AI can do more when it understands the work. Kai uses Kaduno’s business data, with limits set per store and changes you confirm.
A practical view of the process. The exact setup depends on your business and the capabilities you use.
Open Kai with “Spør Kai” at the bottom of the page, or with Ctrl+J (⌘J on Mac). Ask about an order, product or purchasing need. Kai knows which page you are on. What you get answers on depends on your licence and permissions.
When Kai is about to change something you get a card showing what changes, before and after. For bulk changes, such as a price change across many products, you open a full review where you can untick rows you do not want to change before you confirm.
Kai makes no change unless you confirm the card. The exception is settings changes while the store is being set up, which Kai makes directly for an administrator. Once applied, a change can be undone where undoing is possible. High-risk cards can require you to type a confirmation.
Configuration, rules and boundaries to review before rollout.
Under “Hva Kai får lov til” (what Kai is allowed to do) the store sets the highest level Kai may propose: Kun lesing (read only), Lesing og utkast (read and drafts), Trygge endringer (safe changes) or Alt (everything). Starting on read only or drafts is a good way to build confidence and agree workflows. Kai writes drafts of texts, but never sends or publishes on its own.
Changes are divided into safe changes and high risk, and the card marks them “Trygg skriving” (safe write) or “Høy risiko” (high risk). The levels are risk categories, not a promise that every change is harmless. Actions above the level the store has chosen are not proposed by Kai at all; Kai says they must be done manually.
The store can also list tools Kai may not propose (one prefix per line, for example “payment.”). What Kai may change is decided in Kaduno, not by the instructions Kai has been given: Kai cannot change its own limits, budget or security settings.
The store’s stage (automatic, under setup or live) decides, among other things, whether Kai can make settings changes directly for an administrator. Review the stage and the limits before the team starts using Kai.
Not manually for everything, but changes always go through a card you confirm. What Kai can propose is decided by the store’s limits. We review the levels and which actions suit you.
Yes, staff can connect ChatGPT, Claude or another assistant to the store. An assistant that proposes a bulk change does not get it written itself; the proposal lands as a review with the person it acts for. The store chooses whether customers’ personal data is sent in full, masked or hidden. Kai in the admin always sees all data.
We choose a task you actually have, show the context and review proposals, permissions and policy. That is more useful than a generic chatbot demo.
Tell us where it hurts — and we will show you how Kaduno solves it. A 30-minute demo, in Norwegian or English.