Skip to content
ZANARIX
All work
  • Concept

Huolvera.

An enquiry inbox for a fictional property maintenance company: every customer message sorted by topic and urgency, with a suggested reply that a person checks, edits and approves before anything is sent.

Try the demo

A working inbox with simulated AI and sample data. Nothing is ever sent.

Type
Enquiry automation · Inbox that sorts messages and drafts replies
Scope
Identity · Inbox app · Sorting rules · Reply drafts
Company
Fictional, created by Zanarix
Working demo
Live · Simulated AI · Sample data
Service desk inbox · Working demo · Simulated AI · Sample data

The problem

Every message looks the same in an inbox, until it's a leak.

We set ourselves a brief: a property maintenance company looking after apartment buildings. Its service desk gets dozens of messages a day, from burst pipes to invoice questions, all in one shared inbox and in arrival order.

An urgent leak can sit under ten routine questions. Staff write the same replies again and again, and when someone is off, nobody quite knows what is waiting for whom.

The idea and approach

Let AI do the sorting and the first draft. Keep people in charge.

Huolvera is an invented name. The identity is practical and calm: a hex-nut mark with a small roof inside, harbour blue for actions, and colours kept for meaning: brick only for urgent, signal yellow for today.

Each incoming message is sorted by topic and urgency, with a plain-words reason, the key details picked out, and a suggested reply. Nothing goes out on its own: a person reads the draft, edits it if needed and approves it. Anything the AI can't know, like a technician's arrival time, is flagged to check first.

Key features

What the demo does.

All of it works in the demo, with scripted answers instead of a real AI.

  • Sorted by urgency

    Urgent, Today or Normal, with the reason in plain words, so a leak never waits behind an invoice question.

  • Sorted by topic

    Leaks, heating, repairs, keys, billing, yard or safety, with filters and counts for each.

  • Details picked out

    Address, problem and timing lifted from the message, plus a suggested next step such as an urgent job.

  • Suggested replies

    A draft for every message, in the company's tone. Edit it freely, or ask for another version.

  • People approve

    Nothing is sent without a person's approval, and things to check first are clearly flagged.

  • Corrections

    If the sorting is wrong, staff correct it in two clicks, and the correction is visible to the team.

Identity and how it works

Fast for the team, careful with customers.

Huolvera's identity, the path of a message, and the sorting rules the team controls: a fictional company, designed by Zanarix.

Huolvera Property Care

Primary logo

Mark on Slate

Palette

  • Harbour

    #2F6690

    Primary, actions

  • Slate

    #1B2733

    Text

  • Frost

    #EEF2F5

    Background

  • Signal

    #E9B949

    Needs doing today

  • Brick

    #C2553D

    Urgent only

  • Lichen

    #6E9E7F

    Done, replied

Typography

Aa

Instrument Sans

Plain and sturdy, easy to read in a busy inbox and on a phone in a stairwell.

Guidelines

  • Brick is only for urgent. If everything is red, nothing is.
  • Every colour comes with a word: Urgent, Today, Normal.
  • Replies sound like a helpful neighbour, not a ticket system.

The path of a message

The AI does the sorting and the first draft. A person always makes the decision.

  1. A message arrives

    From the company's normal inbox, just as it does today.

  2. It's sorted

    Topic and urgency, with the reason in plain words and the key details picked out.

  3. A reply is drafted

    From the company's approved answers, with anything uncertain flagged to check.

  4. A person approves

    Staff edit if needed and approve. Only then is anything sent.

Rules the team controls

The company decides what counts as urgent and who is told. The AI follows these rules; it doesn't invent its own.

Sorting rules · Concept screen · Sample data
Try the demo

A working inbox with simulated AI and sample data. Nothing is ever sent.

Design highlights

Colour only where it means something.

  • Urgency in words

    Every urgency label has a word next to its colour, so nothing depends on colour alone.

  • The reason is always shown

    Staff can see why a message was sorted the way it was, and disagree with it.

  • Drafts, not auto-replies

    The reply sits in an editable box with one clear button to approve it.

  • Works on a phone

    On-call staff can open, check and approve a reply from their phone, one message at a time.

Technology

How we would build it.

The demo runs on scripted answers. For a real company, we would build it like this:

Email connection
Messages read from the company's existing service desk inbox; replies sent from the same address.
AI model
A language model from a provider chosen with the client, with EU data handling, sorting against the company's own topics and rules.
Approved knowledge
Drafts written only from the company's approved answers, prices and procedures, never from guesses.
Human approval
Every reply approved by a person; urgent messages also notify the on-call phone.
Next.js + TypeScript
One web app for the inbox, on desktop and phone.

What it would do for a business

What an inbox like this is designed to do.

For a real service desk, this approach aims to:

  • Put urgent messages in front of the right person first.
  • Save time on routine replies, while a person still checks every one.
  • Make it clear to the whole team what is waiting and how urgent it is.
  • Keep replies consistent, whoever is on shift.

Intended outcomes, not measured results.

Services involved

Each one links to more of our work in that service.

All concepts, designs and content © Zanarix. All rights reserved.

All brands, products, people and data in Zanarix concepts are fictional. Any resemblance to real companies is coincidental.

Next step

Is your inbox doing the sorting for you?

Tell us what kinds of messages your team handles every day. We'll show you what could be sorted and drafted, and what should stay with people.

Start an automation project