There are many users’ guides for artificial intelligence (the one) that drive online, most of which are free, but some pay a certification rate fee.
This guide is different. Most of the incentive techniques here come from the technology companies that these assistants created in the first place: Openai, Google, Anthropic and Meta. If one knows how to promote, they are.
What is a quick one?
A quick is a requirement to do a task, written in natural language – compared to computer language – given a chatbot he or assistant like chatgt, twins, claude, microsoft copilot, meta and others.
Think of an assistant to him as a very smart and quick young employee. Can help an emails, reports and essays; Can analyze business data and create reports; Can do market research; It summarizes the meetings and automatically sends the following email; May the mind storm and think. Some assistants of he can also generate images, code, video and audio.
How well they perform depends on the quality of one’s instructions for them. Some also do not make real -time information. Ask chatbot to be provided.
Why does it stimulate the issue
The better the fast, the better the answer. Good products generate more important, accurate, penetrating, efficient and useful answers.
For businesses, a well -created email can mean the difference between a good and excellent client experience; A detailed speed in market research for a new product can help CMO make sharper decisions; And a rich summary of a plus meeting, e -mail generated can automatically help that employees do not lose the main topics and actions of the meeting.
Examples of Basic Business Inciting:
- Summarize today’s market trends.
- Enter a customer service email asking apology for a delayed order.
- Generate a five -step plan to start a new laundry cleaner.
Internal tips
Before jumping into the best practices to promote, here are internal techniques that may not be on everyone’s radar:
- Make your longest instructions: the most fruitful products on average about 21 words with the relevant context, however the requirements that people try are usually less than nine words, according to Google’s stimulating guide.
- Use action verbs, determine the length and format (for example, essay) and specify a target audience, according to Google.
- Give the context and limitations of it: Avoid getting a response that sounds like a general production generated by it, be more specific and add restrictions, according to Wharton Mollick Professor.
- Reassess speed to see if the results improve. Also, ask the aide to make the best fast, according to Google.
Guidelines for incitement
1. Give clear, direct and detailed instructions.
For example:
- Basic: “Summarize the meeting.”
- Better: “Summarize the meeting notes in a single paragraph. Then enter a list of notes of the speakers and each of their main points. Order next steps or action items suggested by speakers, if any. “
Even better, add the purpose of the task, its audience and, if applicable, how it fits into the overall workflow.
For example:
- Basic: “Craft a marketing campaign by email for new AC security software features.”
- Target: “Your task is to create a 200 -word marketing email for our q2 AC security feature issuance. Write to a target audience of medium -tech companies with 100 to 500 employees. Highlight three main features of Clouds: Crying of advanced data, inter-platform and monitoring and real-time discovery. To mention ‘new security features’.
2 Add the context, examples and approve a person
It works best when you understand the backdrop of the task. Instead of a general speed, provide a scenario.
For example:
- “Our CS team is overwhelmed with non -structured feedback. Your task is to analyze the feedback and categorize issues for our product and engineering teams. Use the following categories: UI/UX, Performance, Characteristics, Integration, Equation and Other .
- Here is an example:
- Input: The new panel is a mess! Need to charge forever to load, and I can’t find the export button. Fix this asap!
- Category: UI/UX, Performance
- Feeling: negative
- Priority: High
- Now, analyze this reaction (attach current feedback data) ”
Another example:
- “You are the general adviser to a Fortune 500 technology company. We are considering this Agreement for Licensing Software for our main data infrastructure: (attach contract)
- Analyze it for potential risks, focusing on reparation, responsibility and ownership of IP. Give your professional opinion. “
Why can he play roles to improve results? A data scientist can see different knowledge than a marketing strategist who looks at the same data.
3. Use reference texts to reduce errors
If the assistant he does not have enough information on a topic, he can halogue. Give it the reference material to improve its accuracy.
For example:
- “Use only the following data or document to answer. If the answer is not found, declare ‘I can’t find the answer. “(This is actually the essential function of a technique called the added generation of restore, Rag.)
One advice is to re-direct several times to see if you confirm the answer. This is especially important for mathematics because many are still not good in it, according to Meta, who calls this technique ‘self-consistency’.
4. Destroy the complex tasks in simpler steps
For example:
- Basic: “Analyze customer reviews and suggest improvements.”
- Better: “Step 1: Categorize reviews as positive, neutral or negative. Step 2: Identify the usual threads in negative reviews. Step 3: Recommend improvements based on findings. “
5. Tell him to think through the answers
Although new patterns of reasoning today can do so, this is still a better practice as not all assistants have reasoning skills.
For example:
- “Drafting personalized emails for donors seeking contributions to the children’s care program this year.
- Program Information: (Add Program Details)
- Donor Information: (Add donor details)
- Think before you write the email. First, think through what messages can attract this donor given the donation history and their campaigns they have supported in the past. Then, think about what aspects of the children’s care program will appeal to them, given their story. Finally, write the email of the personalized donor using your analysis. “
6. Experiment and iterate
Change guidelines: Using different guidelines can help the model learn more about the task and produce more diverse and creative production, according to meta. Try using different styles, tones and formats.
Test and refine: Once you have created quickly, try it in the model to see how it performs. If the results are not as expected, try to quickly refine by adding more details or adjusting the tone and style.
7. Remember to review the results
Even the best helper of he can slip. Be sure to check the results for accuracy before using it for business purposes.