Cue helps you draft, analyse, and summarise content. To get the most accurate and useful outputs, your prompts should be clear, structured, and provide context.
Key Components of a Good Prompt
Every good prompt is made up of three components:
- Request: What you want Cue to do.
Example: “Summarise the key GST compliance deadlines for a small business in India.”
- Context: Background information that helps Cue understand your request.
Example: “The business is registered in Maharashtra and files monthly GST returns under the regular scheme.”
- Output: The format and style in which you want the AI’s response.
Example: “…in a table with columns for tax type, due date, and remarks.”
Providing all three elements helps Cue produce responses that are more relevant, structured, and actionable.
Giving the right prompt
Cue can process complex tasks, lengthy documents, and multiple files at once. However, how you structure your prompts plays a major role in the quality of the output.
Using follow ups to divide a query
Avoid combining multiple instructions into a single prompt. Instead, guide Cue step by step.
- Less effective:
“First summarise this tax audit report, then identify errors, and then suggest corrective actions.” - Better:
- Initial prompt:
“Summarise this tax audit report with clear headings.”
- Follow-up:
“Highlight any inconsistencies or red flags in the summary above.”
- Follow-up:
“Suggest corrective actions for each red flag identified.”
Revising outputs
When you want to change the overall tone, style, or approach of the output.
Example:
- Prompt:
“Draft an advisory note on penalties for late filing of ITR.” - Revision:
“Make this more concise and suitable for a non-technical client.”
When a specific section of the output needs to change.
Example:
- Prompt:
“Draft a GST engagement letter for a new client.” - Targeted revision:
“Rewrite the scope of services section to exclude litigation and add a disclaimer for departmental notices.”
Output length
Cue cannot generate content based on page counts (for example, “a 10-page note”). Instead, guide it using descriptive instructions.
- Less effective:
“Write a 10-page analysis of capital gains tax changes.” - Better:
“Write a detailed and structured analysis of recent capital gains tax changes, covering background, key amendments, and practical implications for individual taxpayers.”
Working with long or multiple documents
Cue can analyse lengthy files and multiple documents at once. However, broad questions may lead to less precise outputs.
- Instead of:
“Analyse these three years of GST returns.” - Try:
“Review these GST returns and identify mismatches between GSTR-1 and GSTR-3B for FY 2023–24.”
If needed, follow up with narrower questions to go deeper into specific issues.
Give clear instructions
When using multiple documents, explicitly tell Cue how each source should be used.
Example:
“Use the GST audit report to identify observations and refer to the reconciliation statement to check whether those observations were already addressed.”
Specify region for better context
For tax or any regulatory queries, always name the relevant location and act for more accuracy.
- Less effective:
“Summarise late filing penalties.” - Better:
“Summarise late filing penalties under the Indian Income Tax Act for ITR-3 applicable for FY 2024–25.”
Summarisation
Summarisation works best when you specify the purpose and format of the summary.
- Less effective:
“Summarise this assessment order.” - Better:
“Summarise this income tax assessment order in a structured note with headings and highlight possible grounds for appeal.” - Best (when you know what you’re looking for):
“I am advising a small business assessed under Section 144. Review this assessment order and summarise the key additions, the reasoning used by the Assessing Officer, and potential procedural lapses. Present the output with clear subheadings.”
General prompts
Cue is more than a spell-check or drafting tool. Use it to challenge clarity, surface risks, and identify gaps.
- Instead of:
“Check this email for grammar errors.” - Try:
“Is there anything in this email that my customer might find unclear or confusing? List possible follow-up questions the customer may ask.”