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?
To fetch accurate and reliable output, we can prompt AI with back to back queries , so during this conversation with AI, at some point we have reliable information to be trusted thus eliminating confusion.
Saving Changes...
Stelian ROMANProject Manager| MicroSafetyCarlingford, New South Wales, Australia
Same principle as when we use information from a team member or a stakeholder: read and validate before taking ownership.
...
1 reply by Shawn Robison
May 23, 2026 7:10 AM
Shawn Robison
...
Yes! THIS!!
Saving Changes...
Shawn RobisonProgram Manager| Johnson & JohnsonFort Worth, Tx, United States
May 23, 2026 1:05 AM
Replying to Stelian ROMAN
...
Same principle as when we use information from a team member or a stakeholder: read and validate before taking ownership.
Yes! THIS!! Saving Changes...
Josè Luis VillabrilleProject Manager| ENEL-Engineering and ConstructionPalma De Mallorca, Spain- Baleares, Spain
I think that is a time question. The more IA is being used the confidence increases. Starting from simple activities and giving room for higher responsability in the developed tasks. Checking and checking.
Saving Changes...
Josè Luis VillabrilleProject Manager| ENEL-Engineering and ConstructionPalma De Mallorca, Spain- Baleares, Spain
I think that is a time question. The more IA is being used the confidence increases. Starting from simple activities and giving room for higher responsability in the developed tasks. Checking and checking.
Saving Changes...
SANTOSH BADGUJARCHIEF OPERATING OFFICER| Accumax Lab DevicesAhmedabad, Gujarat, India
Sarah, great question. Having used AI tools extensively across project management, operations reporting, and quality documentation, here are the best practices I've found most effective:
1. Define your goal precisely before prompting. The more specific you are about what output you want — format, length, audience, constraints — the better the output. Vague inputs produce vague outputs. "Write a risk summary for executive stakeholders, focusing on top 3 risks by probability and impact, in bullet format" will always outperform "summarize risks."
2. Provide context, not just a question. AI systems lack your organizational context by default. Include relevant background: the project phase, the stakeholder audience, the constraints you're working within. The richer the context, the more relevant the output.
3. Validate outputs against primary sources. Never use AI-generated data, statistics, or factual claims without cross-checking against authoritative sources. This is especially critical in compliance-sensitive environments like regulated manufacturing or healthcare.
4. Iterate rather than accept the first output. If the first response isn't right, ask for refinement. Specify what's missing or off. AI tools work best as collaborative draft partners, not one-shot oracles.
5. Use domain knowledge to spot hallucinations. AI can produce confident-sounding outputs that are factually wrong or context-inappropriate. Your expertise as a PM is the quality gate. Don't outsource your judgment.
The through-line: AI amplifies the quality of your thinking. Put in clear, well-structured thinking; get out useful output. Saving Changes...
Anonymous
Persona pattern shall be used when you are not the subject matter expert and looking for clarity on the matter. Flipped interaction shall be sued if you are not clear with what to ask and thereby develop the questions iteratively. ReAct shall be used in a dynamic and evolving scenario
PMI Team Member
