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?
When I use AI systems, I usually use different ways to verify that the response is accurate. One method I use frequently is asking different AI models and comparing their answers for similarities.
Saving Changes...
Anonymous
The quality of the data is paramount. Try to build test questions to evaluate the quality and accuracy of your data to ensure that it can be trusted. Ask for references and cross-check as time allows.
Saving Changes...
Anonymous
The quality of the data is paramount. Try to build test questions to evaluate the quality and accuracy of your data to ensure that it can be trusted. Ask for references and cross-check as time allows.
Saving Changes...
Anonymous
The quality of the data is paramount. Try to build test questions to evaluate the quality and accuracy of your data to ensure that it can be trusted. Ask for references and cross-check as time allows. Saving Changes...
Jens SquiresProject Office Lead| KyndrylKent, United Kingdom
As someone who’s spent the past 20 odd years setting up and running PMOs across large IT outsourcing programmes, I’ve am starting to see a real shift in how we use AI tools to support our decision‑making and shaping of our PMO solution. AI has become incredibly useful, but like any other piece of kit in the PMO, the value you get from it depends on how you use it.Here are a few practical habits that have helped me get improved responses that are more accurate, relevant, and genuinely helpful outputs from AI without losing sight of the original intent. Firstly, start with a clear and focused question. One thing I’ve learned is that AI responds best when we are clear about what we want.In the PMO world, that means taking a moment to define: ·What problem are we trying to solve? ·How will the output be used? ·Who will the information be for? For me, iIt’s a bit like writing a Statement of Work, if the brief is fuzzy, the outcome will be too. Secondly, treat AI as a smart assistant, not a final authority. AI can sound incredibly confident even when it’s wrong. So, I always make sure we validate anything important against the contract schedule, my personal expertise and, where relevant, either our company established sources (providing AI with examples) such as programme plans, risk logs, toolsets, baseline data, governance reports, and so on. When you’re managing contractual obligations, SLAs, or financials, there simply isn’t room for leaving up to AI, the human review of the output stays essential. Following on from the above, I would suggest that you use trusted data wherever possible. If you want AI to give you sensible answers, feed it with sensible information. So, as an example, for large outsourcing deals, our company and even my personal file directory means I am sitting on years of experience and captured output from Programmes that consist of Lessons learned, performance data, delivery metrics, quality audits and process frameworks and user guides that can act as input against the original question posed. AI can draw on well-governed internal data, the quality of its suggestions increases dramatically when supplied, and you spend far less time filtering out unhelpful noise. I am starting to consider using a PMO library to create reusable prompts for PMO use cases. The one thing I’ve found helpful is building prompt patterns for common PMO tasks, for example. Summarising risks and issues, drafting governance updates, explaining variances, reviewing RAID logs and even producing executive‑level commentary for important meeting minutes as required. It can therefore assist in standardising the quality of output, reduces rework, and keeps the AI focused on our actual goals and refrain from having to drift into creative writing mode. Also of importance is not simply taking the output as a given, we need to also need to bear in mind the accountability of what is produced front and centre. No matter how clever AI becomes, it shouldn’t replace professional judgement. Therefore, within PMOs, we need to be clear about what data we’re feeding into AI tools, how the outputs are being used, where the boundaries are (particularly with client data and confidentiality) and who ultimately remains accountable – spoiler alert: that will be us ;0) AI can support the governance, but it cannot be the governance. We must continually review and refine as part of PMO continuous improvement. We must regularly review our AI usage regularly to understand what’s working well, where the output is still a bit hit‑and‑miss, what prompts need fine‑tuning and determining which new PMO processes could be enhanced by effectively using AI prompts.This approach will keep us moving towards a more mature, more intelligent PMO ecosystem without abandoning the core disciplines that keep projects on track.
In short, AI can be a brilliant enabler for any PMO, but only when we approach it with clarity, discipline, and a healthy dose of professional scepticism. If we combine AI’s strengths with rigorous PMO governance, we end up with insights that are not only accurate and relevant, but aligned with the strategic outcomes our organisations expect from us. Saving Changes...
Anonymous
I'm beginning my GenAI journey so my response will be theoretical. It seems the best practices would include:
Loading up to date and accurate data
Asking specific questions
Detailing desired outputs
Providing context
Outlining the problem to be solved
Saving Changes...
Norman WokomaProject Management| Tecnimont Nigeria LimitedPort Harcourt, RI, Nigeria
Iterative approach and continuous refinement using the CREAT prompting will go a long way in producing the needed results.
Don't fall into the trap of thinking that everything can be done on its own. However, you need to use human intelligence and be patient in providing all the necessary information.
PMI Team Member
