1. Overview

As Chartered Accountants, we often need to prepare technical articles based on information available across several documents and websites. This involves finding reliable information, reading and understanding it, bringing it together, drafting the article, checking whether anything has been missed and then refining the final article.

This use case demonstrates how ChatGPT, Microsoft Copilot and Perplexity can be used at different stages of this process.

The idea is simple: instead of depending on one AI tool, different AI tools are used for different purposes such as drafting, reviewing, comparing and improving the article.

AI supports the work, but the final responsibility remains with the Chartered Accountant, who reviews the content and applies professional judgment.

2. Problem / Challenge Addressed

Preparing a good technical article from information available across multiple sources can take considerable time.

A professional generally has to:

1. Search for relevant and reliable information;
2. Read and understand several documents;
3. Extract the information relevant to the article;
4. Bring information from different sources together;
5. Prepare the first draft;
6. Check for missing points;
7. Remove repetition;
8. Improve the structure and language;
9. Review the article again; and
10. Ensure that the final article is technically correct and useful to the reader.

When these activities have to be repeated several times, a significant amount of professional time is spent on collecting, drafting and reviewing information.

The challenge is therefore to save time and reduce repetitive work without compromising technical quality, reliability or professional judgment.

3. Step 1 – Collecting and Preparing Authentic Source Material

Before using any AI tool, I first collect the relevant material from authentic and recognised sources.

The sources may include:

1. Official Government websites, particularly .gov.in websites;
2. Regulatory and statutory websites;
3. ICAI and other recognised professional institutions; and
4. Other recognised websites and reliable professional literature.

The objective is to start the AI process with specific source material that I have identified and selected, rather than allowing the AI to independently search the internet and decide which information to use.

Creating a Simple Source File

After identifying the relevant sources, I copy the textual content of the relevant material into a Notepad file.

I intentionally do not copy unnecessary graphics, images, website formatting or other visual elements. The purpose is to retain the relevant text in a simple and lightweight format.

This keeps the file size as small as possible and makes it easier to provide the material to the AI tool.

The original websites and source material are retained separately for reference and final verification.

Uploading the Source File to ChatGPT

The Notepad file containing the collected source material is then uploaded to ChatGPT as the first step of the AI workflow.

A specific instruction is given to ChatGPT to:

Use only the information contained in the uploaded source file for preparing the article and do not carry out additional research or obtain information from any other website or source.

This is an important control in the workflow.

It ensures that the first draft is prepared from the source material selected by me, rather than from information independently obtained by the AI.

The process can therefore be summarised as:

Authentic Websites / Literature → Relevant Text → Lightweight Notepad File → Upload to ChatGPT → Controlled First Draft

4. AI-Based Solution

Once the source material has been collected and uploaded, I use three AI tools, with each tool being used for a specific purpose.



The purpose of using three different tools is to get more than one perspective on the same article.

For example, if ChatGPT prepares the first draft, Copilot and Perplexity can review that draft independently. Their observations can then be considered while refining the article.

Importantly, an AI suggestion is not accepted merely because an AI tool has suggested it. The suggestion is checked against the original source material before it is incorporated.

5. Workflow – Multi-AI Review Loop

The complete workflow can be explained through the following steps.

Step 1 – Identify Authentic Sources

Relevant material is collected from authentic and recognised sources, including government websites, particularly .gov.in sources, and recognised professional literature.

Step 2 – Create a Lightweight Source File

The relevant text is copied into a simple Notepad file.

No unnecessary graphics, images or website formatting are included. This keeps the source file as small and focused as possible.

Step 3 – Upload the Source File to ChatGPT

The Notepad file is uploaded to ChatGPT.

A specific instruction is given:

Do not conduct any additional research. Do not search other websites. Prepare the article only from the information contained in the uploaded source material.

This creates a controlled starting point for the first draft.

Step 4 – Generate the Initial Article

ChatGPT is given a structured prompt explaining:

1. The subject of the article;
2. The intended audience;
3. The required structure;
4. The desired professional tone; and
5. Any specific requirements for the article.

ChatGPT then prepares the initial draft using the uploaded source material.

Step 5 – Draft → Microsoft Copilot

The first draft is given to Microsoft Copilot for an independent review.

Copilot is asked to look for:

1. Missing points;
2. Inconsistencies;
3. Repetition;
4. Structural issues; and
5. Possible improvements.

Step 6 – Draft → Perplexity

The same draft is reviewed using Perplexity.

This provides another perspective and may highlight points that were not identified during the first review.

Step 7 – Review Comments → ChatGPT

The observations received from Copilot and Perplexity are brought back to ChatGPT.

ChatGPT is asked to consider these observations and improve the article, while checking the suggested changes against the original source material.

Step 8 – Source Material → Copilot & Perplexity

The source material is also provided independently to Copilot and Perplexity to prepare alternative versions of the article.

This allows me to see how different AI tools approach the same technical material.

Step 9 – Compare the Different Drafts

The different versions are brought back to ChatGPT for comparison.

The comparison focuses on:

1. Which version explains a point better;
2. Whether any relevant point has been missed;
3. Whether the structure can be improved;
4. Whether there is unnecessary repetition; and
5. Which suggestions are supported by the original source material.

The useful points are then incorporated into the final draft.

Step 10 – Final Review by the CA

The final article is reviewed by the Chartered Accountant against the original source material and, where necessary, the original websites and documents.

The CA makes the final decision on what should or should not be included in the article.

This is an important part of the workflow because AI assists in preparing the article, but the professional remains responsible for the final output.

6. Practical Demonstration – 10-Minute Hackathon Presentation

The presentation will focus mainly on a live demonstration rather than a theoretical explanation.

The demonstration will show the complete journey of the article:

1. Identifying authentic and recognised source websites;
2. Collecting the relevant material from those sources;
3. Copying the relevant text into a simple Notepad file;
4. Keeping the file lightweight by excluding unnecessary graphics and formatting;
5. Uploading the Notepad file to ChatGPT;
6. Giving ChatGPT a clear instruction not to conduct any external research;
7. Providing the structured prompt for preparing the article;
8. Generating the first draft from the uploaded source material;
9. Giving the draft to Microsoft Copilot for independent review;
10. Giving the draft to Perplexity for another independent review;
11. Bringing the observations back to ChatGPT;
12. Refining the article based on relevant observations;
13. Providing the source material independently to Copilot and Perplexity to generate alternative drafts;
14. Comparing the different versions in ChatGPT;
15. Incorporating useful and source-supported improvements; and
16. Carrying out the final review as a Chartered Accountant.

The objective is to show how the tools can actually be used in a professional CA workflow, rather than simply discussing what AI can do.

7. Benefits / Impact

The main benefits of this approach are:

Controlled Source Material

The process starts with source material selected from authentic and recognised websites and literature.

Reduced Uncontrolled Research

By specifically instructing ChatGPT not to conduct additional research for the initial article, the first draft remains focused on the source material provided.

Saving Time

AI can reduce the time spent on repetitive activities such as summarising, drafting and comparing different versions.

Faster First Draft

Instead of starting with a blank page, the professional can obtain a structured first draft from the selected source material and then focus on reviewing and improving it.

More Than One Perspective

Using different AI tools gives the professional different perspectives on the same material. This can help bring attention to points that may otherwise be missed.

Better Structure and Presentation

AI can help organise technical information, improve the flow of the article and make complex information easier to present.

Easier Comparison

Alternative drafts can be compared more quickly, allowing the professional to select the better explanation or presentation.

Professional Control

The most important benefit is that the CA remains in control. AI suggestions are reviewed and accepted only when they are appropriate and supported by the underlying material.

In simple terms:

AI helps with the work, but the CA makes the final professional decision.

8. Role of Professional Judgment

The output generated by AI is not treated as the final answer.

The Chartered Accountant reviews the article against the original source material and checks:

1. Whether the technical content is correct;
2. Whether important points have been missed;
3. Whether the statements are supported by the source material;
4. Whether the interpretation is appropriate;
5. Whether the article is relevant to the intended reader; and
6. Whether the final article is professionally suitable.

Therefore, the approach can be summarised as:

AI-Assisted Article Writing + Multi-AI Review + Human Professional Judgment

9. Key Innovation

The main feature of this use case is that different AI tools are used to review and improve the work of each other, starting with information collected from authentic sources.

A simple approach would be:

Source Websites → AI → Final Article

My proposed approach is:

Authentic Sources → Lightweight Source File → Controlled ChatGPT Draft → Copilot Review → Perplexity Review → ChatGPT Refinement → Alternative Drafts → Comparison → CA Review → Final Article

This creates a multi-AI review loop with a controlled source base.

The objective is not simply to ask AI to write an article. The objective is to use different AI tools to draft, question, compare and improve the article, while the Chartered Accountant remains responsible for the final content.

10. Technology / AI Stack



The multi-AI approach allows the outputs of different platforms to be compared rather than depending entirely on one AI tool.

11. Future Scope

The same approach can be developed further in the future.

Possible enhancements include:

1. Standard prompts for different types of CA articles;
2. Linking each important statement in the article to its source;
3. Checking specific clauses or provisions against source documents;
4. Automatically identifying statements that do not have adequate support;
5. Comparing the final article with the original documents;
6. Creating standard article formats for different audiences;
7. Maintaining versions and tracking changes; and
8. Connecting the workflow with an organisation's internal knowledge repository.

These improvements could make the process more consistent and further reduce the time spent on routine drafting and review.

12. Conclusion

“Article Writing using Different Authentic AI Tools” demonstrates a practical way in which AI can be used in the day-to-day work of a Chartered Accountant.

The process starts with authentic source material, which is collected from government and recognised sources and converted into a simple, lightweight text file.

The source file is then uploaded to ChatGPT with a specific instruction to work only with the information provided and not to conduct additional research. Different AI tools are subsequently used to draft, review, compare and improve the article.

The use of ChatGPT, Microsoft Copilot and Perplexity helps in drafting faster, obtaining different perspectives, identifying possible gaps and improving the final article.

At the same time, the process does not remove the need for professional expertise.

The Chartered Accountant remains responsible for checking the content, applying professional judgment and approving the final article.

The overall approach can therefore be summed up as:

Authentic Sources + Controlled AI Input + Multiple AI Perspectives + Human Professional Judgment