Sneha and Dazza working doc

# Use Case Exercise #1:

Example Prompts for GPT-4:

• "Summarize the key points of the IRS transfer pricing documentation best practices."
• "Transform the key points into actionable advisory points for corporate clients."
• "Organize these advisory points into a logical structure."
• "Draft a comprehensive tax advisory for corporate clients about these tax rules."

**Sneha Demo**

Prompt 0:

candidate: Please draft a tax advisory for corporate clients from a big-4 firm about this recent tax guidance by the IRS. Include examples and best practices as well as warnings and cautions.

Prompt 1:

Please provide a summary of the information at the following URL. Please use bullet points: https://www.irs.gov/businesses/international-businesses/transfer-pricing-documentation-best-practices-frequently-asked-questions-faqs

Prompt 2

Please draft a tax advisory for corporate clients from a big-4 firm about this "Transfer Pricing Documentation Best Practices Frequently Asked Questions (FAQs)" guidance by the IRS. Include examples and best practices as well as warnings and cautions.

Prompt 3

Very good, now please take the same information about the IRS documentation and help me draft talking point for a big 4 tax advisor who is about to speak to clients. Please provide talking points about these tax rules first for Jane Doe to advise Hewlett Packard, and then for John Smith to advise Verizon. Please write both sets of talking points in memo and quick reference formats.

# Use Case #2: Working With Tabular Data

### Context:

In this assignment, you'll be working with a CSV file that contains data about labor and software purchases. Your task is to conversationally manipulate this structured data to answer specific questions and perform certain calculations.

### Instructions:

* Examine Data: Start by examining the given CSV file. Understand the data structure, types of data, and any patterns or anomalies in the data.
* Formulate Questions: Based on the examination, formulate a series of questions that you want Bing Chat or EYQ to answer. For example, you might ask Bing Chat or EYQ to calculate the total amount of input VAT (ie: VAT that can be reclaimed) for hardware vendors based on the applicable jurisdiction or the total spend on labor.
* Create Prompts: Create specific prompts to answer each question. Be sure your prompts are clear and specific.
* Test Prompts: Test each prompt with Bing Chat to see if it correctly answers your question. If not, refine your prompt and try again.
* Perform Calculations: Use Bing Chat or EYQ to perform specific calculations, such as calculating the estimated VAT tax for hardware purchases in each jurisdiction and in the aggregate.
* Review and Refine: Carefully review the outputs provided by Bing Chat or EYQ. If the output is not what you expected, refine your prompt and try again.

### Example Prompts for Bing Chat:

Prompt 1

In the above table, please 1) identify the enterprise hardware payees, and 2) the aggregate spending on hardware

Prompt 2

Thanks, now please provide a list of the EU payees.

Prompt 3

Based on the provided data, how much VAT could be reclaimed for the transactions with payees subject to 20% VAT? Please list the sum for each payee and the total sum.

Prompt 4

Thanks! Now, for payees in the Netherlands, how much VAT could be reclaimed for the transactions with those payees? Please list the amount of VAT for each transaction separately and provide a total.

### Sample Data for Exercise:

ID,Payee,Date,Amount,Category,Address,Phone Number,Input VAT
100 24,McDonalds,1/25/22,15,Food,,,
100 84,McDonalds,3/26/22,15,Food,,,
116 12,Amazon,1/13/22,900,Hardware,"Mages Corner, 5 Av. Anatole France, 75007 Paris, France",,20%
120 42,Amazon,2/12/22,900,Hardware,"Mages Corner, 5 Av. Anatole France, 75007 Paris, France",,20%
116 72,Amazon,3/14/22,900,Hardware,"Mages Corner, 5 Av. Anatole France, 75007 Paris, France",,20%
96 14,Furniture Warehouse,1/15/22,1000,Furniture,"Mages Corner, 5 Av. Anatole France, 75007 Paris, France",,20%
100 44,Furniture Warehouse,2/14/22,1000,Furniture,"Mages Corner, 5 Av. Anatole France, 75007 Paris, France",,20%
96 74,Furniture Warehouse,3/16/22,1000,Furniture,"Mages Corner, 5 Av. Anatole France, 75007 Paris, France",,20%
104 10,Desktops Galore,1/11/22,800,Hardware,"London SW1A 0AA, UK",,20%
112 40,Desktops Galore,2/10/22,800,Hardware,"London SW1A 0AA, UK",,20%
112 100,Desktops Galore,4/11/22,800,Hardware,"London SW1A 0AA, UK",,20%
112 8,IBM Hardware,1/9/22,700,Hardware,"Museumplein 6, 1071 DJ Amsterdam, Netherlands",,21%
108 38,IBM Hardware,2/8/22,700,Hardware,"Museumplein 6, 1071 DJ Amsterdam, Netherlands",,21%
112 68,IBM Hardware,3/10/22,700,Hardware,"Museumplein 6, 1071 DJ Amsterdam, Netherlands",,21%
108 98,IBM Hardware,4/9/22,700,Hardware,"Museumplein 6, 1071 DJ Amsterdam, Netherlands",,21%
120 2,Microsoft,1/3/22,500,Hardware,"Mages Corner, 5 Av. Anatole France, 75007 Paris, France",,20%
116 32,Microsoft,2/2/22,500,Hardware,"Mages Corner, 5 Av. Anatole France, 75007 Paris, France",,20%
120 62,Microsoft,3/4/22,500,Hardware,"Mages Corner, 5 Av. Anatole France, 75007 Paris, France",,20%
116 92,Microsoft,4/3/22,500,Hardware,"Mages Corner, 5 Av. Anatole France, 75007 Paris, France",,20%
108 6,Oracle,1/7/22,600,Hardware,"London SW1A 0AA, UK",,20%
100 36,Oracle,2/6/22,600,Hardware,"London SW1A 0AA, UK",,20%
100 96,Oracle,4/7/22,600,Hardware,"London SW1A 0AA, UK",,20%
0 61,Alice Smith,3/3/22,3000,Catering,"123 Main St, San Francisco, CA 94105",[phone],
0 81,Alice Smith,3/23/22,3000,Catering,"123 Main St, San Francisco, CA 94105",[phone],
0 11,Bob Jones,1/12/22,2900,Labor,"456 Pine St, Los Angeles, CA 90012",[phone],
0 65,Emily Green,3/7/22,2700,Labor,"1213 Sunset Blvd, Los Angeles, CA 90026",[phone],
0 85,Emily Green,3/27/22,2700,Labor,"1213 Sunset Blvd, Los Angeles, CA 90026",[phone],
0 9,Grace Liu,1/10/22,2600,Labor,"1617 Mission St, San Francisco, CA 94110",[phone],
0 29,Grace Liu,1/30/22,2600,Labor,"1617 Mission St, San Francisco, CA 94110",[phone],
0 17,Irene Miller,1/18/22,2700,Labor,"2021 Collins Ave, Miami, FL 33140",[phone],
0 37,Irene Miller,2/7/22,2700,Labor,"2021 Collins Ave, Miami, FL 33140",[phone],
100 76,Hilton,3/18/22,200,Travel,,,
112 20,Hyatt,1/21/22,400,Travel,,,
104 50,Hyatt,2/20/22,400,Travel,,,
112 80,Hyatt,3/22/22,400,Travel,,,
112 48,Marriott,2/18/22,300,Travel,,,
108 78,Marriott,3/20/22,300,Travel,,,

//////////

# BAD PROMPTS

Certainly! Crafting an effective prompt is key to obtaining relevant and quality answers. Here are some examples of "bad" prompts juxtaposed with improved "good" versions to demonstrate the difference:

1. **Bad:** "Stuff about the French Revolution?"
**Good:** "Can you provide a brief overview of the main events of the French Revolution?"

2. **Bad:** "Some science thing?"
**Good:** "Can you explain the concept of photosynthesis in simple terms?"

3. **Bad:** "How do those things fly?"
**Good:** "How does the aerodynamic design of an airplane allow it to fly?"

4. **Bad:** "Books?"
**Good:** "Can you recommend five classic novels that have had a significant impact on literature?"

5. **Bad:** "That math problem with the apples and stuff."
**Good:** "If John has 5 apples and gives away 2, how many apples does he have left?"

6. **Bad:** "Some famous dude in history?"
**Good:** "Can you give me a brief biography of Sir Isaac Newton and his contributions to science?"

7. **Bad:** "Help with the thing in my kitchen!"
**Good:** "I'm having trouble with my blender. It's not turning on. Do you have any troubleshooting suggestions?"

8. **Bad:** "Some movie with that actor."
**Good:** "Can you list movies that feature Leonardo DiCaprio as the main actor?"

9. **Bad:** "How do I do the stuff on my computer?"
**Good:** "How can I create a new folder on my Windows desktop?"

10. **Bad:** "Music things?"
**Good:** "Can you suggest some influential jazz musicians from the 20th century?"

By being more specific and clear in your prompts, you're more likely to get relevant and detailed answers that address what you're looking for.