Table of contents
- The Same Tool, Completely Different Results
- Why ChatGPT Gives Bad Results — The Real Reason
- Prompt Engineering for Beginners — The Four Elements
- Five Prompt Mistakes Almost Everyone Makes
- How Canadian Professionals Are Using Prompt Engineering at Work
- Why Prompt Engineering Is Now a Professional Skill — Not Just a Trick
- What Advanced Prompt Engineering Actually Involves
- A list of FAQs that you might find helpful
Why Two People Get Completely Different Results From ChatGPT and How to Be The One Who Gets It Right
The Same Tool, Completely Different Results
Two individuals at the same Toronto firm. With the same subscription and AI tool. In the same week. One person comes out disappointed, as the outcome was generic and unhelpful. The other emerges with a brilliant draft in hand, as well as competitor analysis and email to the client all ready.
As for the AI, it didn’t treat them any differently, nor did the tool work for one and fail with the other. What made a difference between the results is the communication with it.
This is a key essence of prompt engineering — that is the ability to write instructions that help AI systems generate what you actually need. It appears to be a very useful skill for a specialist, regardless of his/her expertise.
It is interesting to know that prompt engineering is said to be one of the fastest-growing skills on LinkedIn. Employers in Canada in various industries like finance, healthcare, marketing and operations are hunting for specialists who can work with AI efficiently, not only use it.
Why ChatGPT Gives Bad Results — The Real Reason
Before we discuss writing good AI prompts, we need to know why a vague prompt produces vague answers. The reason is pretty obvious.
Think of a situation when you have hired a really smart employee. He can do a lot of things, has lot of knowledge about the subject and is willing to help you. On the first day, you tell him to “write something on marketing”.
He is looking at you and then produces some normal piece of writing. This is happening because the instructions you gave were vague and did not explain anything, so he had to write something supposed to be “ok”.
Then think of the situation when you told your employee: “You are a senior marketing strategist. You need to write a 200-word LinkedIn post for our company based in Vancouver and providing SaaS services to HR directors. The tone of writing should be professional but informal.” In this case, he knows what exactly he needs to do and does everything properly.
The same worker. Entirely different results. ChatGPT works the same way. The clearer you state your needs — to whom you are speaking, which context is important, how the output should look like — the better the outcome. Vague input gets vague output and vice versa.
Prompt Engineering for Beginners — The Four Elements
In any useful prompt—whether it involves email composition, document evaluation, or report writing—one must incorporate at least four essential components. Although not all four components are vital, the number included must be kept high enough to increase chances of ensuring the final result is what you require.
Role — Who Do You Want the AI to Be?
When you begin a prompt with a role that the AI must portray, you clearly indicate the angle the AI research must be done from. For example, “You are a senior financial advisor,” will yield significantly different results when compared to “You are a creative writing coach,” even in the case of keeping the rest of the prompt unchanged.
Example: “You are an experienced HR manager at a mid-sized Canadian firm…”
Context — What Does the AI Need to Know?
Context gives the AI all necessary information that allows the AI to provide specific answers. it is actually the part where the majority of users put the least effort. AI will not know anything about your sphere or your audience until you say it.
Example: “…We are getting ready for a performance review cycle, where a team of 12 remote workers is located in three different parts of the world…”
Task–What is the Goal of the Work?
Although it may sound obvious, ambiguous task assignments are the biggest reason why ChatGPT produces poor outcomes. “Write something” does not constitute a task. “Prepare a 150-word summary” is real task. “Analyze this text” is not a task. “Identify the three most significant risks mentioned in the text and clarify each of them in one sentence” is a task.
Example: “…Prepare a 10-question performance review template focusing on online collaboration, communication and quality of output rather than physical presence in the office.”…
Format - The Way in Which Output Should Be Structured
When you inform the AI about the expected structure of the answer, it helps avoid the need for major edits. There are various formats that you can require, including: bullet points, numbering, tables, short paragraphs, formal or conversational tone, etc. Without your direction, the machine will decide how to format the answer that may or may not fit your requirements.
Example: “…Make the output in the form of a numbered list. Limit the length of each question to 20 words at the maximum. Use the simplest words.”
The combined full prompt is as follows: “You are a knowledgeable HR manager in a medium-sized Canadian firm. We have started preparing for the performance review cycle that includes 12 virtual workers in three different time zones. Compose a 10-question template that will help evaluate performance in terms of remote cooperation, communication, and quality of work instead of presence in an office. Make a numbered list. Limit each question to 20 words.”
Five Prompt Mistakes Almost Everyone Makes
Knowing about the four elements is only part of the story. The other side is knowing what not to do, and here are five common mistakes, and why they are significant.
Mistake 1- Failing to Specify the Outcome
“Write me a report” is not a prompt; instead, it is an instruction for certain manual verticals or animates. What type of report? For who? How long? What do you want it to conclude with? The AI is going to fill in every single gap you leave with its best guess-mind you, its best guess might not necessarily match your actual need. Be clear on the outcome and not just the activity.
Mistake 2: Failing to give contextual information
Without any contextual information, AI, including technology like ChatGPT, will have no idea about your area of expertise or your audience or operations. Knowing the significance of context distinguishes good people from bad ones.
Mistake 3: Failing to name the audience
The phrase “Explain machine learning” brings out different responses depending on the audience: if you talk to a 7-year-old child, to a marketing manager or to a software engineer. You should tell the AI who it is that is being addressed to.
Mistake 4: Forgetting about Tone and Format
If you want to see the information as a bullet point list and forget to tell the AI this, you will see only paragraphs as a result. Also, if you require a formal tone, but don’t mention this, then the outcome will most probably be informal. Even though this is a fairly simple request to fulfill, so many people forget to add this and in the end, they spend a lot of time making edits on the generated text.
Mistake 5: Not Trying Multiple Times
One of the most underutilized methods of getting good results from ChatGPT is simply iterating instead of starting over when the first result isn’t perfect. If the output is not as desired, analyze what is missing and change instructions accordingly.
How Canadian Professionals Are Using Prompt Engineering at Work
This skill is not confined to the tech industry. In Canada, individuals working in finance, healthcare, marketing, and operations are applying prompt engineering to achieve optimal results with existing AI tools.
Finance – Toronto
A financial analyst employed at a medium-sized investment company employs a systematic approach by using a pre-defined prompting method for creating customer reports. This method comprises all important details such as her own job title, the type of risk profile of the client, the relevant legal documents necessary for any investor in Canada, requirements of format style and tone. As a result, she completed her reports in forty minutes instead of four hours previously taken without such an approach.
Marketing – Vancouver
A marketing manager working for a startup in Vancouver utilizes a technology to develop initial drafts for the campaign text, social media materials, and e-mail campaigns. The process works quite well because she uses a special context block, which consists of a paragraph about the company and target audience and the tone of the company.
Human Resources – Calgary
The HR employee from an energy company in Calgary managed to create job descriptions, testing questions, and on-boarding info using the prompts. In her prompts, she identifies the level of the position in the company, the necessary legislative clauses, and the company’s values and principles as well as level of literacy for readers.
Want to go beyond the basics and master prompt engineering properly?
Proleed Academy’s AI programme covers prompt engineering in depth — zero-shot, few-shot, chain-of-thought, ReAct, and advanced prompt design for real business applications. Live instructor-led sessions, globally recognised certification.
Why Prompt Engineering Is Now a Professional Skill — Not Just a Trick
The attitude toward prompt engineering is still somewhat negative — like something trendy you heard of reading a list of articles and tried once. However, the change to this attitude is seen rapidly.
As the popularity of prompt engineering continues to grow, it can be found more and more in Canadian job vacancies from various industries that have nothing in common with programming or software development. Professions related to marketing, analytics, human resources, operations, listed in their requirements “AI proficiency” and “knowledge of prompt engineering.”
The matter is simple: organizations are using more and more of the AI technologies and their problem now is not the availability of technologies. The problem is the capability to use them. And prompt engineering is the way to solve the problem of the effective implementation of AI technology.
A specialist with the knowledge of prompt engineering is more productive than his or her counterparts working without this knowledge. The professional with this knowledge becomes the specialist whom other employees ask for help when their completed tasks go wrong and it is a good position to occupy in the times of quick adoption of AI technologies in Canada.
What Advanced Prompt Engineering Actually Involves
All that was discussed above- the role, context, task and format constitutes what is known as the foundation. This is able to work for the majority of people bringing a significant improvement in their experience with ChatGPT and any other AI tools available today.
However the prompt engineering does go way deeper than that especially if it comes to any professional involved in AI system building.
Use of Chain-Of-Thought Prompting
The essence of this technique is that instead of just making the AI come to the conclusion you require you ask it to show how it has arrived at this conclusion through reasoning, making the model do so is likely to lead to the more effective response it is capable of generating.
Prompting Through Few-Shot
In most cases, through few-shot prompting, you give two or three prompts to the AI so that it knows what you want. The AI then tries to reproduce the output on the given examples for different incoming data. This is how organizations manage to get the work done with AI assistants without having to carry out numerous revisions.
ReAct Prompting
ReAct is the abbreviation of reasoning and acting which means that AI is able to think about the action it has to perform, take the action, see the outcome, then think about the next step. This principle of operation is more useful in case of many-step tasks.
System Prompts
In commercial applications of AI, system prompts are elements of control which are usually used in the background and define the AI’s general outlook. It is quite important to work properly with the prompts as it is a really important part of using AI.
These techniques are covered in Proleed Academy’s AI program — moving from foundational prompt design through to advanced techniques used in production systems. The curriculum covers each one with live instruction, real examples, and hands-on practice rather than just theoretical explanation.
Related read: Understand how prompt engineering connects to building full AI agent systems — What Are AI Agents — And How Are They Already Changing the Way People Work?
Related read: See why even well-prompted AI systems can still give wrong answers — and how RAG fixes it — Why AI Chatbots Give Wrong Answers — And How RAG Systems Fix That
A list of FAQs that you might find helpful
Are you ready to master prompt engineering?
Proleed Academy’s AI program encompasses all aspects of prompt engineering, including theories, principles, and advanced systems in production AI applications. Module XXIII leads you from an understanding of prompts through their application in real projects.
