The Ultimate Beginner's Guide to Claude Cowork for Executive Assistants

What Is Claude Cowork?

Most EAs hear "AI tool" and think: chatbot. Ask it a question, get an answer. Helpful sometimes, forgettable mostly.

Claude Cowork is a different product.

It is Anthropic's agentic AI platform. "Agentic" means it can plan and execute multi-step tasks on its own, not just respond to questions. You describe what you need, it builds a plan, shows it to you for review, and then gets on with it. It connects to Gmail, Google Calendar, Slack, Notion, Google Drive, and many other tools. It can read your files, write documents, draft and organise emails, prepare briefings, and run recurring workflows automatically, at a time you set, without you being in the room.

For an EA, this is different in kind to any other AI tool you have tried. The reason most EAs feel disappointed by AI is not that the tools are bad. It is that they are using a general tool with no context about who they are, who their exec is, or how their office works. Every conversation starts from zero. You re-explain everything. The output is generic.

Claude Cowork changes that through three things: Projects, custom instructions, and connected apps. Together, they let you build a workspace configured around how you actually work, one that knows your exec, your tools, and your standards, and holds that context permanently.

This guide will walk you through how to set that up from scratch.

What you will need before you start:

• A Claude account (free account works for Projects; Pro at $20/month is the minimum for daily professional use and covers all Cowork features including scheduled tasks)
• 60 to 90 minutes for initial setup
• Your exec's bio, communication preferences, and any standard templates you already use
• The Whiteprints Claude Cowork Starter Kit (download link at the end of this guide)

BTW: If you're non-technical and need unlimited, on-demand AI support across any tool/software you already use or plan to use, join the Whiteprints club here:

whiteprints.vip/membership

The Plan You Need

Before anything else, get clear on which plan fits where you are now.

Free Plan

Claude's free tier runs on a capable model with daily usage limits. As of early 2026, free users now have access to Projects and Artifacts, which were previously paid features. If you are testing Claude for the first time or your workload is light and irregular, start here.

The limits will frustrate you if you are doing real daily work. When you hit the ceiling, output stops until the next reset window. For professional use, free is not a reliable foundation.

Pro: $20/month

This is the right starting point for most EAs. You get significantly higher usage limits, access to Claude's most capable models, full Projects functionality, and scheduled tasks in Cowork. If you are doing inbox triage, meeting prep, drafting, and research on a daily basis, and you want recurring workflows running automatically, Pro covers all of it.

Teams: $25 per user per month

Designed for more than one person. If you and your exec, or your EA team, want to share a workspace and collaborate on Projects, Teams gives you pooled usage and shared access. The right tier once Claude is embedded in how your office actually runs.

Max: $100 to $200 per month

This unlocks Cowork's full autonomous capabilities: computer use (Claude navigates your screen directly for apps without a connector) and the 1 million token context window. Max is the right tier for the heaviest workloads and the most complex autonomous tasks.

Where to start:

Start on Pro. It covers everything in this guide including scheduled tasks. Move to Max when workload regularly hits usage limits.

Setup - 30 Minutes

The first thing you need is the Whiteprints Starter Kit. Download it at whiteprints.vip/resources/claude-cowork-starter-kit before anything else. Everything in this section assumes you have it.

The Starter Kit is a folder structure that becomes Claude's operating system for your role. Inside it is a context folder with six files:

• about-me.md — your role, priorities, and how you like to work
• about-exec.md — your exec's decision style, calendar preferences, communication preferences, and red lines
• key-contacts.md — your stakeholder map, escalation tree, and delegation defaults
• working-preferences.md — your output formats, execution guardrails, and quality bar
• my-voice.md — your tone, phrase rules, and writing standards
• comms-strategy.md — channel strategy and message architecture

Fill these in. This is the most important step in this entire guide. Every output Claude produces for you, every draft, every briefing, every triage, draws from these files. A half-filled context folder produces half-useful output. A complete one means Claude knows your exec, your standards, and your workflow before you type a single word.

It takes about 20 to 30 minutes to fill in properly. Do it once, update it when things change, and you'll never need to explain yourself to Claude again.

Once your context files are filled, open Claude Desktop and install the Starter Kit as a plugin. Claude will read your context folder automatically and the skills become available as slash commands.

The first time you run /daily-brief and get a morning summary that already knows your exec's calendar priorities, your VIP senders, and your format preferences, is the moment this stops feeling like a chatbot.

Custom Instructions

The context folder is what Claude knows. Custom instructions are how Claude behaves. You need both.

The Starter Kit's working-preferences.md and CLAUDE.md handle most of the operating rules automatically. But if you are working inside Claude's web Projects (claude.ai rather than Claude Desktop), or building additional Projects beyond the Starter Kit, custom instructions are how you apply those rules.

Custom instructions are plain-English rules written inside a Project's Instructions panel. Claude applies them to every conversation in that Project.

What to include:

Role definition. Tell Claude what it is doing in this Project. "You are supporting [exec name]. Your job is to help the EA prepare communications, briefings, and materials that match their voice and working style. See about-exec.md for full context."

Communication style. Be specific. "Emails are direct and professional. Short paragraphs. No filler phrases. Always include a clear next step. Subject lines are under 10 words."

Format rules. Tell Claude how to structure output. "Use bullet points for lists of three or more. Briefings use three sections: Context, Key Points, Recommended Action. Action items always include owner and due date."

Guardrails. "Never send anything without explicit sign-off. Flag anything involving legal, financial, or HR matters before proceeding."

Reference pointers. Point Claude to your context files. "See about-exec.md for executive preferences. See key-contacts.md for stakeholder routing."

How long should instructions be?

300 to 500 words is the practical range. Longer than that and Claude drifts from the sections buried at the bottom. Keep the rules that matter most at the top.

If you are not sure what to write, ask Claude. Describe the Project, what good output looks like, and what generic output looks like. Ask it to draft the custom instructions. Refine from there.

The Five Workflows Every EA Should Set Up First

The Starter Kit comes with five skills built in. These are the ones to run first.

1. `/daily-brief` — run every morning

Produces a 10-minute executive morning brief: today's calendar, priority inbox threads, a RAG-ranked summary of what needs attention, top three priorities, and any risks or blockers in the next 48 hours. Everything is decision-first, no narrative filler.

Run this before you open your exec's calendar or inbox. It tells you exactly where to start.

2. `/inbox-triage` — run midday

Clears the inbox by categorising every unread thread as Reply Now, Delegate, Defer, or Archive. It separates calendar invite traffic automatically, surfaces same-day blockers, and drafts response starters for anything that needs a reply. Your exec's VIP senders and red-line contacts (from key-contacts.md) are factored in before anything is classified.

3. `/prep-meeting` — run before any high-stakes meeting

Produces a full pre-meeting dossier: who is in the room and why, critical context, explicit decision asks, a time-boxed agenda, and executive-ready talking points. It draws from your context files so the briefing already reflects your exec's communication preferences and known stakeholder dynamics.

Give it the meeting name, date, attendees, and objective. It does the rest.

4. `/followups` — run end of day

Collects every unresolved thread from email, chat, and meeting notes. Classifies by urgency, drafts reply starters for anything blocking approvals or next-day priorities, and produces an action tracker with owners and due dates. Nothing slips.

5. `/weekly-exec-report` — run every Friday

Compiles the week into a board-ready status packet: wins, misses, metrics snapshot, risks needing decisions, and next-week priorities. One-page executive summary up front, detailed workstream appendix behind it. Schedule it to run automatically at 4pm on Fridays and it is ready before the week closes.

The Mistakes That Slow Beginners Down

These patterns come up in nearly every EA's first month with Claude. Knowing them before you hit them saves time.

Starting every conversation from scratch.

If you are not using Projects, you are re-explaining your context every single session. Claude has no memory across conversations. Set up your Projects first, before anything else.

Putting everything in one long chat.

As a conversation grows long, Claude's quality degrades. This is called context rot. Unrelated tasks compound the problem. Start a fresh conversation for each task inside the right Project. The Project keeps your context. The conversation stays clean.

Vague prompts.

"Write me an email about the meeting" will produce something unusable. "Draft a 3-sentence email to Sarah at Apex confirming Thursday's 2pm call, professional but concise, from my exec's voice, see exec-bio.pdf" will produce something you can send with one edit.

The output quality is a direct reflection of the input specificity.

Correcting Claude in the same conversation, twice.

If Claude misses what you need and you correct it once, it usually course-corrects. If you correct it twice on the same thing, the conversation context gets polluted with failed attempts and the output gets worse, not better. Start fresh with a better initial prompt.

Expecting Claude to know things you have not told it.

Claude knows nothing about your exec, your company, your preferences, or your standards unless you tell it through custom instructions or uploaded files. When output feels generic, the problem is almost always missing context, not Claude's capability.

Treating it like a search engine.

Claude is a reasoning and execution tool, not a lookup database. Give it context, a role, a task, and a format. Do not just ask questions.

Other Cowork Features

Scheduled Tasks

You describe a recurring workflow, set a time and frequency, and Claude runs it without you being present. It pulls from your connected apps, processes the information, produces the output, and delivers it wherever you need it.

• /daily-brief at 7am every weekday: your morning summary is ready before you open your laptop
• /followups at 5:30pm: every open loop from the day is captured before you close out
• /weekly-exec-report at 4pm every Friday: the week is compiled and ready for exec review

You set these up once. After that, they run.

Dispatch

Dispatch lets you assign tasks to Claude from your phone and come back to finished work on your desktop. You pair your phone with Claude Desktop through the Dispatch tab. From anywhere, you send a task — Claude executes it on your computer using your files, connectors, and skills. When it is done, you get a push notification.

For an EA, this changes the equation. You are away from your desk. Something comes up. You open Claude on your phone, type the task, and your computer handles it. You come back to the output.

Dispatch is in research preview for Pro and Max subscribers.

App Connectors

Cowork connects natively to Gmail, Google Calendar, Slack, Notion, Google Drive, Microsoft 365, Figma, and 30+ other tools. When connected, Claude reads, summarises, drafts, and organises within those tools without you pasting anything in. This is what makes scheduled tasks and Dispatch genuinely hands-off.

Computer Use

For apps without a native connector, Cowork operates your computer directly: clicking, typing, navigating screens. You describe the task, Claude executes it on your desktop. This handles every edge case where no connector exists. Currently macOS only.

Sub-Agent Coordination

For tasks involving multiple files or workstreams at once, Cowork splits the work across parallel sub-agents running simultaneously. What would take 30 minutes sequentially can take 4 minutes in parallel. Useful when you need Claude to process a full week of meeting notes, multiple inboxes, or cross-reference several documents at once.

Mid-Task Steering

Cowork shows its plan and reasoning as it works, not just at the end. You can jump in at any point to redirect, correct, or add context. Long tasks do not require you to stand back and hope. You stay in control throughout.

The five skills in the Starter Kit become a different product entirely when they run on a schedule and can be triggered remotely from your phone. You stop managing the workflows. They run. That is the shift from using Claude Cowork to running on it.

Your Next Step — The Starter Kit and EA Ops

You have the foundation. The next step is setup.

Download the Whiteprints Starter Kit.

The Starter Kit is your Claude Cowork operating system. It installs as a plugin and gives you everything in this guide, ready to use from day one:

• context/ folder with six pre-structured files to fill in: your profile, your exec's profile, key contacts, working preferences, your voice, and comms strategy
• memory/ folder: glossary, decision log, and lessons learned — persistent context that builds over time
• projects/ folder: pre-built project templates for active work
• Five built-in skills: /daily-brief, /inbox-triage, /prep-meeting, /followups, /weekly-exec-report
• A CLAUDE.md master context file Claude reads automatically before every session
• A comms drafter agent for executive communications

Fill in the context files. Install the plugin. Your workspace is ready the same day.

Download at whiteprints.vip/resources/claude-cowork-starter-kit.

The next level: EA Ops

The Starter Kit covers the daily workflow. EA Ops is the full operating system.

EA Ops is Whiteprints' expanding library of skills and agents built specifically for EA work, configured around your exec and your tools. It includes everything in the Starter Kit plus:

• /calendar-optimizer: restructures your exec's week for focus time, decision flow, and reduced context switching
• /stakeholder-brief: one-page intelligence brief on any person before your exec walks into a room with them
• /travel-ops: full executive travel brief with itinerary health check, logistics checklist, daily run-of-show, and contingency plan
• /extract-tasks: pulls every action item, implied follow-up, and commitment from any email or meeting note
• /calendar-inbox-audit: weekly reliability check that catches misses between calendar changes and inbox handling
• /eat-the-frog-daily and /eat-the-frog-weekly: picks the hardest task and sets your targets for the day or week
• /plan-event: full event plan from a single brief: task list, invitation, agenda, run-of-show, and follow-up

Members get new skills as they are built and can request builds for workflows specific to their role.

If you want to turn Claude Cowork into a powerful co-assistant, without building it yourself, join the EA Ops waitlist here:

whiteprints.vip/cowork.

Whiteprints. The AI Company for Executive Assistants.

whiteprints.vip