Quick answer (TL;DR): In most jobs, work arrives from everywhere — a chat message here, an email there, a comment on a ticket, a line in a shared tracker, a hallway aside someone forgets they said. Things get dropped because no single surface holds the whole picture. The reconcile is a periodic sweep across every channel where work lands, cross-referenced against your own tracker, to surface what’s been dropped. It’s tedious, mechanical cross-referencing — exactly what humans are bad at and AI is good at — which makes it one of the highest-leverage things an AI assistant can do for you.
This is the most immediately useful habit in the entire AI working method series, because almost everyone has this problem and almost no one has a system for it.
The scattered-input problem
Think about where work actually reaches you in a given week:
- A direct message: “hey, can you look at this when you get a chance?”
- An email thread you got CC’d on with a buried ask.
- A comment on a ticket asking whether something should be cancelled.
- A new row in a shared tracker assigned to you that nobody mentioned.
- A request in a meeting that was never written down anywhere.
- A “we should do X at some point” that someone said three weeks ago and forgot.
No single one of these surfaces holds the complete picture of what’s on your plate. So things slip — not through negligence, but through fragmentation. The request was real; it just never landed anywhere you’d reliably see it again. And in many organizations, the people sending these asks fire them across different channels and then forget they ever sent them, which means you are the only one tracking the full set — except you’re not, because no one can hold that many scattered threads in their head.
This is where an AI assistant earns its keep in a way that feels almost unfair.
What the reconcile is
The reconcile is a periodic, systematic sweep across every channel where work or requests can arrive, cross-referenced against your own source of truth, to answer one question: what’s fallen through the cracks?
It’s the kind of task that’s miserable for a human — switching between five tools, scanning for open items, holding them all in working memory, comparing against your tracker — and trivial for an AI assistant that can read across those surfaces and do the cross-referencing mechanically. The machine doesn’t get bored, doesn’t lose the thread, and doesn’t forget the message from three weeks ago.
How to run a reconcile
Here’s the process, step by step.
1. Enumerate your input surfaces
List every place work or requests can reach you:
- Messaging / chat tools
- Ticketing or project systems
- Shared trackers and spreadsheets
- Shared docs and comments
You can only sweep what you’ve named, so make the list complete.
2. Sweep each surface for open items
Have your assistant go through each surface looking for:
- Open requests addressed to you.
- Recent messages that contain an implicit ask.
- Comments or questions awaiting your response.
- Items assigned to you that you didn’t initiate.
3. Cross-reference against your tracker
This is the heart of it. Compare what the sweep found against your own source of truth:
- What’s here that isn’t tracked? (Dropped or never-captured work.)
- What’s tracked that’s actually already done? (Stale items to close.)
- What’s tracked as active but actually waiting on someone else? (Blocked items to reclassify.)
4. Produce one consolidated list, with the new stuff flagged
The output is a single reconciled view of everything open, with the genuinely new or dropped items flagged distinctly from the things you already knew about. This is the moment the reconcile pays off — the dropped threads become visible.
5. Decide, then update
For each surfaced item, make a call: action it, defer it, delegate it, or close it. Then update your tracker and your second brain so the reconcile compounds. Next time, you’re reconciling against a more accurate picture, and the sweep gets faster and cleaner.
What a reconcile typically surfaces
In practice, a good reconcile tends to turn up a predictable mix:
- A genuinely new request that came in through a side channel and was never captured anywhere.
- A decision someone is waiting on — a “can this be cancelled?” comment sitting unanswered on a ticket.
- A relabeling or correction — something tracked under the wrong name or number.
- Confirmation that several things you were vaguely worried about are actually fine — which is its own kind of relief.
That last one matters more than people expect. A big part of the value isn’t just catching dropped work; it’s the confidence of knowing you’re on top of everything, rather than carrying a low-grade anxiety that you’ve forgotten something.
A diagnostic benefit: spotting the pattern
Run the reconcile a few times and a meta-pattern often emerges: who keeps dropping asks across channels, and which channels are the leakiest. That’s useful management information. If the same person routinely fires requests across three tools and forgets them, that’s a process conversation worth having — and now you have the evidence, because the reconcile documented it. The reconcile doesn’t just catch dropped work; it reveals why work gets dropped, which lets you fix the upstream cause.
When to run it
Run a reconcile whenever the inputs have outpaced your certainty — when you have that nagging feeling you might be forgetting something. In practice, a weekly cadence works well for most people, plus an ad-hoc reconcile after any particularly chaotic stretch (a big launch, a return from time off, a reorg). The more fragmented your inputs, the more often it’s worth doing.
The reconcile turns “I think I’m on top of everything” into “I know I am.” That shift — from anxious guessing to grounded certainty — is worth the few minutes it takes.
Frequently asked questions
What is a “reconcile” in a work context? It’s a periodic sweep across every channel where work arrives (chat, email, tickets, trackers, docs), cross-referenced against your own tracker, to surface requests and tasks that have been dropped or never captured.
How do I stop dropping tasks when requests come from everywhere? Run a regular reconcile: list all your input surfaces, sweep each for open items, cross-reference against your source of truth to find what’s untracked, then consolidate, decide, and update. An AI assistant makes this fast because it’s mechanical cross-referencing.
How often should I run a reconcile? Weekly works for most people, plus an extra one after chaotic periods like a launch, a reorg, or returning from time off. Run one anytime you feel uncertain whether you’ve captured everything.
Why is AI good at this specifically? The reconcile is tedious, repetitive cross-referencing across many tools — exactly the kind of task humans do poorly (we get bored, lose the thread, forget old messages) and AI does reliably and quickly.
This is Part 6 of a series on building a working method with AI. Next: how to let AI produce real deliverables — drafts, documents, decks — without letting it embarrass you.