I have tried almost every inbox app that promised to change my life.

Notion Mail. HeyHelp. Apple Mail. Reclaim. Fyxer. Zapier hacks that required more maintenance than my actual client work. Unroll.me, which I used for years until it quietly stopped working, and nobody told me.

Every single one of them either broke, cost more than it saved, automated things I didn’t ask for, or just didn’t play nice across all my accounts. Some of them sent disaster drafts out that I still think about when I am playing my “greatest hits of failures” highlight reel at night instead of sleeping. And when you’re a fractional with 9 to 10 active inboxes at any given time, “doesn’t play nice across all accounts” is really bad. REALLY BAD.

Embarrassing confession. So here’s what I actually use: Microsoft Outlook for Mac. Yes, I know. I have talked smack about SharePoint and Microsoft products in many a group chat, and the Teams ringtone still gives me PTSD.

But I’ll swallow my pride because it’s the only client that handles a chaotic mix of Gmail, Microsoft, POP, and hosted accounts without complaining. If I had clients NOT on Microsoft, I might try something else, but at this point, client compatibility and familiarity win. It’s not glamorous. It works.

Anyway, the tool is almost beside the point. The system is what matters.

The problem with most inbox advice

Most inbox advice assumes you have one email address and a normal human amount of emails. I get somewhere between 1,200 and 1,500 emails a day in my consulting account alone. Generic advice about unsubscribing and using just folders doesn’t really cover that territory.

What actually works is treating your inbox like a triage unit, not a to-do list. Because it took me years to get it to sink in, I’ll say it again. Your inbox isn’t a to-do list; it is a triage unit.
Those are two very different operating modes, and confusing them is why most people feel like their inbox is the world’s most impossible escape room.

The 3-block triage system

As I mentioned in my post about my ADHD toolbox, I check my email about three times a day. That’s it. The rest of the time, it’s closed or blocked.

Morning (15-20 minutes): I start with LittleBird (use code J6FD24JY for 2 months free), which runs a morning briefing across all my inboxes and flags anything urgent or on fire before I even open Outlook. Then I do one fast pass through each account: sorting, flagging, forwarding, archiving, and deleting. The goal is to get as close to inbox zero as possible before I touch anything else. I am not responding in this block. I am just sorting. If I have time at the end of the block, I will respond to the Urgent tasks.

Afternoon (15 to 45 minutes): This is my main response block. Anything flagged urgent gets actioned first, then I work through This Week. For drafting, I use a combination of Gemini for quick Gmail replies and Claude or LittleBird for anything that needs more thought. Willow Voice is my secret weapon here: I’ll dictate what needs a response and what I want to say, and it creates the drafts. Talking is faster than typing. I will do it all in one prompt, so I don’t have to wait for each individual draft to populate. I continue sorting/deleting, moving things along while it works. Then I deploy.

End of day (30 minutes): Final sweep. Anything that didn’t get done either goes back into This Week or becomes a task in Motion for tomorrow. The inbox closes when the timer goes off. I will admit that this is the session most likely to run over its time. I hate leaving things undone. Progress, not perfection.

“Wait a minute, what about the emails you do have to think more about and reply to?”

Ah, glad you asked. Heavy emails that require real thinking don’t get answered in triage. They become Motion tasks with a specific block to respond. That way, I’m not losing 40 minutes to one complicated thread when I should be triaging.

The folder architecture

Here’s what the sidebar actually looks like for my main consulting account:

• 🔥 Urgent (auto-filtered by keyword, contact list, and billing triggers)

• 🟧This Week (mostly manual sort during morning triage)

• 🐿️To Reply (Substack replies auto-route here -- I ask readers to reply so I know they’re human), then I respond

• 💼 New Leads / Opps to Look At (referrals, alerts, prospect emails - feeds my sales dev time block)

• 🤝 Partner To Do (partner communications and opportunities - feeds my partner power hour time block)

• 📝 Research Requests / Surveys (I have a task for this folder, 10 minutes 3x a week in Motion)

• 💲Invoices To Process

Newsletters, blogs, and reading material don’t live in my inbox at all. I forward them straight to Readwise Reader with a keyboard shortcut or filters and delete the original. When I have reading time, I go there. Not before. This is what feeds my personal development block, which I mentioned in my workflow post.

The AI prompt that sets up your filters

If you’re staring at your inbox right now, wondering where to even start, here’s the move I used when I first built this out:

Open Claude, Gemini, or whatever you’re using and run this prompt. Claude Cowork is great for this if you want to point it directly to your main inbox:

“I want to build a smarter inbox filtering system in [Gmail / Microsoft Outlook]. Before I set anything up, I want you to help me identify my patterns.

First, here’s some context about how I work: [briefly describe your role, how many inboxes you manage, and who your most important senders are -- clients, prospects, vendors, etc.]

Based on that context, please:

1. Identify the most common categories of email I likely receive (urgent, FYI, promotional, platform notifications, newsletters, etc.)

2. Suggest which folders would be most beneficial for me to use to triage my inbox.

3. Based on those folders, suggest specific filter rules I can set up for each category, including keywords, sender patterns, and what action to take (archive, label, flag, delete on arrival)

3. Give me the exact language I can use to set up those rules in [Gmail / Outlook]

4. Tell me which filters to set up first, in order of impact.”

Run that once, implement the top three suggestions, and then run a second pass a week later once you’ve seen what’s still slipping through. The system gets smarter every time you tighten it.

One thing I’d add: if you’re not sure what your urgent email patterns actually look like, ask it to analyze that too. The prompt version I used was something like: “What types of emails do consultants and fractional executives typically need to respond to within 24 hours? Help me build an urgent filter based on those patterns.” Then I layered in my own specifics on top of what it gave me.

It’ll surface patterns you already know but haven’t acted on yet. That’s what started this whole system for me.

The ADHD tax on inbox management

One thing nobody talks about: for an ADHD brain, the inbox is a dopamine slot machine. Every refresh is a potential hit. Every unread number is a nagging open loop.

When you work across multiple clients, the inbox anxiety compounds. Missing an email from one client feels like negligence. Missing one from a prospect feels like lost revenue. Missing one from a vendor feels like a project derailment. My brain reads every unread email as a potential crisis I haven’t dealt with yet. What if that’s the one? What if I miss something important? What if someone thinks I’m ignoring them? I spent years with my inbox open all day, not because it helped, but because closing it felt worse than the chaos. It didn’t help. It just meant I was always half in my inbox and never fully anywhere else.

In truth, if something were a real emergency, you probably wouldn’t get an email about it.

As for turning off the slot machine, the pomodoro timer is non-negotiable for me. While it’s running, the only thing open is my inbox. I have it set up to block Chrome (sometimes to my annoyance). No link-clicking. No rabbit holes. No “I’ll just quickly check this one thing.” When the timer goes off, I close it. Even if things are left undone (and yes, that genuinely hurts, especially the first few times), I trust the system to come back to it.

Steal this:

• Use the AI prompt above to identify your folder and filter categories before you build anything manually.

• Set up your urgent filter first. Add keywords, contact lists, and billing-related terms.

• Set up other filters and forward rules (or set the shortcut keys).

  • Send all newsletters to Readwise (or a dedicated reading folder) on arrival. Delete the original.

• Run three timed blocks. Close the inbox between them. (You can do it, I know it is scary.)

• If an email needs real thinking, make it a task. Don’t answer it during triage.

• Test it for a month and then repeat the analysis to improve it.

If you try this, hit reply and tell me what broke. I’ve probably already been there.

⭐ Thanks for being here. Now go do the thing you’ve been putting off.

Katherine / Chief Chaos Officer

The recurring topic of conversation the last few weeks has been, “I want to switch over to Claude, but how do I not lose my history?”

Additionally, with some platforms throttling usage earlier and earlier in the day to prompt you to upgrade, it can seem like you have started paying more for 3-4 AI tools than you would have for a personal assistant by now. Consolidation of tools may be the driving force, as it was for me. Claude has emerged as the leading partner in my AI work, thanks to their skills and coworking capabilities, and for how they integrate into my existing workflow rather than forcing me to create new ones.

If you have also reached the conclusion that you want Claude to run your AI life, then this article is for you.

Understandably, you have spent months teaching ChatGPT (or other platforms) who you are. How you speak, what frameworks you like, and how much you hate em-dashes. Your time has gone into loading up its memory and having conversations about dinner planning and strategic planning. The sunk cost feels very real.

This guide will help you move the important context from ChatGPT to Claude. It will take you about 30 minutes.

(Note: Feel free to leave ChatGPT as a backup for when your usage limit gets in the way. Just make sure you bring over important conversations later.)

Let’s do it!

Step 1: Extract what ChatGPT knows about you. There are three parts here to complete this.

Part A: Your Personality
Go to ChatGPT Settings > Personalization and manually copy your Custom Instructions into a separate document. These are [hopefully] rules you wrote telling ChatGPT how to personalize - tone preferences, output formats, things it should always or never do. (Mine says no em-dashes, but alas, it was always ignored) They live separately from stored memories, and the export prompts below will not capture them.

Part B: Memories and Projects
Open a new chat in ChatGPT and paste this prompt (straight from Anthropic’s migration guide):

Export all of my stored memories and any context you’ve learned about me from past conversations. Preserve my words verbatim where possible, especially for instructions and preferences.

## Categories (output in this order):

1. **Instructions**: Rules I’ve explicitly asked you to follow going forward — tone, format, style, “always do X”, “never do Y”, and corrections to your behavior. Only include rules from stored memories, not from conversations.

2. **Identity**: Name, age, location, education, family, relationships, languages, and personal interests.

3. **Career**: Current and past roles, companies, and general skill areas.

4. **Projects**: Projects I meaningfully built or committed to. Ideally ONE entry per project. Include what it does, current status, and any key decisions. Use the project name or a short descriptor as the first words of the entry.

5. **Preferences**: Opinions, tastes, and working-style preferences that apply broadly.

## Format:

Use section headers for each category. Within each category, list one entry per line, sorted by oldest date first. Format each line as:

[YYYY-MM-DD] - Entry content here.

If no date is known, use [unknown] instead.

## Output:

- Wrap the entire export in a single code block for easy copying.

- After the code block, state whether this is the complete set or if more remain.

Copy the whole output it returns.

Part C: Everything

Your full chat history isn’t included in the above prompts. Claude’s import tool transfers your memory and preferences, not your full chat history. If you need to reference old conversations, export your ChatGPT data and upload specific files to a Claude Project. That’s the workaround for conversations you actually need to keep access to.

Go to: Settings > Data Controls > Export Data

It produces a zip file containing conversations.json, which is every message you have ever sent. It will take anywhere from a few hours to a few days, depending on how much you talk to your robots. It is not necessary for migration.

Step 2: Edit before you import. (This is the step people skip.)

ChatGPT sometimes infers things incorrectly from context, so check everything. Before you paste into Claude, open Part A and Part B docs and do a quick prune. Kill anything outdated, projects you’ve closed, tools you stopped using six months ago. A well-edited profile is more useful than a bloated one, so don’t be afraid to cut aggressively. Think of it less like moving boxes and more like finally unpacking only the stuff you actually want in the new place.

Think of it like any database or CRM migration: junk in, junk out.

Step 3: Import into Claude

From Part A: Add your custom instructions/personalization settings from Part A to Claude’s personal preferences (Settings > General > Personal Preferences)

From Part B: Go to Claude.ai > Settings > Capabilities (or Memory) and paste your cleaned-up context directly into the memory field. Claude updates its memory, and you’re good to go. From there, Claude learns your preferences across conversations and keeps project context separate so nothing bleeds together.

Bonus move for Cowork users/Part C: Upload your conversations file directly to Cowork and run this prompt: “I uploaded my exported ChatGPT conversation history. Analyze it like a strategic chief of staff and operating partner. Your job is to extract the most useful, durable context about me from my full conversation history, not just summarize chats.” Then copy that output back into Claude’s memory to complete the loop if it’s relevant.

You can upload specific pieces or conversations over into new Claude Projects if you feel complete context is necessary.

I kept a copy and didn’t bring anything over from the Part C full export. I am just including it here for completeness because it makes me feel better to have it.

REPEATING THIS COMMON MISTAKE: Don’t dump the raw ChatGPT export directly into Claude without editing it first. You’ll import a bunch of stale context and then wonder why Claude keeps referencing a project you finished two years ago.

Key note:

One more thing worth knowing: memory import is still labeled experimental as of March 2026, and Claude focuses primarily on work-relevant context. Personal details that aren’t connected to how you use Claude professionally may not stick.

Adjust accordingly if you really need to remember what the meal plan was for a week when you only had garlic and apples in your fridge.

⭐ Thanks for being here. Now go do the thing you've been putting off.

-Katherine, Chief Chaos Officer

THE TIP

Before you build that AI workflow, run it through the Good Enough Gate. Three questions. Two minutes. Zero wasted afternoons.

Why it matters: AI is genuinely impressive. It can do a lot. Which is exactly the problem. “Can I build this?” is a very different question from “Should I?” and the gap between those two questions is where hours go to die. (Confession: I once spent four hours building out a full content repurposing system in ChatGPT that I was never, ever going to use. It was a great experiment. But it also turned into a full-on squirrel moment—definitely not the best use of a Tuesday when I had other things to be working on.)

The issue is that AI reduces the friction of starting something so much that it eliminates the natural pause where you used to ask yourself whether it was worth doing.

This is the gate.

Subscribe now

The Good Enough Gate (3 questions, run them in order):

1. Will I actually use this more than once? If the honest answer is “probably not,” stop. You’re not building a workflow. You’re doing a fun experiment. That’s fine, but name it correctly and timebox it to 20 minutes max.

2. Does this replace something I currently do manually, or am I inventing a new task? New tasks dressed up as efficiency wins are one of AI’s sneakiest traps. If you didn’t have a process before, building an AI version of a process you didn’t need isn’t progress. It’s procrastination with better branding.

3. If this works exactly as planned, does it save me meaningful time OR meaningfully improve an outcome? “Meaningful” means at least 30 minutes a week, or noticeably better output on something that actually matters. If the answer is “well, kind of,” that’s a no.

How to use it:

1. Next time you catch yourself thinking “ooh, I could build a workflow for this,” or “wow, this framework will change my life,” pause before you open a new chat.

2. Run the concept through all three questions. Takes 90 seconds.

3. Two or more “no” answers? Put it in a “someday maybe” list and walk away. One clear “yes” across all three? Build it, but set a 45-minute timer before you start.

BONUS HACK:

If you use a custom GPT (or a saved Claude project), add this to the system instructions so it asks you the gate questions automatically before helping you build anything new:

“You are my AI workflow filter. Before helping me build any new workflow, automation, prompt system, or tool integration, you are required to stop and run the Good Enough Gate.

Ask me these three questions one at a time and wait for my answers before proceeding:

1. Will you actually use this more than once, or is this a fun experiment you’re calling a workflow?

2. Does this replace something you currently do manually, or are you inventing a new task that didn’t exist before?

3. If this works exactly as planned, does it save you at least 30 minutes a week or meaningfully improve an outcome that actually matters to your business?

Based on my answers and stored knowledge, you are allowed to push back. If I’m rationalizing, call it out. If I’ve asked you to build something similar before and abandoned it, remind me. If I answer yes to all three, help me build it efficiently with a defined scope and a hard stop. If I answer no to two or more, tell me to park it and ask: what’s the real problem you’re trying to solve today?”

Paste that in. Done.

Now your AI is the thing stopping you from going down an AI rabbit hole. The irony is not lost on me.

Common mistakes:

Skipping question two. Inventing AI tasks feels productive. It is not the same as automating real ones.

• Before: Four hours deep into a beautifully structured content system inside ChatGPT that existed purely because the technology made it possible.

• After: A two-minute gut check that sent three “clever ideas” directly to the parking lot, and freed up actual time to do the work.

If you try this and it saves you from a rabbit hole, join the conversation/chat and tell me what you almost built.

If you’re anything like me, your default mode is: save it, screenshot it, send it to yourself, never look at it again. So much great STUFF and…..you never use it. And the gap between “that’s interesting” and “I use that” is where most good ideas go to collect dust like craft supplies you swore you’d get to. (I used to work at JoAnn Fabrics way back in the day, so trust me, I know what hoarding potential is all about.)

The Gap Nobody Talks About

Here’s the uncomfortable truth: more content won’t fix it. More bookmarks won’t fix it. More tips won’t fix it. A better note-taking app won’t fix it.

The only thing that closes the gap between “saved” and “used” is a small, consistent action that forces you to actually test what you’re collecting.

And if you’ve got any flavor of ADHD brain (diagnosed or just suspicious), this pattern probably hits harder. The dopamine hit comes from the save, not the use. Finding the thing feels productive. So we keep finding more things…..

…until your saved folder looks like a craft drawer full of supplies for projects you’re never going to start.

Here’s your fix: The 15-Minute Friday Install.

Step 1: During the week, capture as you go (keep it dumb simple).

When you see a tip, prompt, or framework worth testing, clip it to ONE LOCATION. Use the Notion web clipper, Apple Notes, a simple task list, whatever you already have. Don’t build a system. Just capture. You don’t need a fancy template here.

If you REALLY want a template, search “prompt organizer” or “AI organizer” in Notion’s template library. There are good free options. But honestly? A basic list works fine until you know what’s worth keeping. I use a Notion board that is just a basic task manager.

Step 2: Run the install.

1. Block 15 minutes on your calendar (not “when I have time”). Mine is Friday morning.

2. Pick ONE thing from your captures. One. (Don’t know what to pick? Drop the list into AI and ask which would be most beneficial to you right now.)

3. Test it. Use it immediately in one real place:

   • one client workflow

   • one email

   • one meeting agenda

   • one prompt you actually run

   • one template you actually fill in

4. If it works, move it to a “keepers” folder or copy it to wherever you store things you actually reference.

5. If it doesn’t, delete it. No guilt. That’s the point.

Step 3: Let the losers go.

Most things you save won’t stick. That’s fine. You’re not building a library. You’re filtering for the handful of things that actually make your work easier. Don’t fall into the trap of keeping it for “someday.” If it didn’t stick the first time, you won’t remember it exists later.

Common mistake: Don’t wait until you have the “perfect system” to start testing. Capture first. Organize later. The system is not the point. Using what you find is the point.

Why This Matters More Than You Think

Every tip you save and never use is a small promise you break to yourself. Do that enough times and you stop trusting your own follow-through.

But flip it: every time you test something (even if it fails), you build evidence that you actually do things. You’re not just collecting ideas. You’re making decisions.

That shift matters more than any individual prompt or framework ever will.

What this newsletter will always aim to give you:

• A clear takeaway

• A small action

• A path from “interesting” to “implemented”

Your job is to pick one thing and try it. My job is to make that easy. So save the articles you want to test.

Try the 15-Minute Friday Install this week. Reply and tell me what you picked and whether it worked. (Flops count too. I want to know what’s landing and what’s not.) If it helps you, share it with a friend.

Bonus Acorn Tip: Hook up your folder to AI (like Claude CoWork) and let it tell you which tips are most likely to improve your life and which ones you are most likely to ignore. Delete the duds to narrow the list even more.