AI essentials and ChatGPT orientation

Work with AI.
Keep judgment.

A 60-minute core session for corporate knowledge workers.

Your trainers

Learn with Orfloat’s founders.

Practical guidance for safe, reviewable work with ChatGPT.

Portrait of Akram Ahmed
Akram AhmedCofounder and CTO, Orfloat
Portrait of Mufassir Ahmed
Mufassir AhmedCofounder and CEO, Orfloat

Your outcome today

Select a suitable task. Use ChatGPT safely. Review the result.Success means that you can explain your choice and show the checks you made.

What generative AI does

Creates a likely next response.
  • 01It uses patterns learned from data.
  • 02It creates text, images, audio, code, or other content.
  • 03It does not guarantee truth, intent, or approval.

Start with task fit

Choose work where a useful draft has value.

GOOD STARTDraft and transform

Summaries, outlines, rewrites, classifications, and first drafts.

GOOD STARTCompare and explore

Options, themes, questions, risks, and alternative approaches.

STOP OR ESCALATEHigh-stakes decisions

Do not delegate final legal, medical, financial, safety, or people decisions.

A quick selection test

Can I review the result?

  • The task is clear and bounded.
  • The information is approved.
  • A draft will save useful time.
  • I can detect a poor answer.
  • A person owns the final decision.
  • No hidden action can cause harm.

The core practice

Input. Work. Review.

01 / INPUTGive the task and allowed context.

Share only what is needed. State the output you need.

02 / WORKLet ChatGPT create a first result.

Stay ready to answer questions or change direction.

03 / REVIEWCheck before you use or share.

Correct, improve, or stop. You remain responsible.

Limits

Useful does not mean correct.

A fluent answer can include invented facts, weak logic, stale information, missed context, or biased assumptions.

  • 01Check names, dates, numbers, and calculations.
  • 02Open important sources. Read the relevant passage.
  • 03Check omissions, tone, and audience fit.
  • 04Ask a qualified reviewer for high-impact work.

Safe information boundary

Use the minimum information needed.

Usually safer

  • Synthetic or approved training data
  • Public facts from approved sources
  • Redacted text with only relevant details
  • Labels instead of real names

Stop and check

  • Personal, client, or confidential data
  • Credentials, access tokens, or security details
  • Contract, health, payment, or employee records
  • Anything your policy does not approve

Before every work task

Check the active workspace.

Personal and managed workspaces are separate environments. Switching does not move or copy data.

ChatGPT orientation

Four areas to notice.

1. Sidebar and new chat
2. Active workspace
3. Conversation canvas
4. Composer and tools

The conversation

Keep the evidence close to the answer.

Source links support review. Status shows that work is still in progress. An approval prompt asks before an important action. These elements appear only when the selected feature and task use them.

  • 01Open sources. Do not trust the label alone.
  • 02Read status updates and answer questions.
  • 03Approve only an action you understand.

Choose the smallest suitable surface

Chat, Search, Deep research, or Work?

CHAT

Think and draft

Use a conversation for short, interactive work with approved context.

Examples: rewrite, outline, compare
SEARCH / DEEP RESEARCH

Find and support

Use Search for quick current facts. Use Deep research for slower, documented synthesis.

Open and verify sources
WORK

Run longer work

Use Work for a longer, multi-step task that creates a finished deliverable.

Follow, redirect, and approve

Chat

Use conversation when you want rapid turns.

Chat can use text, approved files, and available tools. The visible tools and models depend on the account and workspace.

GOOD FIT“Turn these approved notes into a five-point agenda.”

You can inspect the notes, compare the result, and revise it in a few turns.

Search and Deep research

Match depth to the question.

Search quickly returns current information with links. Deep research takes more time to compare sources and produce a documented report.

  • FASTSearch: “What is the current public holiday date? Link the official source.”
  • DEEPDeep research: “Compare current regulatory approaches across three jurisdictions.”
  • CHECKOpen key sources and confirm that they support each important claim.

ChatGPT Work

Use longer execution only when the boundary is clear.

You can follow progress, answer questions, change direction, and approve important actions. Day 5 teaches controlled execution.

Availability check

Your screen can differ from this map.

ACCOUNTPlan and workspace

Features and limits can differ.

CONTROLRole and admin policy

Permissions can allow or block access.

CONTEXTPlatform and region

Web, desktop, mobile, and country availability can differ.

ACCESSProvider and rollout

Connected-account permission and staged release still apply.

Guided practice / safe request

Use approved sample data only.

Request: “Summarise these synthetic meeting notes into five action items. For each item, show an owner label and a due-date placeholder. Do not invent missing details. Mark missing information as ‘Not provided’.”
6 MINUTES / CHAT

Synthetic practice notes

Project Cedar check-in

• Team A will confirm the draft scope.
• Finance needs the sample cost table.
• A review meeting is planned, but no date is set.
• The project owner was not named.
• Use the approved internal template.

Do not fill the gaps.

  • 01Did it keep “owner” as not provided?
  • 02Did it keep the date as not provided?
  • 03Did it avoid adding commitments?

Review the result

Evidence before polish.

  • Each action comes from the notes.
  • No person, date, or fact was invented.
  • Missing details are visible.
  • The format has five clear items.
  • The wording does not imply approval.
  • A person will review before use.

Completion evidence: save the request and mark three checks you performed.

Day 1 recap

Three habits protect the value.

Choose

Start with a bounded task and a result you can review.

Protect

Use an approved workspace and the minimum allowed information.

Check

Verify facts, sources, numbers, omissions, and final use.

Next: Day 2 prompting

Today you chose the work.
Next, you will direct it.Day 2 turns objective, audience, context, sources, output, and constraints into a complete professional brief.

Day 1 complete

Keep judgment
in the loop.

You can now select a suitable task, use ChatGPT safely, and review the result.