Home / AI Courses / ChatGPT Guide

ChatGPT in Practice

Advanced65 min13 chaptersFree

Course Contents

  1. 6.1 Interface Tour and Basics
  2. 6.2 Conversation Skills That Matter
  3. 6.3 File Upload and Analysis (PDF, Excel, Images)
  4. 6.4 Custom GPTs: Build Your Own Assistant
  5. 6.5 Web Browsing and Current Information
  6. 6.6 Image Generation (DALL-E 3)
  7. 6.7 Voice Mode and Translation
  8. 6.8 Canvas: Collaborative Writing
  9. 6.9 Code Interpreter: Data Analysis for Everyone
  10. 6.10 Memory and Personalisation
  11. 6.11 Team and Enterprise Plans
  12. 6.12 Ten Workplace Scenarios, Fully Worked
  13. 6.13 Summary + Workplace Project

6.1 Interface Tour and Basics

The ChatGPT interface is deliberately simple — one input box — but five features change how much you get from it: custom instructions (persistent context about you), temporary chat (no history saved), model selection (speed vs capability), shared links (collaboration), and the mobile app's voice mode. Spend ten minutes setting these up once; they pay off daily.

6.2 Conversation Skills That Matter

The golden opener You are my [industry] consultant. Context: [one line about my business]. Task: [what you need]. Before answering, ask me the 3 questions you need clarified.
"Ask before answering" is the most underused trick: output quality after AI's clarifying questions far exceeds one-shot attempts. Try it once and you won't go back.
Chapter checklist

6.3 File Upload and Analysis (PDF, Excel, Images)

Business Example

Quarterly reports in 30 minutes

A pet-wholesale manager reads four 40-60 page supplier market reports each quarter. Her flow now: upload → "List the 5 key numbers and 3 implications for a small wholesaler from each report" → merge the four summaries → "Overall, which categories should I stock up on next quarter?" Half a day became 30 minutes, with no skipped pages. The trick is asking for "numbers + implications" rather than "a summary" — numbers force the AI to be concrete.

Common mistakeUploading spreadsheets containing customer personal data
Right approachDelete name/phone columns first — analysis only needs the numeric patterns
Common mistakeGiving up on scanned (image-based) PDFs
Right approachAI can OCR them, with lower accuracy — convert critical documents to text-based PDFs

6.4 Custom GPTs: Build Your Own Assistant

  1. Pick one recurring task with a stable format (e.g. weekly Instagram posts in your brand voice).
  2. Write the instructions: role, task, format, constraints — your tested four-part prompt becomes the GPT's brain.
  3. Add 2-3 example outputs in the knowledge section for tone anchoring.
  4. Test with 5 real cases, refine instructions until consistent, then share with your team.

6.5 Web Browsing and Current Information

Enable browsing whenever your question depends on "now" — prices, news, recent events. Ask for sources with every fact, then click through two citations to build calibrated trust. Browsing outputs are grounded but slower; use it for research, not chat.

6.6 Image Generation (DALL-E 3)

Describe scenes like a camera director: subject, setting, style, lighting, mood. For business use — social visuals, product concept mockups, event posters — the winning pattern is generate four variations, pick one, then refine with targeted feedback ("keep the composition, change the cat to orange"). Always review generated images for weird text artifacts before publishing.

6.7 Voice Mode and Translation

6.8 Canvas: Collaborative Writing

Canvas opens an editable side-by-side workspace where you highlight passages and ask for targeted edits, length adjustments or tone shifts without regenerating the whole document. It is the closest thing to co-writing with the AI — ideal for proposals and long articles where you want surgical changes.

6.9 Code Interpreter: Data Analysis for Everyone

Zero-code data analysis flow

  1. Upload clean data: de-identified sales/booking/inventory spreadsheet.
  2. Let AI orient itself: "Analyse this file — what columns, time span, obvious trends do you see?"
  3. Ask decision questions: "Which product category is growing fastest month over month? Plot it."
  4. Request actions: "Based on the trend, what should I stock up on next quarter? Give 3 reasons."
  5. Verify one conclusion manually: spot-check one month in Excel — validate before you trust.
Chapter checklist

6.10 Memory and Personalisation

Memory lets ChatGPT retain facts across conversations ("I run a pet shop in Kowloon", "I prefer Traditional Chinese"). Manage it deliberately: pin the facts that help, delete ones that don't, and remember memory is convenience — critical business rules belong in your saved prompts, not in memory.

6.11 Team and Enterprise Plans

Team plans add shared GPTs, central billing and stronger data guarantees (conversations excluded from training). The rollout playbook: pilot with power users, build a shared prompt library, measure time saved before scaling.

6.12 Ten Workplace Scenarios, Fully Worked

ScenarioOpener prompt
Product description"You are my copywriter. Product: [x]. Audience: [y]. Write 3 descriptions, 80 words each, tone: [z]"
Complaint response"Draft a reply to this complaint: [text]. Acknowledge, apologise, offer [resolution], under 100 words"
Meeting prep"Brief me on [topic] in 5 bullets — what happened recently, what to watch, 2 smart questions to ask"
Data check"[paste table] — what's the trend, any anomalies, what should I investigate?"
Email triage"Sort these 12 emails into: reply now / delegate / archive, with one-line reasons"

The remaining five scenarios follow the same pattern — role, context, task, format. Build your own scenario table in the capstone project below.

6.13 Summary + Workplace Project

Capstone: your ChatGPT playbook
Summary

You now command ChatGPT's full surface: conversation craft, file analysis, custom GPTs, browsing, images, voice, canvas, and code-interpreter data analysis. Sharpen the underlying skill next: Prompt Engineering Complete Guide →

Keep learning

Advanced tools track
Foundations

Frequently Asked Questions

Free or paid plan?
Use the free tier for a month and log the limits you actually hit. If caps and speed genuinely block your output, the subscription pays back within an hour or two of saved time.
Is my business data safe?
Free-tier conversations may improve models. De-identify customer data first; sensitive industries should use plans with enterprise data guarantees.
Can it really analyse my Excel file?
Yes — upload it and ask questions in plain language: trends, anomalies, charts. Verify one conclusion manually before acting on the analysis.

Ready to put AI to work in your business?

Buddylala.ai assigns you a dedicated AI account manager — zero technical overhead, integrating ERP/CRM/Shopify.

Free WhatsApp consultation →