Prompt Engineering
We live in a highly competitive environment and many business owners are keen to find ways in which artificial intelligence can boost productivity, reduce costs and increase profits. For example, AI can assist with drafting reports, writing proposals, social media posts and emails, or researching and planning marketing strategies. Some organisations are developing their own AI business strategy, but many people feel intimidated by the terminology used in relation to AI and assume that in depth knowledge of the technology is required to integrate AI with their business processes. Below is a description of prompt engineering, advice on how to write AI prompts and 20 prompt engineering examples, some of which you might find useful.
Please contact me for a quote to write content for your business.
What Is Prompt Engineering?
The term prompt engineering is used to refer to writing prompts which when entered into a generative artificial intelligence tool such as ChatGPT or Google Gemini will output text, image or audio content. Essentially this involves crafting the best possible text prompt to get the AI to output the optimum answer to a question. This is a skill that can be developed by practicing with different AI tools and learning how use of language and refining questions can deliver good results.
Some companies are offering large salaries to people with qualifications, skills and experience they consider essential to excel at prompt engineering, including computer science and coding in languages such as Python. Consequently SMEs might assume they lack the resources needed to attract the talent that will enable them to compete effectively. They could fear being left behind as AI is widely adopted. However, SMEs might have the potential to not only compete against but outperform larger businesses in their sector. For example, by training their existing personnel.
Although this article uses the term ‘prompt engineering’, it might be better to describe it as ‘prompt writing’. Use of the word engineering can be misleading and cause businesses to focus too much on technical expertise rather than the domain knowledge and experience that is relevant to the products or services that they sell. For example, when using prompt engineering in marketing for a restaurant, a food expert is more likely to get results than a software engineer who lacks domain knowledge of the restaurant industry. Similarly when writing creativity prompts for design thinking, individuals with relevant skills will be able to write more effective prompts for generative AI.
When a business plans to develop a custom software application, it will be necessary to employ the services of people with relevant skills and experience. For example, AI platforms including ChatGPT and Google Gemini, provide an application programming interface (API), which can be used to develop custom applications powered by AI. SMEs typically lack such in-house resources, but increasingly AI is being integrated into applications business owners and their employees will be familiar with, such as Microsoft 365 Copilot.
How To Write AI Prompts
A logical mind, good writing skills and domain knowledge are valuable assets in the writing of effective prompts. Understanding the nuances, strengths and weaknesses of different generative AI tools and LLMs will help users to select the best platform for each task. Business owners and their employees will be familiar with terminology used in their sector, which will help them to write suitable prompts. They will also be better able to recognise the relevance and accuracy of AI output, correct errors and ensure content is appropriate for the intended audience.
Although coding skills are not essential to understanding how to learn prompt engineering, some basic technical knowledge can be useful. For example, awareness of issues around data security and terminology frequently used in AI. Writing concise AI prompts can also benefit from a basic understanding of conditional statements used in programming, such as if-then and if-else. Boolean logic, is a type of algebra that uses three basic operators: AND, OR, NOT to determine if a thing is true or false. Comparison operators, such as equal to, less than or greater than, compare values.
When attempting to formulate a good prompt you should have a clear idea of what you want to achieve. For example, prompt engineering for business use, such as project planning or researching a particular topic. You could begin by asking generative AI to suggest suitable prompts, which might help you to get the results you are looking for. It is likely that you will spend some time refining your prompts and you might find that some unexpected answers prove useful. As your prompt writing skills improve you might begin to build up a range of effective prompts that you can share with colleagues in a knowledge base.
Writing efficient prompts can reduce the time required to get useful results and save money by reducing the usage of tokens, which is also more sustainable as it consumes less electricity. Rather than a specialist skill, writing effective AI prompts is likely to become something that most people will learn, as it can help them to succeed in their role. However, just as with other software tools, such as spreadsheets and word processors, those who learn more advanced skills will be able to achieve more.
The structure and quality of prompts will determine the output, as referred to in the phrase ‘garbage in, garbage out’. If prompts are vague the output will likely lack relevance, be too generic, or fail to align with business goals. In addition to clarifying the scope, depth, purpose and target audience of the output, a prompt can request details of factors that influenced conclusions and recommendations. This might inspire solutions or innovations that promote growth, while analysis of big datasets could provide insights that assist decision making.
The Art Of AI Prompts
A musician who develops an understanding of musical theory and a feel for their instrument, can apply this in their process that draws upon their creative intelligence. Writing effective prompts is a skill, with theory that can be understood and techniques that can be learned. However, it could also perhaps be considered an art form, in which the AI will respond more effectively to prompts that are written with a feel for the nuances of the AI platform.
LLMs recognise patterns and relationships among the datasets upon which they are trained. In response to a text prompt, the AI model generates answers that most closely correspond to what it predicts will come next in a sequence. The phrasing, structure and content of a prompt will determine the output. Rather than asking simple questions, the art of prompt writing involves guiding the AI towards the desired output. The audience for the required output could also be specified, to ensure the depth of complexity and language used are appropriate.
20 Prompt Engineering Examples
Writing effective AI prompts is an iterative process. It will likely require you to experiment with and refine the structure and phrasing used. This could involve a conversational chain, in with each prompt improves upon the previous one. The resulting content could provide a good starting point, but will probably require additional work to ensure it is error free and meets specific requirements. These prompt engineering examples might inspire you. They reflect the wide range of possible prompt engineering use cases.
Instructional prompts provide explicit instructions. Example: ‘Provide a step-by-step explanation of the most efficient method that will solve the following mathematical problem.’
Priming prompts include examples that reflect the desired output from a particular input, in order to get the AI to generate something similar. Example: ‘The generated output should be similar in style, tone and content to these examples listed below.’
Socratic prompts can instruct an AI model to explain its reasoning process. Example: ‘Describe the advantages and disadvantages of employing the following strategy to resolve the following issue, which includes your reasoning process and describes sources of verified data.’
Mixed prompts include a combination of different prompt types. Example ‘Based upon the previous output (context), describe advantages and disadvantages of using this technology (instruction) and provide relevant supporting examples to explain your decisions, with evidence and sources (Socratic).’
Generating Code Example: ‘Write the HTML CSS, and JavaScript code for a fully responsive web page featuring a unit converter to convert between metric and imperial values of weight, length, volume and temperature. Put the JavaScript and CSS in external files, to ease maintenance and updating. Include comments explaining how the code works.’
Data Research and Analysis Example: ‘Generate a CSV file of the ten largest businesses selling (product or service) in the (region), which includes available data to populate the following fields of data (data required) within each record. The CSV file will be opened in an MS Excel spreadsheet to display data in a table, requiring suitable record and field names.’ (*CSV = Comma Separated Values)
Rules and Regulations Example: ‘Detail requirements for (project description) at (location) using (product) on (dates and times). Based upon the previous conversation and information regarding our business, highlight key issues and recommend any actions that should be taken to ensure compliance with rules and regulations.’
Strategy Planning Example: ‘Generate a SWOT analysis for a UK based SaaS startup selling cloud services to SMEs that has the following characteristics:’
Report Summary Example: ‘Generate a summary, between 80 and 100 words in length, for a sales pitch suitable for a non-technical audience that highlights key information from the following report:’
Social Media Posts Example: ‘Create a series of engaging marketing messages for (product or service) that will appeal to (customer persona) suitable for the following social media platforms:’
Email Responses Example: ‘Generate a professional response, highlighting key features of (product or service) detailed in the report (report name) summarised previously, that answers the questions in the following client email:’
Social Media Research Example: ‘Generate a report of the most highly commented social media content found online regarding (company name products or services), along with the most frequent positive and negative reviews.’
Market Research Example: Generate a market research plan for a business selling (product or service), identifying key data required. Segment customers based upon their demographic characteristics and suggest strategies that are most likely to persuade them to complete a purchase.
Market Forecasting Example: ‘Estimate the current and potential future UK market size for (product or service) and include a breakdown of relevant data upon which future predictions are based.’
Competitor Analysis Example: Generate a SWOT analysis of current market leaders selling (product or service) and compare the performance of each company. Identify key areas for improvement and generate a marketing and sales plan that should maximise customer engagement, leads and sales.
Consumer Behaviour Example: Analyse available online data regarding buying preferences for the following customer persona types (description). Based upon these findings, suggest a marketing strategy that will appeal to each persona and increase the probability they will complete a purchase.
Event Planning Example: Find a list of suitable venues for a conference to be held in (city name) on (dates) to be attended by (number and description of attendees) at which their will be discussion of (conference topic). Generate a list of suitable speakers and an event plan, with detailed itinerary for each day.
Comparing Product Suppliers Example: ‘Find the 10 best rated suppliers of (product) within (region), listing available data regarding pricing and delivery. Include links to best and worst online reviews, along with any relevant media stories with verified sources.’
Product or Service Research Example: Analyse past, present and projected future trends regarding purchase of (product or service type). Identify target audience characteristics and suggest new or updated product features likely to appeal to them. Include estimated production, storage and distribution costs.
Product or Service Design Example: Generate a list of essential features of (product or service). Create a step by step plan detailing the design process, highlighting issues to be aware of. Include key findings from relevant research data, to identify unmet niche demand and opportunities to differentiate the design from those already on the market.