AI development techniques

Artificial intelligence refers to computer programs that can complete cognitive tasks
typically associated with human intelligence. There are two main techniques used to design
AI programs:

 Rule-based techniques involve creating AI programs that strictly follow predefined

rules to make decisions. For example, a spam filter using rule-based techniques might
block emails that contain specific keywords using its predefined logic.
 Machine learning techniques involve creating AI programs that can analyze and
learn from patterns in data to make independent decisions. For example, a spam filter
using these techniques might flag potential spam for the recipient to review,
preventing automatic blocking. If the recipient marks emails from trusted sources as
safe, the spam filter learns and adapts its logic to include similar emails from that
sender in the future.

AI tools can use either rule-based or ML techniques, or even a combination of both. In

general, rule-based techniques are commonly used for tasks that require rigidity, such as
blocking messages from untrusted senders that are obviously spam, like requests for bank
transfers or private information. Conversely, ML techniques are better suited for tasks
demanding flexibility and adaptability, like learning to recognize that messages from trusted
senders containing typos are not spam.

Approaches to training ML programs

Recall that machine learning is a subset of AI focused on developing computer programs
that can analyze data to make decisions or predictions. AI designers often use ML in their AI
programs because it doesn’t have the limitations of rule-based techniques.

There are three common approaches to training ML programs:

 Supervised learning
  Unsupervised learning
 Reinforcement learning

Supervised learning

In this approach, the ML program learns from a labeled training set. A labeled training set
includes data that is labeled or tagged, which provides context and meaning to the data. For
instance, an email spam filter that's trained with supervised learning would use a training set
of emails that are labeled as “spam” or “not spam.” Supervised learning is often used when
there's a specific output in mind.

Unsupervised learning

In this approach, the ML program learns from an unlabeled training set. An unlabeled
training set includes data that does not have labels or tags. For instance, ML might be used to
analyze a dataset of unsorted email messages and find patterns in topics, keywords, or
contacts. In other words, unsupervised learning is used to identify patterns in data without a
specific output in mind.

Reinforcement learning

In this approach, the ML program uses trial-and-error to learn which actions lead to the best
outcome. The program learns to do this by getting rewarded for making good choices that
lead to the desired results. Reinforcement learning is commonly used by conversational AI
tools. As these tools receive feedback from users and AI designers, they learn to generate
effective responses.

Each ML technique has its own strengths and weaknesses. Depending on the type of data
that's available and what's needed to solve the particular problem, AI designers may use one,
two, or all three of these techniques to produce an AI-powered solution.

Generative AI
Advancements in machine learning have helped pave the way for generative AI—AI that
can generate new content, like text, images, or other media. This type of AI often uses a
combination of supervised, unsupervised, and reinforcement learning to create original
content.

For instance, all three approaches play distinct roles in conversational AI tools. Supervised
learning equips conversational AI tools with foundational dialogue data, enabling them to
respond to common conversational cues appropriately. Unsupervised learning enables them
to interpret nuances in language, like colloquialisms, that occur naturally in conversation.
Reinforcement learning further strengthens these tools by allowing them to improve their
responses in real-time based on user feedback. This enables them to adapt to the
conversational context and engage in natural conversations.

Generative AI's ability to create and innovate offers a range of benefits to all sorts of
workplaces and professions, such as marketing, product development, engineering, education,
manufacturing, and research and development. These benefits include:
  Greater efficiency: Generative AI can automate or augment routine tasks, allowing
workers to focus on other work priorities.
 Personalized experiences: Generative AI can tailor its interactions to individual
preferences and needs.
 Better decisions: Generative AI can quickly analyze vast amounts of data to uncover
useful insights.

AI tools and AI models

Terms like AI tools and AI models can be confusing because they sound similar but refer to
different things. Recall that an AI tool is AI-powered software that can automate or assist
users with a variety of tasks. An AI model is a computer program trained on sets of data to
recognize patterns and perform specific tasks.

To better understand the relationship between these two concepts, consider a car and its
engine.

 The car: An AI tool, like a car, gets you to a “destination,” such as a completed task
or an output. And AI designers and engineers, just like auto engineers, add various
features and controls into AI tools to provide a user-friendly experience.
 The engine: An AI model is the underlying component that makes the “car” run. It's
under the hood, you might say, processing user input and allowing you to drive the
car.

Cars come in various shapes and sizes, and people choose the vehicle they drive based on
their needs, like a truck to transport materials or a bus to transport riders. In a similar way, AI
tools are developed for a wide range of applications, like generating text, image, or code. And
whatever the tool’s specific function is, it's powered by an AI model.

Note: AI tools sometimes use multiple AI models in order to have more flexibility and
perform a wider range of tasks. You'll explore these types of tools later in this course.
The process of training AI models
AI designers and engineers develop AI models through a process called training. Here’s an
example of the typical steps a designer might take in this process, in this case for building a
model that predicts rainfall:

1. Define the problem to be solved. AI designers and engineers want to predict rain to
help people stay dry when commuting to and from work. They start by considering
AI’s capabilities and limitations before identifying an AI solution.
2. Collect relevant data to train the model. AI designers and engineers gather
historical data of days when it rained and days when it didn't rain over the past 50
years.
3. Prepare the data for training. AI designers and engineers prepare the data by
labeling important features, such as outdoor temperature, humidity, and air pressure,
and then noting whether it rained. It's also common to separate the data into two
distinct sets: a training set and a validation set to test with later.
4. Train the model. AI designers and engineers apply machine learning (ML) programs
to their rain prediction model, which helps it recognize patterns in its training data
that indicate the likelihood of rainfall. Those patterns might include high
temperatures, low air pressure, and high humidity.
5. Evaluate the model. AI designers and engineers use the validation set they prepared
earlier to assess their model's ability to predict rainfall accurately and reliably.
Analyzing a model's performance can uncover potential issues impacting the model,
such as insufficient or biased training data. If any issues exist, the AI designers and
engineers may revisit an earlier step in this process to try a different approach. Once
the model performs well with its validation set, the process continues to the next step.
6. Deploy the model. When the AI designers and engineers are satisfied with their
model's performance, they deploy it in an AI tool—helping people in their city stay
dry on their way to work!

Model training is an iterative process. AI designers and engineers can repeat each step as
many times as necessary and make adjustments until they create the best model possible.

But the process doesn't stop at deployment. Once users interact with a model in practical
situations, the model might be exposed to new challenges. AI designers and engineers should
continuously monitor and collect feedback on their models, ensuring their models continue to
perform reliably and to identify areas for improvement. It's this iterative process of continual
refinement that makes AI models precise and versatile, which ultimately leads to effective,
reliable AI tools. When you understand how AI models are developed, you can make
informed decisions about when and how to use an AI tool to accomplish your goals.

Workplace uses for generative AI

The generative AI tools that you’ll use will depend on your industry, role, and specific
workplace needs. Please note that each tool’s uses are not limited to the industries listed here.
Also note that the cost varies for each tool. Some tools in this list are free or offer a free trial;
others require a paid subscription to access. The examples that follow include whether the
tool is stand-alone, integrated, or both. Stand-alone means it’s an individual organization and
tool that you can add and use in your workflow. Integrated means the tool is part of an
existing type of software, which you’ll need to access if you want to use that AI feature set.

Conversational AI tools
Conversational AI includes general-purpose tools that can simulate a human conversation, as
well as provide answers to questions on a wide variety of subjects. Workers might use
conversational AI tools to help with work tasks, such as brainstorming or finding answers to
low-stakes questions.

Example industries: Human resources, marketing, public relations, sales, education, project
management, retail, copywriting, creative writing, product management

Example tools include:

 Anthropic Claude
o Description: Anthropic Claude can complete problem-solving tasks, like
finding mathematical solutions, translating between languages, and
summarizing long documents.
o Stand-alone or integrated: Stand-alone
 Gemini
o Description: Supercharge your creativity and productivity with Gemini. Chat
to start writing, planning, learning and more with Google AI.
o Stand-alone or integrated: Both
 Microsoft Copilot
o Description: Integrated with Microsoft Edge, Microsoft Copilot can help with
online searches to find information, compare products, and summarize web
page content.
o Stand-alone or integrated: Both
  ChatGPT
o Description: ChatGPT can generate ideas, plan schedules, debug code, and
proofread text.
o Stand-alone or integrated: Stand-alone

Productivity and writing assistants

AI productivity and writing assistants can help with workplace tasks. They might provide
grammar or spelling suggestions, generate a summary of a long document, or solve problems.
Here are some examples:

 Clockwise
o Description: Clockwise is a calendar tool that learns users’ work habits to
automatically schedule and manage calendar events.
o Example industries: Consulting, technology, sales
o Stand-alone or integrated: Stand-alone
 Grammarly
o Description: Grammarly is a writing assistant that can help users edit and
write clear, concise text.
o Example industries: Creative writing, education, marketing
o Stand-alone or integrated: Stand-alone
 Jasper
o Description: Jasper is a writing assistant intended for marketing tasks, like
drafting social media posts, emails, and landing page content.
o Example industries: Copywriting, marketing, sales
o Stand-alone or integrated: Stand-alone
 NotebookLM
o Description: NotebookLM integrates into document apps, like Google Docs,
and helps summarize or ask specific questions about text, notes, and sources.
o Example industries: Content writing, finance, sales
o Stand-alone or integrated: Both
 Notion AI
o Description: Notion AI is a writing assistant built into Notion, a productivity
and note-taking software tool.
o Example industries: Development, marketing, product management, sales
o Stand-alone or integrated: Integrated
 AI by Zapier
o Description: AI by Zapier is a built-in productivity tool that allows AI
automation to be integrated with the apps and workflows already connected
through Zapier.
o Example industries: Engineering, marketing, project management,
technology
o Stand-alone or integrated: Integrated

Code-generative AI tools
Code-generating tools can help generate, edit, or complete code for a variety of programming
tasks in many different programming languages. Examples include:
  Android Studio Bot
o Description: Built into Android Studio, Studio Bot can generate code and
answer questions about Android development.
o Example industries: Data science, software development, web development
o Stand-alone or integrated: Integrated
 GitHub Copilot
o Description: Built into GitHub, Copilot can write and suggest code, suggest
descriptions for pull requests, translate multiple languages into code, and
index repositories.
o Example industries: Data science, software development, web development
o Stand-alone or integrated: Both
 Replit AI
o Description: This tool, built into Replit, is a cloud-based Integrated
Development Environment (IDE) for programmers that can make suggestions,
help explain code, and turn natural language into code.
o Example industries: Data science, software development, web development
o Stand-alone or integrated: Integrated
 Tabnine
o Description: Tabnine can be a plugin to many popular code editors to help
speed up delivery and keep code safe.
o Example industries: Data science, software development, web development
o Stand-alone or integrated: Stand-alone
 Jupyter AI
o Description: Jupyter is an open-source platform for coding, and this built-in
tool includes a chat interface, which can be used to generate code, fix coding
errors, and ask questions about files.
o Example industries: Data science, software development, web development
o Stand-alone or integrated: Integrated

Image- and media-generative AI tools

Media-generating AI tools help workers with tasks like generating and editing images, video,
and speech. Examples include:

 Adobe Firefly
o Description: Built into the Adobe suite, Firefly can generate and edit images.
o Example industries: Design, education, marketing
o Stand-alone or integrated: Integrated
 Canva Magic Design™
o Description: Canva Magic Design is a tool that generates text and image
content in Canva, an online graphic design tool.
o Example industries: Design, education, marketing
o Stand-alone or integrated: Integrated
 DALL-E
o Description: Integrated with ChatGPT, DALL-E generates images from text
prompts.
o Example industries: Design, education, marketing
o Stand-alone or integrated: Integrated
 ElevenLabs
 o Description: ElevenLabs is a speech AI tool that can generate spoken voice-
over audio from text in different languages.
o Example industries: Content creation, education, marketing, production
o Stand-alone or integrated: Stand-alone
 Google Ads
o Description: Google Ads helps businesses reach customers around the world,
driving growth and performance. Google Ads makes it easy to create
campaigns, measure impact and improve your results. Put Google AI to work
for your business with the Google Ads AI Essentials. Learn more with the AI
Explored video series.
o Example industries: Marketing, Advertising
o Stand-alone or integrated: Integrated
 Midjourney
o Description: Integrated into Discord, Midjourney can generate images from
text prompts.
o Example industries: Design, education, marketing
o Stand-alone or integrated: Integrated
 Runway
o Description: Runway can generate a new video from a text prompt or edit an
existing video’s style or focus area, and remove people or other elements.
o Example industries: Content creation, design, marketing, production
o Stand-alone or integrated: Stand-alone

Prompts for different purposes

Recall that a large language model, or LLM, is an AI model that is trained on large amounts
of text to identify patterns between words, concepts, and phrases so that it can generate
responses to prompts. As you’ve been learning, good prompt engineering can help guide an
LLM to generate useful output for workplace tasks. In this reading, you’ll further explore
how to write clear and specific prompts for a variety of workplace use cases.

alt=""

Use cases
As you explored previously, you might use an LLM at work to help boost your productivity
and creativity and complete any of these useful tasks:

 Content creation

 Summarization

 Classification

 Extraction
  Translation

 Editing

 Problem-solving

Note: The following examples illustrate best practices; they aren’t exact templates to copy
for every situation. Your results will vary based on a number of factors, including the specific
LLM you’re using. Remember to critically evaluate all LLM output and to iterate on your
initial prompt to get the most useful output.

In general, here’s how to make your prompts more effective:

 Consider what you want the LLM to produce. The LLM will generate more useful
output when you include a specific instruction in your prompt, like create, summarize,
classify, extract, translate, edit, or solve.

 Provide necessary context. The LLM will generate more useful output when you
include detailed instructions, with specific guidance about the style or format of the
output you want.

Up next, you’ll examine each of the use cases described earlier by considering an additional
example for each.

Content creation

No matter your industry, an LLM can help you create content for a variety of purposes, such
as blog posts, reports, product descriptions, and taglines. For example, suppose you’re
working on an ad campaign for a new line of home appliances. You can ask an LLM to help
you create an engaging tagline for one of the products:
Act like you are a creative advertising professional who can apply innovative thinking to
develop original taglines that project the positive qualities of a product. Create a concise
tagline for a washing machine that gets clothes extra clean, has 25 settings, and fits in a
small space.

The prompt begins by describing the LLM’s role as a creative advertising executive. Next,
the prompt clearly states that the task is to create a concise tagline for a washing machine.
Finally, the prompt specifies the product features to include in the tagline.

Pro tip: Assign the LLM a role, job, or function to reinforce the purpose of the prompt and
help guide the LLM to produce useful output.

Summarization

An LLM can help you summarize many types of texts: reports, customer surveys, meeting
notes, emails, and more. For example, the following prompt asks an LLM to provide a
summary of a lengthy email:

The following text is an email from a software vendor. Summarize its main points in a
bulleted list:

"I hope this finds you well. It was a pleasure to chat with you at the conference last week.

Following up with more detail on our pricing plans. Our bronze level subscription gets you
access to three of our most popular software products as well as training videos for these
products. If you have additional software needs, you can subscribe at the silver level. This
level allows you to choose two additional software products, with training videos on those
products, as well as gets you 24-hour support on any difficulties you encounter while using
the products. Finally, our gold level membership gets you access to all ten of our software
products. You get training for all ten products as well as 24-hour support, and you are also
the first to enjoy any beta additions to our products.

Please contact me for the pricing of the level that interests you. We offer monthly
subscriptions and a reduced rate for a yearly subscription."

The prompt begins with useful context about the relevant email. Next, it clearly states that the
task is to summarize the main points of the email. Finally, it specifies that the output should
be formatted as a bulleted list.

Note: Be aware that LLMs can sometimes hallucinate, or produce AI outputs that aren’t true.
In this case, the LLM might add details to the summary that aren’t included in the source
email. Always evaluate LLM output for accuracy before using it.

Classification

Text classification is another common workplace application for LLMs. An LLM can help
you sort customer service emails into categories based on the content of the email, categorize
content in social media posts, and analyze the sentiment or feeling of customer feedback. The
following prompt asks an LLM to analyze the sentiment in a customer review:
Read these customer reviews and tell me whether the sentiment of the reviews is positive,
negative, or neutral.

Customer Review: I don't know where to begin. We had reservations for 7:00 but they seated
us at 7:45. Then, no one came to our table for at least 30 minutes. Our appetizer and main
course were mediocre. I did love the dessert, but that wasn't enough to change our
experience.

Customer Review: I love this restaurant. The food is delicious and the service is excellent.

The prompt begins by clearly stating that the task is to analyze the sentiment of a customer
review and then specifies the options: positive, negative, or neutral. Then, the prompt
includes the relevant reviews under the label “Customer Review.”

Extraction

You can also use an LLM to pull data from text and transform it into a structured format
that’s easier to understand, known as extraction. For example, this prompt asks an LLM to
review a blog post and extract information about products mentioned in the post.

Read the blog post below and extract all of the references to items of clothing I can buy and
how much each item costs. Create a bulleted list of just these items.

Blog post: Hey everybody, I want to share what I’m wearing on campus this fall. If I’m going
out for the evening, I prefer the raw selvedge denim jeans ($150) paired with the cashmere
crew neck sweater ($250). For a more casual look, I like the fleece hoodie ($99) and fleece
sweatpants ($129). I also love every color of the striped socks ($15). They pair well with both
the jeans and the sweats.

The prompt begins by clearly stating that the task is to extract all the items of clothing
mentioned in the blog with their corresponding prices. Next, the prompt specifies that the
format of the output should be a bulleted list of the items. Finally, the prompt includes the
relevant blog.

Translation

You can leverage an LLM to translate text between different languages very quickly. For
example, the following prompt asks an LLM to help translate product descriptions from
English to Spanish:

Translate our product descriptions from English to Spanish. Maintain the same structure and
casual tone that is used in the English version in the Spanish translation.

Bicycle: Whether you’re exploring city streets or forest paths, our sleek and durable bicycle
has it all.

Rollerblades: Roll into summer in style with our smooth and stylish rollerblades.

The prompt begins by clearly stating that the task is to translate product descriptions from
English to Spanish. It also specifies that the Spanish translations should maintain a similar
structure and tone as the English originals. Finally, each example contains a label that
introduces the product description: “Bicycle" and “Rollerblades.” This format indicates that
the LLM should present the output in a similar form.

Note: As a best practice, confirm that an LLM’s translations are accurate by cross-checking
with another translation tool.

Editing

You can also use an LLM to edit and rewrite text. The LLM can help change the tone of the
text from formal to casual, or complete a grammar check. For example, the following prompt
asks an LLM to help edit a technical report so that it’s less jargony and easier for
stakeholders to understand:

Edit the language of the following paragraph so that it's easy for a general audience to
understand it. Use simpler vocabulary and grammatical structures but maintain the same
ideas.

Site selection for expansion is a complex and multifaceted process. The west side site offers
several advantages, including zoning for industrial use and direct access to both a major
highway and railroad. However, the site is also located in a jurisdiction with a complex and
time-consuming permitting process, and its distance from residential zones may necessitate
higher wages to attract workers.

The prompt begins by clearly stating that the task is to edit the text to make it easier for a
general audience to understand. Then the prompt specifies that the text’s vocabulary and
grammar should be simplified while its main content should remain the same. Finally, it
includes the relevant paragraph.

Problem-solving

One more use case is problem-solving. You can use an LLM to generate solutions for a
variety of workplace challenges, from analyzing sales data to planning an event. For example,
the following prompt asks an LLM to help organize a program for a nonprofit organization:

We are running a community program to teach children gardening skills. The program runs
from June 1 to August 15. We want the children to be able to grow plants that will be ready
for harvest by the time the program ends. First, identify a list of 10 plants that can be planted
and grown in that time period. Include sources that support the time to harvest for each
plant.

We want the children to grow three plants. These plants should be as different from each
other as possible. So next, choose three plants from the list that will provide the children with
this variety.

The prompt begins with useful context about the program, such as its main purpose and
timeline. Next, the prompt breaks the problem down into two main steps: First, identify a list
of 10 plants that fit the timeline, and second, choose three plants from the list that are unlike
each other. The prompt also asks the LLM to include sources for the list of 10 plants. Asking
the LLM to cite its sources in the output helps you verify the accuracy of the information
used to solve the problem.

Pro tip: Break a problem down into steps to help the LLM process the request and improve
the overall accuracy of the output.
Mark as completed

Explore chain-of-thought prompting

As you’ve learned, there are prompting techniques that can guide a large language model
(LLM) on how to complete tasks. Few-shot prompting is a technique that provides two or
more examples in a prompt. And one-shot prompting is a technique that provides a single
example in a prompt.

Chain-of-thought prompting, introduced in this reading, is a third technique you can use to
help a conversational AI tool solve a problem involving step-by-step reasoning. You’ll first
learn about the main workplace applications of chain-of-thought prompting, and how it can
improve the overall quality of LLM output. Then, you’ll review a detailed example of how to
write a prompt using this technique.

alt=""

Overview
Chain-of-thought prompting is a technique that involves requesting a large language model
to explain its reasoning processes. Chain-of-thought prompting is useful for solving problems
that involve step-by-step reasoning. This technique improves the quality of an LLM’s
answers in certain cases.

Benefits

Chain-of-thought prompting has two main benefits:

1. It can improve the overall accuracy of an LLM’s output. When you divide a task into
more manageable steps, you help the LLM produce accurate and consistent results.

2. Instructing an LLM to break down the problem helps you understand the intermediate
steps used to arrive at the solution. You can then evaluate the output more easily by
reviewing each step to identify possible errors or miscalculations.

Applications

Chain-of-thought prompting is useful for solving problems that involve mathematical or

logical reasoning. For example, you might use chain-of-thought prompting when you make
purchasing decisions, analyze sales data, or recommend products based on customer
requirements.

Prompt design

Chain-of-thought prompts often include one or more examples that demonstrate how to solve
a problem in discrete steps.

Suppose you want to use an LLM to help manage your organization’s annual budget. Begin
your prompt with an example of the step-by-step process your team used to calculate
expenditures for last year's budget. Then, instruct the LLM to calculate expenditures for this
year's budget in the same manner.

Example: Create a purchasing code

Here’s an example of how to design a chain-of-thought prompt for a task at work. Suppose
that an organization with thousands of employees assigns purchasing codes that its employees
use when buying supplies or equipment. A technical support specialist needs to create a
unique purchasing code for each employee. The specialist uses an internal conversational AI
tool, approved for use with company and employee information, to help them create the
purchasing code with the following chain-of-thought prompt:

Our organization assigns purchasing codes by combining an employee's department and ID

number. All alphabetic characters are lowercase in the purchasing code. Review the
examples and then answer the question that follows in the same manner. Explain the steps
involved in determining each employee's purchasing code.

Q: Tiana B works in the Marketing department and has an ID number of 9283. What is Tiana
B's purchasing code?

A: The purchasing code for Tiana B is marketing9283. To determine this, first combine the
department (Marketing) with the ID number (9283). This results in Marketing9283. Then,
change all alphabetic characters to lowercase. This creates the purchasing code
marketing9283.

Q: Sylvie E works in the Sales department and has an ID number of 2379. What is Sylvie E's
purchasing code?
A:

This prompt contains three main parts. First, it provides context. Next, it includes an example.
Third, it states a request for the LLM to answer. Let’s look more closely at each part.

Provides context

The prompt first provides useful context to solve a specific problem:

Our organization assigns purchasing codes by combining an employee's department and ID

number. All alphabetic characters are lowercase in the purchasing code. Review the example
and then answer the question that follows in the same manner. Explain the steps involved in
determining each employee's purchasing code.

The prompt describes the organization’s method for creating a purchasing code. Because this
is a chain-of-thought prompt, the prompt instructs the LLM to follow the example and to
explain the steps that determine the purchasing code.

Includes an example

The next part of the prompt includes an example of the steps used to create a purchasing
code, presented as a question and answer pair:

Q: Tiana B works in the Marketing department and has an ID number of 9283. What is Tiana
B’s purchasing code?

A: The purchasing code for Tiana B is marketing9283. To determine this, first combine the
department (Marketing) with the ID number (9283). This results in Marketing9283. Then,
change all alphabetic characters to lowercase. This creates the purchasing code
marketing9283.

Presenting the example in a question and answer format makes it easier for the LLM to
follow.

The question portion of the example includes relevant information about the employee's
department and ID number, and asks for Tiana B’s purchasing code based on this
information.

The answer portion of the example demonstrates the step-by-step reasoning that the specialist
wants the LLM to use to determine the purchasing code.

Note: You may not always be able to provide a useful example in your prompt. In that case,
simply state that you want the LLM to explain its reasoning. The quality of your results will
depend on your prompt and the specific LLM you’re using. Try including the following
language in your prompt to get the best results:

 "Solve the problem in a step-by-step manner."

 "Explain each step used to determine the answer."

States a request

Finally, the prompt includes the specific question the LLM should answer:

Q: Sylvie E works in the Sales department and has an ID number of 2379. What is Sylvie E’s
purchasing code?

A:

The question follows the pattern of the previous example to make it easier for the LLM to
provide a similar answer. The field after the label “A:” is blank to indicate that the LLM
should complete the answer.