Team, with AI engines "ready" to answer your questions, prompts have become critical towards "engineering" to get the response you want. I have often foudn that creating the appropriate prompt is challenging. For example, do I give a situation, context, specific instance or lead to what exactly I want from the AI application? I recently checked to get a project management plan for a problem and found myself re-designing prompts multiple times to get the answer I wanted. I was amused to see some of my colleagues in a bistro sitting and discussing ways and means of creating a prompt (sharing tips and tricks!)
Does anyone have a suggestion on how to design a prompt? Have you faced this situation? or do we ask ChatGPT what should the prompt be??
Your thoughts please. Saving Changes...
Sort By:
Thomas WalentaGlobal Project Economy ExpertHackenheim, Germany
Prompts for AI are the equivalent of an interview questionnaire and planning for humans. I do not think of prompts as one-off questions to AI but rather as the start of a dialogue, as you do with humans.
Designing prompts needs time and tries to project what AI might consider. It is predictive planning, and as we know, plans fail when they confront reality in an unlimited knowledge space.
can you help create formatted pdf documents with synthetic data for these "1. IT compliance document types? HIPAA Compliance Documents: Ensure that employees handling health information comply with the Health Insurance Portability and Accountability Act. 2. GDPR Compliance Records: Document employee adherence to data protection regulations for handling European Union citizens' personal data. 3. Code of Conduct Acknowledgment: A document signed by employees to confirm they understand and agree to adhere to the company's code of conduct. 4. Anti-Bribery and Corruption Training Certificates: Records that employees have completed training on anti-bribery and corruption laws." please create 12 copies for each of them and package them for me to download. please ensure synthetic data generated is close to real world, information rich, include any dummy signatures, date properly populated. These documents are needed for testing Unstructured data products in DataMarket
Here i could not find attachement to share my documents to restructure. Are we planning to have similar like ChatGpt in future or guide me how to use for such scenarios
Understanding prompt patterns and how GenAI behaves helps.
Coursera offers a class on prompt patterns from Vanderbilt University that I've found helpful. GenAI is designed to provide an answer, which is why it sometimes "hallucinates" instead of saying "I don't know." There are times when you need to rework your prompt. I've found that prompts that worked in one version behave differently in the next release and in other GenAIs.
You can also ask GenAI to provide you with a better prompt when you ask it a question. Saving Changes...
PMI's prompt engineering course provides some useful templates and approaches for the iterative approach required to get better educated guesses from your favorite Gen AI tool.
We need to remember that this is a temporary problem. At some point, the natural language processing capabilities of AI tools will become advanced enough to require no more clarification than a colleague would.
Kiron Saving Changes...
Sergio Luis ConteHelping to create solutions for everyone| Worldwide based OrganizationsBuenos Aires, Argentina
Yes. You have to take into account different prompt frameworks that you can use according to the information you have on hands and the expected response. Here comes:
R-T-F: act as Role, create Task, show as Format
T-A-G: define Task, select the Action, clarify the Goal
B-A-B: explain the problem Before, state outcome After, ask for the Bridge
C-A-R-E: give the Context, describe the Action, clarify the Result, give the Example
R-I-S-E: specify the Role, describe Input, ask for Steps, describe the Expectation Saving Changes...
Product Operations Program ManagerBarcelona, Cataluña, Spain
It’s essential to include sufficient detail to make the content both comprehensive and easy to understand. When refining the text, consider that the AI-powered language model might not be familiar with the specific context or nuances of the subject matter. Therefore, it's important to provide ample context and specific details to prevent any ambiguity. In this case, the principle of 'less is more' does not apply; instead, offering more information is crucial for ensuring that the language model can generate the most accurate and relevant response. Saving Changes...
Frank SpiegelSenior Projectmanager, PMP, PMI-ACP| Commerzbank AGOberursel, Germany
I recommend the new PMI course "Talking to AI: Prompt Engineering for Project Managers".
I think this course is a good starting point/entry point for systematic prompt engineering, from which you can evolve with increasing experience. Saving Changes...
Community Champion
