When using AI systems, what are some best practices for ensuring the results you receive are accurate, relevant, and aligned with your original goals?

Yes, I agree with the statement "predictive text on steroids". ChatGPT is now considered "old" and dependent on business structure an organization may or may not need to have additional human resources for a new business unit. At my previous organization the entire staff was challenged to conduct prompt engineering as it relates to individual departments as opposed to creating a new arm in the business. More experienced developers were responsible for model training.

AI requires human oversight and quality data to be effective. While generative AI has revolutionized text prediction, its impact on non-tech sectors is still limited. Implementing AI systems demands a multidisciplinary approach, involving specialists from various fields to ensure responsible and effective deployment.

Very good analogy Sergio

"AI without Data (today called data science discipline or big data or whatever) is the same thing that live without oxygen."

I like it :-)

Putting the content to the test for expert scrutiny has been my take on using the responses generated by AI. We tend to fall victim to AI Hallucination if we don't verify the data and responses generated by AI which could sometimes be catastrophic if major project decisions are made merely based on the output of AI.

Best practices for accuracy are specificity and use of CREATE formula

Be specific, provide examples, use a persona.

Generative AI is just an ancient model but everything "explode" when Google published the new architecture called transformer in 2017. So, with that said, take into account that generative AI is just "predictive test with steroids" just simplifying the model. With that said, two key points has to be taking into account when somebody works with AI: 1-human in the loop. 2-AI without Data (today called data science discipline or big data or whatever) is the same thing that live without oxygen. Talking about generative AI all related to technology has almost not impact with relation to all related to non-technological roles and activities. What you stated about accuracy and things like that are easy to implement because there are a lot inside disciplines like statistics. Most of them to make things "a priori" to prevent instead of cure. Few organizations taking into account that when generative AI environments are put in place almost a new business unit has to be created where roles like lawyers, linguistic, diversity and inclusion specialist must be hire to help on put it in place.

Sergio, I believe you are spot on both in your history and simplification of defining AI. Same is true regarding (At least for the short term 2-5 years) there will be people representing the existing business units (Legal, H/R, sales), etc....
Long term, once AI matures, I expect an oversight group consisting of directors or VP Businesspeople, will exist on an "as needed bases".
I strongly believe in AI and the potential productivity gains when the technology matures.

you need to work with iteration and to review constinuously the outputs ... the human is accountable for the outputs

In my experience GenAI models available in market vary in terms of their accuracy as they are built (trained) differently. Some are more suited to certain fields like technology etc. Having this knowledge helps to select a model. During the prompting I found providing context as the most effective way followed by providing examples and reflection of my visual representation that we may have in our minds. Additionally asking AI to summarize helped since it extracted what it thought was key part of its response. Once it reached a certain level then I could trigger the "unpack"

In my experience using tools like Salesforce and Qualtrics in project settings, I’ve found that getting accurate and relevant results from AI really depends on how clearly you frame your goals from the start. I always try to be specific with my inputs and make sure they align with what I actually need from the system, whether that’s analyzing feedback or streamlining reports. Once I get results, I don’t just take them at face value; I double-check them against what I already know or what the project requires. If something doesn’t feel right, I tweak the prompt or ask follow-up questions until it makes sense. And honestly, getting a second opinion from a teammate has helped a lot too, sometimes another set of eyes can catch what you might miss. So for me, it’s a mix of clarity, iteration, and not relying on AI to do all the thinking for you.

Good explanation. Thanks

Keep it simple:

1-Clean your data.

2-Ask specific questions.

3-Review the results.

The exact approach will vary with the AI tools you choose and the goals you want to achieve. Once those elements are clear, you’re all set.

Hi
You will need to do an algorithmic impact assessment before implementation and then evaluate your entire AI system from end to end to ensure there is no drift, bias or other ethical dilemmas (like MechaHitler Grok). some good guidelines can be found in the IEEE AI series of standards.

Hi
Evaluating the performance of commercial AI models is more difficult because those are black box. Third party AI Audits may be useful.

In my experience, ensuring AI outputs are accurate and aligned starts with defining clear objectives and success criteria from the outset. I apply human-in-the-loop validation for critical outputs, use prompt engineering to guide relevance, and test across diverse scenarios to catch edge cases. Regularly reviewing outputs against benchmarks and incorporating feedback loops also help refine accuracy and maintain alignment with project goals.

Good analogy.

Thank you for your detailed perspective. I agree that the breakthrough with transformers in 2017 fundamentally changed the scope of generative AI. Your point about keeping humans involved and the important role of data is very valid. Without them, the system can’t deliver meaningful results. I also appreciate your focus on “prevention” through statistical methods instead of fixing issues after deployment. The idea of forming a dedicated business unit with diverse roles like legal, linguistic, and inclusion specialists is insightful. AI's impacts go beyond just technology. This complete approach seems essential for sustainable adoption.

Good scenario, Generative AI functions with both the human and data considerations

thanks

Good explanation !

Note that a project manager should be working with a team that also adds unique project details and business requirements. Not every detail will need to come from AI via prompt engineering.

Thank you for your view!

Oversite and validation is very important for effective use of AI. In the end it is a tool which works based on previously gone through data and scenarios.

I agree.

Iterate

I validate your point. As you mentioned, while Gen AI can be incredibly valuable for providing information. It's important to have subject matter experts involved to validate and assess the responses generated to avoid errors and ensure the best outcomes in decision-making.

Agree with the OP. We conduct a multi-faceted verification of outcome accuracy by comparing outcome with known trustworthy sources and subject matter experts as well as comparing outcome with authoritative databases. In addition, we conduct regular team introspections to examine and act on feedback.

I appreciate your explanation, Oliver; as I've been wondering this myself. Restricting AI to an area where I have sound knowledge would seem to ensure I could validate it's output based on personal experience.

I agree it is extremely critical that you review the content for accuracy which I think is best done by using subject matter experts and or determining the source of the output generated.

BLEU, ROUGE, and METEOR are all automatic evaluation metrics used to assess the quality of machine-generated text, particularly in tasks like machine translation and text summarization.

I agree. Using AI in a nonfamiliar area requires proper insight to navigate informational or technical mishaps which it may generate.

Excellent response, Oliver. I believe your response is very professional, and I strongly believe, like you, that even though there are many ways to improve communication using different techniques also, there are no precise defined metrics that one can use a model to evaluate an AI-generated response always there is an area of opportunity and continuously improve AI response and provide feedback to LLM behavior. You provided a very smart analysis; I like your level of critical thinking. Excellent job!

Good points! Lately I keep hearing garbage in, garbage out. The first step is to ensure that your prompts are asking for what you truly desire ina viable format. Also to ask AI to focus on accuracy than speed and to validate it's answers with the sources it used. When applicable it is good practice to upload documents and ask AI to reference specific text or valid websites. And of course to validate the information that you receive to ensure that it is accurate and timely because just like the best human experts/SMEs AI can also make mistakes.

That is a great explanation Oliver. Thanks for the help. Chip

That is a great explanation Oliver. Thanks for the help. Chip

SME verification sounds about right in that context.

AI, as useful as it is, is always going to require the knowledge of a human to help determine how coherent its responses are. Obviously, if the human doesn't know their domain area, the AI will steer them in any direction. From my point of view I prefer to see AI as a very capable assistant with the ability to access a great deal of knowledge. But, in what parcel of knowledge, are the answers we are looking for, still require criterion and good judgment from a human. It is this joining of forces that really makes AI powerful.

Good point!

I like the point about having a subject matter expert review GenAI output that I'm not familiar with. GenAI is just like us humans. Sometimes it understands and answers as we need. Sometimes, we may have to walk away, rethink, and rewrite the prompt to ensure the results you receive are accurate, relevant, and aligned with your original goals. We might need to verify results with legal statutes, codes, and procedures. Perhaps we might run the same prompt through several LLMs to see if all return the same results. We might explain what we want to have a peer review of our prompt and rewrite it accordingly to see if we get a better result.

The question is pretty difficult to address, given some limitations that has been associated with the use of AI. Nonetheless, an iterative and comparison of the outputs could increase the chances of getting a reliable results.

I agree that the reliability of any AI-generated response ultimately depends on how well it aligns with verifiable facts and the user’s subject knowledge. Restricting AI use to areas where one already has solid knowledge was also how I initially tested its capabilities when I first began using it. Your observation about confidence increasing with use is very true. Over time, users learn how to ask more effective questions, establish clear boundaries, and apply critical thinking when evaluating responses.

I agree that the reliability of any AI-generated response ultimately depends on how well it aligns with verifiable facts and the user’s subject knowledge. Restricting AI use to areas where one already has solid knowledge was also how I initially tested its capabilities when I first began using it. Your observation about confidence increasing with use is very true. Over time, users learn how to ask more effective questions, establish clear boundaries, and apply critical thinking when evaluating responses.

I believe that iteration is crucial to get more precise results but at the end we may have to rely on the accuracy of the response. Do you think that consulting an expert about a portion of the AI answer (let's say 30%) would make you fill confortable of the overall result?

this is a prudent approach

I normally go in discussion mode with AI to refine the responses I get. Somehow I use the RTF and CREATE model to get responses that are relevant and to the point.

Great question. I also find it a challenge to verify outputs that I am not conversant with. I have in the past requested the LLM to provide sources/references and also compared the outcome across various LLMs.

Great point. When working in an area, where I have good knowledge, I use GenAI in discussion mode. Here I present the initial request with context and expected response format. But when using AI in an area I am unfamiliar, I ask the AI to provide sources for all provided information and insights, while refining my requests with each response.

Correct and precise input with available support documents will help to get better outcome.

With sensative data, you need to generalize and provide the available document to be compliant with company policy and generic outcome can be used to align with your requirements.

AI is not an oracle of the ancient times that intermediates responses from the gods. It is much more similar to a dedicated interim that will deliver literally deliver what it was asked to.

So Mr. Chitsamatanga's advice of submitting AI suggestions to SMEs is wise.

Recently I heard a professor saying AI should be preferably used in two scenarios:

1. You know a lot about a subject and need to refine or evolve on it

or

2. You know nothing about a subject and need some directions.

I tend to strongly agree with him.

De acuerdo, si no se cuenta con un experto en el tema es muy complicado validar los resultados y no se puede asumir que la IA es el experto

I totally agree with this, you have to be versed with the subject matter, then you can verify the information that the AI model has provided for you

Always use the CREATE model of prompts for better results. reiterate the prompts accordingly for best results. The output should always be validated before publishing.

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.

Be as clear and specific as possible and provide context.

I agree and the perspective of the report audience is also fundamental

Ensuring the results from an AI system are accurate, relevant, and aligned with specific goals requires a systematic approach. Here are some best practices for validating outputs and maintaining alignment:

1. Establish Clear Criteria for Success

• Define Objectives: Clearly articulate the goals of the AI system and its intended purpose.

• Set Metrics: Use measurable benchmarks such as accuracy, relevance, precision, recall, or user satisfaction to evaluate performance.

• Contextual Relevance: Ensure outputs are aligned with the domain-specific requirements or the user’s intent.

I am a new fan of the "CREATE" model which incorporates Character, Request, Example, etc. into a prompt engineering framework. It really helps to get refined and precise results.

I really enjoy the ReAct method because you can get really specific and validate your answers.

Totally agree, it all depends on how better your prompt is to generate the reply in that context.

One has to know their expectations of the creation they wish the ai to fulfill. I don’t mind spending time to reach perfection. It’s much better to train especially while hypothetically implementing optional strategies. I do so for my personal portfolio. On the contrary if a person types a generalized prompt such as “build me a project plan” and has minimal experience in the field, it can be difficult to assume ai generates the correct answer. It surely will not be at its highest potential given this is a broad statement. Ai performs as accurately as it was created. I would advise to try and build one if the time permits the pm. Build, build, build! 😊

I usually just use the Question Refinement method on a conversational style. The AI's response then makes me think (and learn) how to ask better questions (ie provide detail, structure, context).

Exactly my thought too. no ambiguity, just clear, specific prompts and you keep adjusting till you get what you really want

having prompt formulas allow you for an easy way to create more specific and complete prompts which leads to a better response outcome. I look forward to incorporating them into my daily use of the AI tool.

Totally agree

Yes, also ensuring the model has the relevant information first. For closed AI models in particular, it's important to provide the specific documents or other information relevant to your prompt to ensure your output is accurate. You should always validate your responses

AI is like a baby who needs guidance.

Yes, it is important to use precise and specific prompt to get better result. We can use the RTF and CREATE techniques to have responded well from AI

I totally agree, If you really want the AI to give your meaningful and effective output for what you are looking for, then it's really important that you be specific with your ask, and the format of output also the role that AI is expected to play when giving you output..

I agree!

An educated consumer is still the best customer. I have found that when refining prompts and referencing prior conversational information, AI has outstanding recall and the ability to helpfully match relationships between the information both input and output. I think a terrible misuse of AI is to underestimate its capabilities to provide insights.

I think that especially a concise context is important.

Be precise and specific in your instructions. Run test case against known responses. Consult subject matter and project experts. Update with new information and data. Provide constraints round data and information to be considered.

and of course ensuring the responses are in line with your thought process and desired outcome This should be by validating every response with clear and continued refinement of the prompts.

Agree

hummmm...that sounds pretty efficient, concise and...agile...totally agree eventhough we do know sometimes things aren't always so simplistic.

I agree and further than asking AI to review it's own responses, there needs to be a QA/QC "sanity check" before external stakeholders see the results. This is fairly important as we get more used to AI and our relationship appears to be close to flawless, that's when we need to remain careful about reviewing results.

I completely agree with this response, you need to test the results as you receive them, but also validate the information with other sources.

iterative approach with checks on smaller portions of the whole solution

This resonates, especially when seeing it as a new model that has many glitches and still needs to learn. This perspective also adjusts expectations on accuracy and time required to actually mine useful responses from the llm.

GOOD

Best Practices for Using AI Effectively : Set clear goals, Know what you want from the AI.Use specific prompts, Provide detailed and focused input. Iterate as needed, Refine prompts to improve results.Fact-check outputs, Validate with trusted sources. Apply human judgment, Don’t rely solely on AI. Define success criteria, Know what a “good” result looks like. Watch for errors, AI can make mistakes.

This is not a reasonable approach because it is too risky to directly build all the connections and formulas without being able to confirm that each step is correct. You should first test each part in sections, verify the functions, and then gradually expand the entire model.

Good point! We have to keep in mind that AI must have data and must be trained, so it has to be verified.

I agree with you Keith.

I agree with you.

Yes! I think so many people want AI to be perfect and immediate and while it is a tool that greatly improves speed of work and efficiency, the subtle refinements made by putting in the time up front to check progress throughout the process of iterative generation keeps your outcomes on track. A small deviation from your course can have a big impact when considered at scale. Think about when you're driving. You're constantly adjusting the wheel a little here and there to keep the car between the lines. If you don't, you're going to end up drifting into another lane and causing an accident.

I fully agree on your feedback. Sometimes, and especially with AI, there are newer pathways of achieving the objective, but the expectation of what the outcome should be, needs to be clearly specified well ahead of building any AI related solutions

Agree - no one wants to fail at the last hurdle!

i like this approach, especially if considering different aspects in a model that you want to gather more data on before implementing.

I agree with this approach. Starting small and scaling over time is always helpful.

When I use AI, I start by clearly starting my goal, the audience, and any constrains so the tool knows exactly what I need. I only use the parts that support my original objective, so I stay accountable for the final result.

Garbage in, garbage out.

the core principle is treat AI as a “decision support system” and not a source of truth

It takes several attempts to get to the right flow of prompts. Otherwise I find the results start becoming vague as more details are added

Yes, the old saying, 'garbage in, garbage out." If you want clear, concise outputs, you have to be sure that what you are putting in is accurate and your goals are clearly defined.

I agree with Elmar's response. You need to be specific. I would use CREATE to ensure I get the best response and continue to refine until I have what I am looking for from AI.

I love this response. Thanks for sharing such detailed insights Herr Saenger

Solid points, AI will explicitly thank you in a sentence when generating a response to a solid prompt.

I´ll keep that answer in my files

Hi, great answer! I agree with you — interacting with AI is very similar to communicating with people. If you want a specific answer, you need to be clear in both the information you provide in the prompt and in your request.

Thank you

Hi, great answer! I agree with you — interacting with AI is very similar to communicating with people. If you want a specific answer, you need to be clear in both the information you provide in the prompt and in your request.

Thank you

i think your point of view is correct as strategy of thinking ,quality of information , understand the LLMs ,also continuous improve the questions to be advance will reflect on the output of the chat bot

Yes, always check the quality of the work to make sure the chatbot is using up to date and relevant information.

Excellent example! I would recommend adding constraints to the prompt by framing the output to be specific to the project and excluding outside resources. I often use this since it's common for information "leaking" into the output from other projects or prompt chats that do not pertain to the project that I am currently working.  

Experience and soft skils allows us to challenge AI Answers and refine questionning to get the most relaible and usable information

Use a proven framework like CREATE and continue to iterate till you get the results you are looking for.

That’s a really smart approach. Adding a confidence threshold along with a verification step not only improves output quality but also adds a layer of reliability to the system. I’m definitely taking notes from this thanks for sharing!

Thanks for your post - very helpful!

When using AI system, to get the results that are accurate and aligned with our proposed goals, the prompt must be specific. it is also a good practice to provide AI with patterns to follow and models to avoid. There should be also a process to evaluate the output as to correct any discrepancy between input and output.

Agree, input provided is very important as well as continue iterations to refine and get an output that satified your need.

I am agree with you, as log as the question is direct it will narrow the answer and make it specific. simplicity of giving the information to AI tools make it clear and the answer will be more accurate. and for sure the more continuing review and improvement, the more improvement of project and results.

I am agree with you, as log as the question is direct it will narrow the answer and make it specific. simplicity of giving the information to AI tools make it clear and the answer will be more accurate. and for sure the more continuing review and improvement, the more improvement of project and results.

Working as a PM in India's dynamic startup ecosystem while leveraging AI tools daily, I've learned that ensuring accurate, relevant results requires a nuanced approach that accounts for our unique operational context—limited resources, rapid pivots, regulatory complexities, and diverse stakeholder environments.
The most critical practice is embedding Indian startup realities into your prompts. Generic AI responses often assume Western business contexts that don't translate to our environment.

Instead of: "What are typical project risks for a fintech startup?" I prompt: "For a fintech startup in India with 15-person team, ₹2 crore funding, targeting Tier-2 cities, operating under RBI guidelines, what are the top risks considering regulatory changes, talent retention challenges, and payment gateway dependencies?"
This specificity is crucial because AI models often default to Silicon Valley contexts. When I was managing a digital lending platform launch, generic AI risk assessments missed critical India-specific factors like UPI integration complexities, regional language requirements, and state-specific compliance variations.
I never accept AI outputs without triangulating against ground truth. For instance, when AI suggested a 3-month development timeline for our e-commerce feature, I cross-referenced it by asking: "What factors could extend this timeline in an Indian startup context with monsoon season disruptions, festival schedules, and potential talent attrition?" This revealed that our initial estimate needed a 40% buffer for local realities.
Indian startups operate differently than global counterparts—we have flatter hierarchies, more informal communication, and relationship-driven decision-making. I continuously refine AI outputs to match these realities.
Startup resource limitations require careful AI query design. I've learned to prompt AI with explicit constraints that reflect our ground reality.
Indian startups navigate complex regulatory landscapes that change frequently. I maintain a practice of updating AI context with current regulatory information.
Indian startup teams often blend different cultural backgrounds, languages, and work styles. I use AI to optimize for these dynamics while applying human judgment for cultural nuances.
I consistently validate AI insights against Indian market research and local data sources. When AI suggested pricing strategies based on global SaaS benchmarks, I cross-referenced with local reports from firms like RedSeer, Bain India, and BCG India to ensure relevance.
I maintain continuous feedback loops with AI based on real project outcomes. After each project milestone, I analyze where AI recommendations succeeded or failed and use this learning to improve future prompts.
Indian business relationships often require personal touch and relationship-building that AI doesn't naturally account for. I use AI for content generation but always adapt for relationship dynamics.
Given the volatility of Indian startup environments, I use AI for comprehensive scenario planning while grounding assumptions in local realities.

The most effective AI integration for Indian startup PMs comes from treating AI as a sophisticated research assistant that requires continuous cultural and contextual calibration. Success depends on your ability to bridge AI's analytical capabilities with deep understanding of Indian business dynamics, regulatory environment, and startup culture.
The goal isn't to get perfect AI outputs immediately—it's to develop a systematic approach to refining AI insights until they align with your project realities and organizational context. This iterative approach has consistently delivered more accurate, actionable results that drive real project success in our unique operating environment.

Couldn't agree more --- Continuous review is key!

I agree with your point of view and I belive that it's important to test and verify the result that AI provide, anf then we improve it.

Definetly agree! I have read very useful comments, but yours is very assertive "keep it simple".

100% right on that. Data has to be clean, the info we are feeding or asking should be clear, concise and specific to what goal we are trying to achieve

Always start with a clear purpose. Think of this like setting a destination before a journey. If you’re using AI to write reports, analyse data, or generate ideas, be clear on what you want the AI to do.

This helps you judge whether the output is actually useful.

Always start with a clear purpose. Think of this like setting a destination before a journey. If you’re using AI to write reports, analyse data, or generate ideas, be clear on what you want the AI to do.

This helps you judge whether the output is actually useful.

I agree keeping it simple with some clear built in validation criteria to ensure it doesn't AI hallucinate and also reiterate and refine the prompt so that the AI learns how best to respond

Agree to keep it simple if you are new using Gen AI. Think of what you want to achieve and how the result should look like. Start small, look at the outcome and add some more context. Check and review it again etc, etc.

Use it like a continuous improvement process and check the outcome. Get the right (technical) people to verify your outcome.

Nice formula, Mashhood! Thnaks

Thank you for putting this together. I'll make use of it when I need to use it.

Keep it simple by providing clear and concise prompts, refine as needed, and give yourself time to learn.

good point on providing the outcome.

I agree that you have to clearly define the objective of the prompt and avoid ambiguity—be as specific as possible. Also, provide relevant background information and examples. It is always good to follow the role, task and format method. It will help you produce accurate, reliable, and relevant results.

Hi Melissa, I strongly agree with your discussion. Be precise, clear, and testing and refining your response is a critical way to ensure the results you receive are accurate, relevant, and aligned with your original goals and desired outputs. Outstanding point of view. Thank you for sharing!

Simply put, continuous iterative prompt refinement will ensure accurate; relevant and original aligned result. Validation from skilled expert could also help.

I like that you mentioned jargons, explicitly breaking down the request and clearly specifying the company specific jargons do help in getting the desired output. As some of these jargons are very specific to the team and organizations.

I agree with your points. It is crucial to elaborate on specialized terminology and industry-specific context so that the outputs are even relevant. And for me sometimes, it can take too much time and effort to keep refining the input data.

Additionally, A key best practice when using AI systems is involving domain expertise—specialized knowledge in the relevant field—to ensure that the system’s outputs are accurate, meaningful, and applicable to real-world needs.

Yes, breaking down project into chunks is definitely helpful. I have also found that Microsoft's Copilot often provides good follow up prompts. The back and forth "discussion" with an LLM will help refine the response.

To ensure that your results are accurate you should keep it simple, input proper prompts, do not use industry specific terms that an AI may not know. After this, you should review the output for accuracy and then refine your prompt for necessary changes.

In addition the course does touch the topic of AI Hallucinations, Non Current Data, misalignment, etc. and associated problems. The best practices or solutions provides additional measures to address problems and symptoms of vague ,outdated, conflicting incomplete responses.

Patterns such a Tree of thoughts, Personas, ReAct etc. provides additional methods of refinements leading to better refined and complete responses.  

I agree with you and believe the following helps to maximize AI effectiveness:

1. Define clear objectives
2. Accurate and up-to-date data
3. Provide necessary background information or constraints
4. Do not forget the need for the "human" in the loop

I agree with this method.

Jabin, I love your analogy. I was thinking how prompting is much like the need to be clear and concise with a new-hire who has no idea what you are looking for. 😉

Hi Jabin-
Completely agree that it is very similar to mentoring associates! The ability to take a defined character or role and tailor the approach or answer to a specific question. Most of the time you are, as the mentor, peeling back the layers to get to the actual problem or question all while re-evaluating responses based on output received from the mentee.

Jabi, that was an excellent discussion. I agree with your point of view when you mention the Tech. The solution that comes in for accuracy is having a specific set of APIs that talk to real and accurate data sources or use 2-3 outputs of LLMs and then analyze and bring the best in output.
Definitely, using an LLM analyst provides a better response when integrating two or more outputs and combining them to create accurate, relevant, and aligned outputs with your original goals and desired outputs. Excellent discussion!

, agile can make AI development faster and more responsive, but it requires special care—for example, keeping ethics, data quality, and thorough records a regular part of every sprint. Properly adapted, agile brings real benefits to AI work.

Agile sometimes neglects thorough documentation, which AI projects need for reproducibility and compliance.

If not careful, the rapid pace of agile could mean teams overlook transparency or ethical concerns with AI systems.

Of course, we can give feedback and also re tweak our prompts

When approaching the challenge of precision and accuracy in AI responses, it’s important to think about this from both a project management and Agile mindset perspective. As PMP professionals, we know that continuous refinement, stakeholder feedback, and adaptive planning are central to achieving better outcomes. The same applies to working with AI systems.

1. Feedback as a Control Mechanism
Just as we monitor and control project work, we must treat AI responses as deliverables to be reviewed. Providing structured feedback ensures alignment with objectives, similar to how mentoring or coaching sessions guide mentees toward the right path.

2. Prompt Re-work as Iterative Planning
In Agile projects, requirements evolve and get refined through sprints. Similarly, reworking prompts is an iterative process where clarity and specificity drive improved outputs. Each iteration should narrow ambiguity and communicate expectations clearly to the AI system.

3. Conciseness and Incremental Adjustment
Agile emphasizes simplicity and incremental delivery. Keeping prompts concise allows us to inspect results quickly, make judgments, and tune the system iteratively minimizing waste and focusing on value.

4. Technical Enablers for Accuracy
From a PMP lens, accuracy is not just process it also requires the right enablers. This could mean integrating APIs that connect to verified data sources, or using ensemble approaches (reviewing multiple LLM outputs and selecting the most accurate). These solutions act like quality assurance gates in a project lifecycle, ensuring decisions are based on reliable information.

Yes, very true. Thanks for those insights

I agree. Utilizing the best prompt formula and adjusting prompts according to the output received.

I have definitely done this before... giving feedback to the AI system because inherently it will get smarter and helps the next individual that interacts with it for a similar prompt.

I agree

estoy de acuerdo con la respuesta , simplificada es establecer los patrones vanzados del promt a travez de cadenas de pensamiento