daBongo LMS AI Training Courses

← Back to Course

Claude Chat, Code, and Cowork – Three Modes, One AI

Lesson 4: Choosing Your Mode – Which Tool for Which Task

Lesson Objectives

By the end of this lesson, students should be able to:

  • Apply a five-question decision framework to any task
  • Explain why using the wrong mode produces worse results than the right one
  • Categorize a list of common tasks by correct Claude mode
  • Build the habit of pausing to choose mode before starting any Claude session

Lesson Content

The wrong mode costs you more than you realize.

Using Chat for a task that belongs in Code produces a response you then have to act on manually – extra steps, extra time, extra room for error. Using Code for a task that belongs in Chat means Claude has access to your file system when it does not need it. Using Chat for a task that belongs in Cowork means you copy-paste between Claude and your real tools instead of having Claude work inside them.

Most users never switch modes because they do not know the framework for choosing. Here it is.

The five-question decision framework.

Work through these questions for any task you bring to Claude:

Question 1: Does this task require Claude to take action on files or your computer?

  • Reading, writing, editing, or organizing files on your local machine?
  • Running commands or scripts?
  • Working with code, databases, or developer tools?
  • If yes to any: Code mode

Question 2: Does this task require Claude to work inside your real apps – email, calendar, documents, Slack?

  • Summarizing or drafting in your actual email inbox?
  • Finding a specific file in your real Google Drive or OneDrive?
  • Posting to or reading from Slack?
  • Any workflow that needs to read or write to a connected cloud service?
  • If yes to any: Cowork mode (with relevant Connector authorized)

Question 3: Does this task involve a repeating workflow you want to run on command?

  • A daily briefing you want every morning?
  • A recurring report format you run weekly?
  • A structured workflow you do the same way every time?
  • If yes: Cowork mode (build a Skill or Routine)

Question 4: Is this task entirely conversational – thinking, writing, analyzing, explaining, or creating?

  • Writing, drafting, or editing content
  • Answering a question or explaining a concept
  • Brainstorming or exploring ideas
  • Summarizing a document you upload yourself
  • Planning, decision support, or analytical thinking
  • If yes: Chat mode

Question 5: Are you unsure?

  • Start in Chat mode. It is the general-purpose mode that handles the widest range of tasks. You can always switch if you find you need something Chat cannot provide.

Quick reference table.

| Task type | Best mode | |———–|———–| | Write a blog post | Chat | | Summarize a PDF you upload | Chat | | Brainstorm campaign ideas | Chat | | Read and organize 200 files in a folder | Code | | Review and edit a codebase | Code | | Automate a daily email summary | Cowork | | Draft a reply using real email history | Cowork | | Find a specific doc in Google Drive | Cowork | | Answer a complex question | Chat | | Build a weekly report Routine | Cowork | | Process a data file and produce analysis | Code | | Explain a concept in plain language | Chat |

What the right mode feels like.

When you are in the right mode, Claude's responses feel immediately actionable. Chat gives you text you can use. Code shows you a plan you can approve and watch execute. Cowork runs a Skill and produces output drawn from your real tools. If you are doing a lot of manual copy-paste between Claude and another application, that is often a signal that Cowork would serve this task better. If you find yourself describing your file structure repeatedly in Chat, that is a signal that Code mode with a CLAUDE.md would serve you better.

Practical Example

A project manager has five tasks to complete before noon.

She pauses before opening Claude and runs each through the framework:

  • "Draft the project status update for the client" – Conversational writing task – Chat
  • "Pull all meeting notes from the last two weeks and summarize the open action items" – Needs to search her real Google Drive – Cowork
  • "Rename and organize the 80 asset files the design team sent" – Local file operations – Code
  • "Figure out the best approach for the upcoming scope change conversation" – Strategic thinking and planning – Chat
  • "Send the weekly Friday status email to the stakeholder list" – Automated recurring task using real email – Cowork Routine

She spends 30 seconds on mode selection and then works without switching mid-task because each session is already in the right mode.

Lesser-Known Tip

If a task genuinely spans two modes – for example, you need to process local files (Code) and then email the results to stakeholders (Cowork) – handle the Code portion first and produce a clear output, then switch to Cowork to handle the communication step. Trying to do both in a single mode forces compromises. Treating mode transitions as natural handoffs between phases of a workflow is more efficient than trying to find one mode that does everything.

Safety Notes

Mode selection is also a safety decision. Code mode's file access means Claude can modify files – verify you are in the right mode before approving any file-writing action in Code to avoid unintended edits. Cowork's Connector access means Claude can act inside your real apps – confirm the correct Connector is authorized before running a Skill that sends email or posts to Slack, to avoid sending something prematurely. When in doubt about what Claude is about to do in either mode, use plan mode to review its full intention before execution.

Practice Task

Take five tasks from your current work – things you are actually doing this week – and apply the decision framework to each one. For each task, write down: (1) which question from the framework applies, and (2) which mode the framework recommends. Then compare that to which mode you would have instinctively opened without the framework. Note any differences and consider what that tells you about your current Claude habits.

Completion Check

You should be able to apply all five framework questions to any task and arrive at a mode recommendation, explain why the right mode produces better results than defaulting to a single mode for everything, and recognize the signals that indicate you are using the wrong mode.

Log in and enroll to access lesson quizzes.

← Back to Course

Scroll to Top