FOIA / Day Desk
Title
FOIA Day Desk

Every morning, each person gets one prioritized list of the FOIA work they need to handle — delivered into the Discord chat they already use.

August 2026 Status — controlled pilot

01 · The problem

The team can be busy while important work still sits.

Production work
Agency replies
Records received
Follow-ups
Invoices & fees
New requests
Calls
Research
Backlog

Nine kinds of work, all arriving at once, all looking equally urgent.

And today, every person has to go and find their own share of it before they can start — then decide by themselves which parts actually matter today.

02 · The change

Stop making the team search for their work.

Today Before a single case moves
Open Notion Filter down to my cases Check which response dates have passed Check whether any agencies replied Check what Production needs Check the other systems Decide, alone, what to work on

The employee searches for work.

With Day Desk Same person, same morning
Start Time Doctor Day Desk checks the whole pipeline The day’s list is already in their Discord chat Read it, change anything, approve Work straight down the list

The work finds the employee.

Notion Still the source of truth Every case, status and task stays exactly where it is. Nothing moves out of Notion.
Discord The daily work desk Where the team already reads and replies all day. Now it is also where the day’s work shows up.
InsanityBot The layer in between Reads the operation, works out today’s list, posts it, and writes the agreed plan back to Notion.

03 · A normal morning

From clocking in to an agreed plan, in about a minute.

STEP 1 Workday starts Time Doctor says who is working, when they started, and how many hours they have.
STEP 2 Day Desk reads the operation Notion, Autobot, email, portals, calls, payments, records, production.
STEP 3 It works out what really needs doing Due? Already handled? Did the agency reply? Whose case is it? What is the actual next step?
STEP 4 — THE POINT The plan appears in their Discord chat One message, in the chat they already have open. Nothing to go and look for.
STEP 5 They review and approve Approve, reorder, add, remove, mark something already done, or flag a blocker.
STEP 6 Both sides know the day Employee and manager are looking at the same agreed list. It is written back into Notion.

Steps 1 to 6 run today, for the two people in the pilot.

Later — not active Step 7 is checking whether the agreed work actually moved, and telling the manager when it did not. That layer is designed but deliberately switched off until the planning is proven.

04 · What it takes

Complexity happens behind the scenes. The employee gets clarity.

Reads Notion Autobot Email Portals Calls Payments Records Production Time Doctor
Asks about every possible task Is this actually due? Did someone already handle it? Did the agency already reply? Have the records already arrived? Whose case is this? What is the real next step? Is it already on someone else’s list? Is it blocked, or just quiet? Does it fit the hours they have?
# naud-chat Discord
IB InsanityBot BOT 09:12

Today’s FOIA plan — Tuesday 25 August

8-hour workday · 8 tasks · proposed

Critical 1Production re-request — Case A 2Answer agency clarification — Case B Must handle 3Review records received — Case C 4Follow up, 41 days overdue — Case D 5Outstanding invoice — Case E Normal 6Follow up — Case F 7Submit request — Case G 8Review — Case H

Reply approve, or tell me what to change.

N move Case C above Case B — Case F is already done

Hundreds of open cases and nine live systems on the left. One message with eight lines on the right.

05 · Why it is not a task bot

An old status is not the same as the next action.

Case  — what each system says

Notion Follow-up overdue — 22 days
Email The agency replied four days ago
Records Some of the files already arrived
A bot that just copies statuses Chase them again
What Day Desk puts on the list Read the records and reply to the agency

The same check runs on every follow-up that looks overdue

The date has passed, and nothing shows a follow-up was ever sent. Send the follow-up
The follow-up did go out — the date in Notion just never got moved. Update the expected response date
The date was moved, but nothing anywhere shows a follow-up happened. Check that it was really sent

06 · Priority

Important work goes to the top. Easy work does not.

1stProduction blockers and hard deadlines Never pushed off the list for lack of hours
2ndAnything waiting on us Agency replies, records, invoices, clarifications
3rdFollow-ups that are due or overdue Only where no follow-up actually went out
4thHigh-priority cases and yesterday’s leftovers Important work that did not get finished
5thNormal backlog Fills the rest of the day, oldest first

It works out what kind of work something is before it decides what order to do it in.

Weighed inside every band:

Deadline How long it has sat Who owns it Case state Production impact Hours available

07 · Shared visibility

One list. Both sides. Agreed before the day starts.

What the bot posts Proposed plan
→  the person reads it, changes what is wrong, replies approve
What both sides then have Today’s agreed plan

The employee

Live in pilot

Knows exactly what they are expected to handle today, and had a say in it.

  • Approve it as posted
  • Reorder it around their own day
  • Add something the system missed
  • Say an item is already done
  • Flag a blocker instead of going quiet

Every change makes a new version of the plan, so what was agreed is never in doubt.

The manager

Live in pilot

Opens that person’s work chat and sees the same list, without asking anyone.

  • What this person is handling today
  • What was moved, added or dropped, and why
  • Which blockers they raised
  • Where the day was already full
  • No chasing across four systems to find out

This is about a shared expectation, not watching people. Time Doctor only says when a workday began.

Later — not active Once planning is proven, the same agreed list becomes the baseline for what actually moved — and only the exceptions come to the manager. That comparison is not switched on yet.

08 · Pipeline impact

Cases lose weeks between the stages, not inside them.

Research
Request
Follow-up
Agency replies
Records
Production

Where the waiting happens — and what lands on someone’s list instead

Follow-ups that slip Every due and idle case is surfaced — but only where no follow-up actually went out.
Replies nobody answers When an agency has already written back, answering them beats chasing anyone else.
Records that sit unopened Files that arrived but still need someone to act become a named task, not a notification.
Re-requests stuck behind backlog Work that is blocking Production goes to the top of the list the day it appears.

These are the effects we expect, not ones we have measured. The pilot is too young for real numbers, and we would rather show you true ones later than estimates now.

09 · Where we are

Built in layers. We only add the next one when the last one holds.

Reading the operation — who works when, and what every system saysBuilt
Deciding what matters — the next action and the priority orderBuilt
The Discord daily plan — posting, replies, approval, writeback to NotionBuilt
Controlled live pilot — two peopleWe are here In progress
Checking the work actually movedNot started
Midday and end-of-day checks, and exceptions for the managerNot started
Personal targets and performance reportingNot started
The whole team, researchers includedNot started

We are proving it gives people the right work before we use it to judge anyone’s performance.

Two people on purpose. Researchers are waiting on one remaining Notion connection, and the production detector is running quietly in the background where we can watch it without it acting. Automation is easy if we accept wrong answers — the work is making it right often enough to run the operation on.

10 · End state

Every day, we should be able to answer five questions without asking anyone.

What does each person need to handle today? Answered now
Live in the pilot — one agreed list per person, in their own Discord chat.
What moved?Next layer
Confirmed from what the systems show, not from what someone remembers.
What didn’t?Next layer
Measured against the plan the person agreed to that morning.
Why not?Next layer
Blockers recorded as blockers, instead of looking like someone did nothing.
What needs you?Next layer
Only the exceptions reach the manager. Everything else just runs.

A faster, more predictable FOIA pipeline — because the right cases get worked on the right day.