sense. It's the set of instructions someone uses to install a Cobra valve, troubleshoot a Hunter decoder, run a tree-injection treatment, route a customer call. The stakes are concrete. If a procedure step is missing, someone in the field is going to call someone else for the answer, and the gap costs you time, callbacks, or worse.
When you upload a guide, a transcript, or a document, Mereon's first job is to figure out what's actually being taught in it. The output is not a summary. It's a graph: every concept Mereon identified (what a riser is, what continuity testing means), every procedure it recognized (how to install a backflow preventer, step by step), every piece of source content it cited as the basis for each concept and procedure. Procedures are linked to the concepts they depend on. Concepts are linked to other concepts they assume. The whole thing is your operations as Mereon understood them.
That graph is what Mereon uses every time it generates a quiz question, answers a learner's chat message, builds a lesson, or plans a curriculum. The Ontology view makes that graph visible, so when a quiz question looks wrong, you can trace it back to the concept it came from, the source content Mereon cited, and the connections it drew. If Mereon got something wrong, you can see exactly where the misunderstanding lives. If it got something right, you can see the work behind the answer.
The conflicts it finds
Operational documents disagree with each other all the time. A wiring spec in last year's training video says one thing, the manufacturer's updated guide says another, and the SOP in your tools folder splits the difference. None of the authors was wrong when they wrote it. The conflict is what happens when they all sit on the same shelf.
When Mereon adds a new item to your graph, it compares the item against the items already in there. If two of them give contradictory instructions about the same procedure step, or claim different things about the same concept, Mereon files a CONFLICT issue and assigns it to whoever owns that area. The conflict shows up with both sides quoted, the source content for each, and a short explanation of where they disagree.
This matters more for operational training than it does for generic "content." A contradictory paragraph in a sales playbook is annoying. A contradictory step in a wiring procedure or a medication-handling SOP is a safety incident waiting to happen. And learners trust the most recent thing they read, which is almost never the right thing to trust by default. The CONFLICT issue is Mereon's way of saying: a human needs to pick the correct version before this becomes a question on someone's bad day.
You don't have to wait for the next training session for these to come up. As soon as a new item lands in the graph, the comparison runs, and if a conflict exists, someone gets pinged.
The gaps it asks about
Every night, Mereon does something a content-management system can't. It walks through every topic in your knowledge base from the perspective of a new learner who's encountering it for the first time. What questions would that person have? Where would the existing content give them an answer? Where would they hit a wall?
When Mereon hits a wall, it files a CONTENT_GAP issue. The issue includes the question that wasn't answered, the reasoning for why a learner would ask it, and a routing to whoever owns the topic. The next morning, the people who maintain the content wake up to a short list of things worth fixing, with the gap already named and the right person already on the hook.
The shift here is about cadence. Most teams find out about coverage gaps at training time: a new hire asks something obvious, the trainer realizes the deck doesn't cover it, an "we should add that" gets added to a list that no one re-reads. The gap stays in the content until the next hire trips on the same edge. Mereon collapses that loop into one overnight pass, and pushes the result at the owner instead of waiting for someone to remember to ask.
You can also tell when Mereon's confident. If a topic generates zero gaps for several weeks, that's Mereon saying the content is genuinely comprehensive. If the same topic keeps generating new gaps every nightly run, something about how that topic is taught isn't holding up. Both signals are useful, and neither requires you to read every nightly summary line by line.
Why showing the work is the point
Opaque automation is a brand risk in operations. When a learner gets an answer that turns into a callback, an injury report, or a compliance finding, "the system said so" is not an answer the business can accept. Someone has to be able to say what Mereon considered when it answered, and where it came up short.
Mereon's graph gives you that audit trail. Every answer traces back to source content. Every quiz question traces back to the concept it tests, and the source the concept came from. Every issue in the issues panel is Mereon proactively naming something it isn't sure about, with the work shown. None of that is a UI gimmick. It's the receipts.
Internally we describe this as "showing the work." The conflicts Mereon surfaces and the gaps it asks about are evidence that it is in fact doing the work. A system that never files a single issue against your operations is either reading content so flat that there's nothing to find, or it's not paying close enough attention. The teams getting real value from Mereon are the ones where it's filing issues and getting them resolved, week after week, on content that was already considered finished.
See what Mereon's reading
The Ontology view is live now for every active organization. The fastest way to see this in action is to open it and pick a topic you know well. Look at the concepts Mereon extracted. Look at the source content it cited. Then check the issues panel for that topic. Either you'll find Mereon agreeing with what you'd expect, or you'll find a conflict or a gap worth resolving. Both outcomes are the system doing its job.
Open your Ontology and tell us what you find.
