Skip to main content
Quorum
ExampleExample case studies

Quillwren works in writing. Now it knows where the writing waits.

Quillwren builds heat-pump monitoring devices and the software behind them, with 40 people spread across Europe and almost no meetings. Quorum's job here was the other half of communication: where written work waits, and whether the team keeps the response times it agrees.

Industry
Heat-pump monitoring: devices and software
Location
Remote across Europe, registered in London
People
40
Work model
Remote-first, async by default
Tools
Google Workspace, Slack, GitHub, Linear, and GitLab

The situation

Meetings are rare at Quillwren: 3.4 hours per person per week. Decisions happen in Slack threads, pull requests, Linear issues and a GitLab project for the firmware. When something stalled, nobody could tell whether it was a slow repository, a busy team or just a quiet week.

Coding agents now open a share of the pull requests — about 25% in this example, 128 of 477 — and the founders wanted to know whether those waited longer for review. What they refused to build was a list of who answers slowest.

What they set up

  • Google Workspace calendars, read-only.
  • Slack with thread consent: counts and times only, never message text.
  • 4 GitHub repositories, 3 Linear teams and the GitLab firmware project, each bound to the workspace by its own webhook secret.
  • Pull requests recorded by the kind of author — a person or a coding agent — never by the account.

What Quorum showed

Quorum read the workspace as async-first — 3.4 meeting hours per person per week, with written work carrying most of the coordination — so the Overview led with threads and response times. The reading changes the order and the headlines, never a figure: a reader who scrolls finds the same numbers whatever leads.

Computed by Quorum on the example workspace modelled on Quillwren.

£53,092
Thread cost, last 30 days
Modelled Reading and writing time modelled from message volume. 708 of 752 conversations priced.
£50,222
Meeting cost, last 30 days
Measured Booked time priced by cost band. 399 meetings; 305 with attendance measured.
1.3 working h
First reply
Median over 694 conversations; 79% answered within a working day.
99%
Focus days
Person-days with a long enough free block, over 28 days.

Meeting cost is measured and thread cost is modelled. Quorum shows them side by side and never adds them together.

Places with a published response timeMeasured
20Repositories, Linear teams, the GitLab project, Slack channels and teams. 1 withheld under the floors of 5 conversations and 3 people.
Pull requests opened by a coding agent, last 30 daysMeasured
585.4 working hours to the first review for 58 agent-opened pull requests, against 5.0 for 147 opened by people. 153 were opened by people.
Threads that outgrew asyncCount
14Conversations Quorum suggested turning into a short call, by their size and length — never by who posted most.
Recorded attendance cost, last 30 daysMeasured
£28,872305 of 399 meetings had attendance published by Google Meet.

What they changed

  1. Agree response times

    The leads set 7 response agreements — on Engineering, one of its teams, the Linear teams, the GitLab project and a repository — each a target in working hours and a share of conversations to meet it.

  2. Remind the owning team

    When a conversation passes most of its target unanswered, the owning team's manager gets a reminder naming the conversation, never a person: 12 reminders so far, 12 conversations close to their target right now.

  3. Report weekly

    A weekly digest reports each agreement: 1 met, 6 not met, 0 withheld under the floors.

  4. Read agents beside people

    Agent-opened and person-opened pull requests are compared on the same places and working hours, so a slow review queue shows as a place to look at, not a person to blame.

The result

Quillwren now has its response times in writing, agreed per place, with reminders that point at the waiting conversation and a weekly digest that says plainly which agreements hold. The first digests are candid: 1 of 7 reporting agreements met. The next step is the team's — a review rota, an owner per repository — and the following weeks' digests are where it will show. No change has been confirmed yet, so Quorum reports no before and after.

Where this lives in Quorum

  • Overview

    Led with thread cost and response time, because the workspace read as async-first.

  • Conversations

    Response time per repository, channel, Linear team and GitLab project, and the agent comparison.

  • Working agreements

    Targets per team and per place, with the weekly digest and reminders.

  • My week

    Each person's own hours and focus, for them alone.

How these figures were made

The workspace behind Quillwren was generated by Quorum's archetype harness — the synthetic companies the product is tested against — shaped to match the company's size, departments and tools, and seeded on Oct 11, 2026. Every figure on this page was then computed by Quorum's own analysis on that data and is read from a file this page cannot change. What the people in the example did — who answered, which changes they made, the settings their administrators chose — is part of the example and is stated where it matters.

An example is not a customer result, a benchmark or a forecast for your organization.

See these figures for your own calendars

Request a calendar auditRead how Quorum measures

Quillwren is an example company. Its name and story are invented to show how Quorum works; every figure on this page was computed by Quorum on a workspace modelled on it, seeded on Oct 11, 2026.