Director, Learning Design & Development| PMIAsheville, NC, United States
Validating and checking outputs is critical when working with AI systems like Generative AI. Such validation approaches may include establishing clear criteria, implementing strong testing protocols, and continuous refinement.
In your experience with AI, what are some best practices for ensuring the results you receive are accurate, relevant, and aligned with your original goals?
Providing the specific context in clear and consise way is essential.
Yes Giorgios. When working with AI we have to be more more specific, describing the persona, the context, the task/request, etc. Vague prompt will produce general answers and not the information we really want.
Also, it is important to evaluate the response with the eyes of a project manager, which seems to be a forgotten actor in all this AI trend. Saving Changes...
Lori WaldoEDI Developer| Tempur Sealy InternationalWinston Salem, Nc, United States
Cast with a wide net, and, as the AI responds, tweak your input to get to the best output. Saving Changes...
Amy HiltunenProject Manager| National Association of State Workforce Agencies (NASWA)Mi, United States
Garbage in, garbage out! Ensuring your results are accurate, relevant, and aligned starts with making sure you have quality inputs (whether it be the data an AI model is trained on or your prompt). Human review of the results is critical as well which could me you as the PM or bringing in subject matter expert. Saving Changes...
Badri N SrinivasanAVP and Lead - Agile Center| Societe Generale Global Solutions CenterBangalore, Karnataka, India
When using AI systems, some of the best practices for ensuring the results we receive are accurate, relevant, and aligned with our original goals are to check and validate the results, compare with other historical / current data, check with other secondary sources (more than one check), discuss with other experts to get a feel of the type of output that you may receive, check with other AI systems to see if you get similar results and finally do a direct check on the data and sources to see if it is accurate and can be trusted. This type of analysis may be done only for specific outputs that are very critical and cannot go wrong.. Saving Changes...
Pedro CarneiroCivil Engineer| EPEI - Escritório de Projetos de Engenharia e Inovação Ltda.São Paulo, São Paulo, Brazil
I currently direct AI work to my domain knowledge, together with the use of PM training and refinement techniques, as well as those presented in previous answers, to ensure data validation security. I believe that in the near future we will also have officially certified “AI validators” who can guarantee validation of work development in various areas of knowledge and thus enable us to advance in a continuous process of progress in meeting the greatest diversity of demands aligned with the original project objectives. Saving Changes...
I have found that the responses from LLMs are better when asked for general relationships than for specific information. I try to validate against known sources, or ask the LLM to show the source of information. If the source is not shown then I ask the LLM to look for it, and it normally improves the answer. Saving Changes...
Validating and checking outputs is a "must" step when working with AI system. Saving Changes...
Ramesh AnandTechnical Project Manager| PhilipsBangalore, Tn, India
When using AI systems, some best practices for ensuring the results you receive are accurate, relevant, and aligned with your original goals is to follow the CREATE Formulae, Prompt to include Character, Request, Examples, Adjustments, Types of Output and Evaluation.
Iterate the Create Formulae till the accurate, relevant and aligned results are achieved to match the objectives. Saving Changes...
I think that by using an iterative process our original goals will be accurate, relevant, and aligned.
Checking outputs and sources and, improving our requirements will help us. Saving Changes...
May be having 2 independent AI tools one for generating response and another for cross checking the same. If there is a difference/gap in the same, we can measure the level of accuracy. Saving Changes...
PMI Team Member
