One Trello card asked for something that would show — not tell — how a Danish company can hand work to an AI colleague. The answer was cases.grn.dk: a self-contained interactive page where the visitor types a task, watches an AI pick it up, gets asked a question, answers it, and sees the finished card land at the top of Done. The brief, the domain, the design, the copy, the sound and the publication were all handled by the AI setup the page describes.
- Client
- GrN.dk — internal build
- Sector
- AI workflow automation
- Period
- July 2026
At a glance
The challenge
“AI can take work off your desk” is easy to claim and hard to feel. A screen recording proves nothing — anything can be edited — and a written explanation of labels, queues and hand-offs loses a business owner in the first paragraph.
The brief was explicit about that: not a video. An interactive page, in Danish, aimed at Danish companies, that shows the whole loop — task in, AI picks it up, AI asks when it is unsure, human answers, finished work lands in Done — using sound and motion where they help rather than decorate. And it had to be published on its own domain.
The solution
cases.grn.dk is a single hand-written HTML file: 24 KB, no framework, no build step, no external assets, no tracking. Everything the page needs is inline, down to the emoji favicon.
The centre of the page is a working kanban simulation that borrows the real vocabulary of the live workflow instead of mocking one up:
- the visitor types a task — or taps one of four presets — and the card drops into To-do;
- a robot picks it up, the yellow
ai:workinglabel snaps on, and a log types out what it is doing; - one task hits a genuine decision, slides to Venter på dig with the orange
ai:waitinglabel and asks its question on the card — the visitor answers with a button and unblocks it; - the finished card lands at the top of Done with a green label, a result note and a chime.
Every sound is synthesised in the browser with the Web Audio API, so there are no audio files to download, and the Danish narration uses the Web Speech API with a da-DK voice when the visitor's device has one — with a mute toggle in the corner for when it does not fit the moment. Below the demo, four steps explain the loop in plain Danish, and then the page makes the point that matters: it was ordered exactly the same way, and the card that ordered it is sitting in Done.
Delivery included the hosting, not just the page: an OpenLiteSpeed virtual host with scripting switched off — there is no server-side code to attack — plus a Let's Encrypt certificate and an HTTP-to-HTTPS redirect.
The results
The page has been live since 19 July 2026 and does in under a minute, on a phone, what normally takes a sales call: the visitor runs the workflow themselves, including the part where the AI stops and asks a question, which is the part people do not expect.
Because it is a static file, it costs effectively nothing to run, shrugs off a traffic spike, and has no database, plugin tree or admin login to keep patched.
And the build is its own evidence. The domain, the TLS certificate, the interaction design, the Danish copy, the synthesised sound and the publication all came out of one Trello card — the same loop the page lets you play with.
Not a video, but an interactive page or whatever it could be best practice mindblowing wow factor 2026.