Most “AI in the product” amounts to a chat box bolted onto a corner of the screen that has no idea what you were doing. Kadaikodi took the opposite route: the assistant lives in the app shell, not on one page, so it is present on the customer discovery grid, the merchant catalog, the worker task list, the investor portfolio, and the platform operator’s control room alike — and it is handed your location in the product before you have finished typing.
That location is deliberately structured, not a guess. Kadaikodi resolves the current route into a concrete entity: open a merchant storefront and the assistant knows it is looking at a service provider; open a delivery you are tracking and it knows the order id; open a worker task, an investment opportunity, or a holding in your portfolio and it knows which record that is. Seven route shapes across the app map to a typed entity this way, and every prompt also carries which of the five consoles — customer, merchant, worker, investor, admin — you are standing in, because “what should I do next?” means something completely different to a rider mid-shift than it does to an investor reading terms.
There is also a full-page AI workspace at its own route in the app for the longer conversations that do not belong in a floating panel — the “walk me through onboarding my shop” or “explain what these analytics are telling me” kind of session — sharing the same product identity and the same permissions as the widget.
What the assistant is careful not to do is invent marketplace facts. Everything it can read, it reads through the same permission-checked API the screens use, so it never sees an order, a roster, or a holding that the person asking is not allowed to see. And because Kadaikodi is a multi-role marketplace where one human can be a customer in the morning and a merchant by evening, that scoping is doing real work, not decoration.
This part is genuinely live in the app today — the widget, the shell mounting, the route-to-entity context, and the dedicated AI page all exist in code. What is not live is a Kadaikodi-specific operating brief that teaches the assistant the marketplace’s own vocabulary and its destructive-action etiquette; that is authored for a handful of Burdenoff products so far and is still on the queue for Kadaikodi.
Pre-launch platform: this describes shipped plumbing in a product that has not opened to the public yet, not a record of real usage.
Do it yourself
Use the floating assistant from whichever Kadaikodi console you are in — and see that it already knows the page you are on and the merchant, order, or task open in front of you.
Sign in and open your dashboard. Look for the floating assistant button in the bottom corner of the screen.
You should see: The assistant is available here and on every other Kadaikodi route — it is mounted in the app shell, not on one page.
Ready to make this your story?
