How ready is your team?

A token is a chunk of text, often part of a word, not always a whole word.

Answer: True AI models read and write in tokens. In English, a token averages about three-quarters of a word.

A context window is the AI’s permanent memory of everything you have ever told it.

Answer: False The context window is the AI’s working memory for one chat. When the chat ends, so does the window.

Without searching the web, an AI knows about events after its training ended.

Answer: False A model learns from data up to a cutoff date. For anything newer, it needs web search.

An AI can state something false in a confident tone.

Answer: True The AI predicts likely text, so a wrong answer can sound as sure as a right one. Check the facts.

The longer a chat gets, the more accurate the AI becomes.

Answer: False A long chat fills the context window, and early details get less attention. Answers drift.

Starting a new chat when you switch topics can improve answers.

Answer: True A new chat clears the unrelated history, so the AI works only on the new task.

A .md (Markdown) file is plain text with light formatting that an AI reads well.

Answer: True Markdown marks headings and lists with symbols such as #, which makes it a good format for instructions.

Uploaded files and the AI’s own replies count toward the context window.

Answer: True Everything in the chat takes up room: your messages, your files, and its answers.

Every AI tool automatically remembers your earlier chats in a new chat.

Answer: False Some tools have memory, some don’t, and memory can be turned off. Check your settings.

Saved instructions or a project let you set your preferences once instead of retyping them.

Answer: True Custom instructions and projects carry your preferences and files into every chat.

Memory features store details about you, and you can review and delete them.

Answer: True In Claude and ChatGPT, you can see what memory holds and remove items in settings.

Anything saved in memory is always correct and current.

Answer: False Memory can keep outdated or wrong details. Review it from time to time.

If web search is off, the AI answers from its training data.

Answer: True Without search, the AI uses what it learned in training, which can be out of date.

A connector lets an AI read from or act in an app such as email, a calendar, or Google Drive.

Answer: True Connectors link the AI to the apps your team already uses, such as Gmail, Drive, and Slack.

An AI can only produce text. It can’t take actions in other apps.

Answer: False With connectors, an AI can draft emails, create files, and update records in other apps.

A skill is a saved set of instructions that the AI loads when a task needs it.

Answer: True Claude calls these skills. They keep long instructions out of the chat until a task needs them.

Thinking models suit hard problems better than quick lookups.

Answer: True A thinking model works through a problem before it answers. It’s slower, so save it for hard work.

The most expensive model is the best choice for every task.

Answer: False Faster, lower-cost models handle most everyday tasks well. Use the largest model for hard problems.

If an AI cites a source, you can skip opening it.

Answer: False An AI can cite a source that doesn’t say what it claims, or doesn’t exist. Open it.

You can paste customer or contract data into any AI tool.

Answer: False Use only the AI tools your company approves for customer data, and follow its policy.