ASK
Help me think.
Think: Chat“What are three ways we could solve this problem?”
✦A plain-English guide to AI
Which AI Door
Should I Walk Through?
You don't need to know every AI feature. You just need to know what you're trying to accomplish.
ChatGPT, Claude, and Microsoft Copilot increasingly do many of the same things—but they don't always call them the same thing.
Choose what you want to do ↓START HERE
What are you trying to accomplish?
Choose the statement that sounds most like you. We'll translate the capability across all three platforms.
HELP ME CHOOSEDescribe the work.
Describe the work.
We'll find your door.
Use plain language. If your answer could mean more than one thing, we'll ask one useful follow-up—not make you complete a technology quiz.
“We want AI to remember our client details every time we work on their project.”
Which AI platform are you using?
Same capability. Different name.
You don't need to memorize what every platform calls its features. Find what you want to accomplish, then scan across to see what ChatGPT, Claude, and Microsoft Copilot call that capability.
| What are you trying to do? | ChatGPT | Claude | Microsoft Copilot |
|---|---|---|---|
| Quick conversation | Chat | Chat | Copilot Chat |
| Persistent work context | Project | Project | Notebook |
| Create visual content | Create | Design | Create |
| Teach a repeatable process | Skill | Skill | Skill |
| Give AI a specific job | GPT / Agent | Agent | Agent |
| Delegate multi-step work | Work | Cowork | Cowork |
| Create interactive/output content | Site / app / files | Artifact | Page / Microsoft 365 files |
| Connect to other systems | Apps / Connectors / Plugins | Connectors / MCP | Connectors / Plugins |
| Give AI an executable capability | Tool | Tool | Tool / Action |
| Deep sourced investigation | Deep Research | Research | Research / Researcher |
| Software development | Codex | Claude Code | GitHub Copilot |
| Remember preferences/context | Memory / personalization | Memory / preferences | Memory / personalization |
| Repeat work automatically | Tasks / automations | Automated workflows | Power Automate / agent workflows |
Keep in mind: These are conceptual equivalents, not always exact feature-for-feature matches. AI products, capabilities, licensing, and terminology change quickly.
WHAT DOES THAT MEAN?
Your AI terminology translator
Hear an unfamiliar term? Pick it below for a quick, practical explanation.
IN PLAIN ENGLISH
Chat
A direct conversation with AI.
WHAT IT'S USEFUL FOR
Thinking, drafting, questions, and one-time tasks.
CHATGPTChatCLAUDEChatMICROSOFT COPILOTCopilot Chat
THE 7 LEVELS OF AI DELEGATION
How much are you handing over?
More delegation can create more value—and requires more intentional oversight.
CONTEXTUALIZE
Help me knowing the background.
Think: Project / Notebook“Based on this project, help me prepare for our meeting.”
STANDARDIZE
Do this using our process.
Think: Skill“Evaluate this using our standard go/no-go process.”
SPECIALIZE
Take on this particular role.
Think: Agent / GPT“Use our Proposal Agent to turn these notes into a draft.”
CONNECT
Use information from our actual systems.
Think: Connectors + Tools“Review our project files and identify open decisions.”
DELEGATE
Take this outcome and handle the steps.
Think: Work / Cowork“Analyze results, find themes, and build the presentation.”
AUTOMATE
Do this when it needs to happen.
Think: Tasks + Workflows“Every Monday, compile last week’s results and prepare a summary.”
!
Higher isn't necessarily better.
The goal isn't Level 7 for every workflow. Choose the appropriate level of AI delegation for the work, risk, and organization.
A FRIENDLY REALITY CHECK
Don't overbuild it.
Just because AI can do something doesn't mean you need to build it. The goal is not maximum AI. It's appropriate AI for the work.
×
Don't use Agent when a Project and good instructions will solve it.
×
Don't use Skill when you’ll only do it once.
×
Don't use Work / Cowork when you simply need to brainstorm.
×
Don't use Connector when the use case doesn’t require outside information.
×
Don't use Automation when the process is still broken.
HUMAN OVERSIGHT THROUGHOUT
More autonomy calls for more intention.
As AI gains access, autonomy, impact, or risk, organizations should be increasingly thoughtful about:
PermissionsData accessSecurityHuman reviewApproval pointsAccountabilityGovernance
Start With the Work,
Not the AI Feature.
AI companies will keep introducing new names, buttons, features, and interfaces. You don't need to chase all of them.
01What are we trying to accomplish?
02What context does AI need?
03Is this something we do repeatedly?
04Does AI need access to another system?
05Are we asking AI to think, create, act, or complete a workflow?
06Where does a human need to review, approve, or decide?
Use case first. Tool second. Human oversight throughout.
Back to the doors ↑