Tech Tips
AI Usage – Productivity
Practical ways to use AI tools more effectively for drafting, structuring, and thinking through work—without replacing judgment.
Interrogate Documents, Don’t Just Summarise Them
What it does:
Helps you extract hidden risks, inconsistencies, and missed obligations from documents—not just summaries.
How to use it:
Upload your document to NotebookLM or ChatGPT
Use prompts like:
* “What assumptions does this document make that may not hold true?”
* “List clauses that can create financial exposure or compliance risk.”
* “What is missing that should ideally be included?”
Pro tip:
Ask:
“Challenge this document like a regulator/auditor—what would they question?”
This flips AI from passive summariser to active reviewer.
Note:
Mask confidential data before uploading.
Use AI as a “Second Reviewer,” Not a Typing Tool
What it does:
Turns AI into a critical reviewer that challenges your conclusions instead of just generating content.
How to use it:
After drafting your answer/position, paste it into AI and prompt:
“Act like a strict reviewer. Challenge my conclusion. Identify weak assumptions, missing risks, and alternative interpretations.”
Optional follow-up:
“Now strengthen my argument after fixing these gaps.”
Where it helps:
• GST classification decisions
• Legal interpretations
• Replying to notices
• Internal memos and opinions
Why it matters:
Most professionals use AI to support their view. That creates blind spots.
Using it to attack your view improves accuracy and defensibility.
Clear prompts produce better AI responses.
Vague input usually leads to generic answers.
Instead of:
“Check this calculation”
Try:
“Review GST calculation logic for RCM on freight under reverse charge.”
Good prompts typically include:
• *Context* (GST, income tax, audit, etc.)
• *Specific task* (review, explain, verify)
• *Relevant condition or rule*
Precise input → more useful output and less rework.
AI works better with CONTEXT, not long text.
Instead of pasting everything, try:
• Who you are (CA / practitioner)
• What you want (summary / checklist / reply)
• Output format (points / table)
Short prompts → clearer answers.
Don’t just ask for answers. Ask for *structured disagreement.*
Use:
“Give me 2 alternative approaches. For each, list: assumptions, risks, edge cases, and when it will fail.”
Optional add-on (power move):
“Now critique the strongest option.”
Why this matters:
• Forces the model to expose hidden assumptions
• Surfaces implementation risks before you commit
• Reveals edge cases (where most professional errors happen)
• Simulates internal review without group bias
Most AI errors don’t come from wrong answers.
They come from unchallenged assumptions.
Single output = efficiency.
Structured alternatives + critique = control.
Never ask AI to “summarise this email” blindly.
Instead say:
“Summarise and highlight ONLY items requiring action or reply.”
Why this matters:
• Prevents missing obligations
• Filters noise
• Saves review time
AI should prioritise work, not just compress text.
Use AI as a drafting assistant, not a data processor.
Good uses:
• Rewriting emails professionally
• Summarising notifications
• Creating checklists
• Explaining unfamiliar concepts
Avoid uploading:
• Client data
• Returns
• Working papers
AI can assist speed.
Judgment and data responsibility stay with you.
AI responds better to context than keywords.
Instead of:
“Draft a reply”
Try:
“Draft a polite reply declining a deadline extension,
in a professional tone.”
Clear intent → usable output.
AI tools answer confidently — accuracy is not guaranteed.
Before relying on output:
• Recheck section numbers & limits
• Verify dates, thresholds, applicability
• Use it as a draft, not a final view
Treat AI as an assistant, not an associate.
Judgement remains yours.
