top of page

The Essential Guide to Partnering with GenAI: Achieve Both Accuracy and Precision

Writer's picture: Jeff HulettJeff Hulett

Updated: Jan 20

The Essential Guide to Partnering with GenAI: Achieve Both Accuracy and Precision

This article will guide you on how to integrate GenAI into your life, offering eight actionable prompt design suggestions to improve your AI interactions. It concludes with a relatable car-buying example featuring Bob, a young professional, to illustrate how these strategies help get the most from our GenAI partnership.


In today’s AI-driven world, partnering effectively with Generative AI (GenAI) can significantly amplify outcomes. However, to unlock the full potential of this partnership, it is essential to grasp the difference between accuracy and precision. Accuracy ensures that GenAI aligns with your specific goals, while precision focuses on delivering consistent and reproducible results. Together, they create a powerful synergy that enables impactful and reliable outcomes.


Lacking accuracy while being precise is like being exceptionally skilled at implementing someone else’s strategy—it may be executed perfectly, but it does not align with your objectives. Conversely, being accurate but imprecise is akin to being well-intentioned but sloppy, where results may be on target occasionally but are unreliable overall. In both cases, the desired result remains out of reach.


About the author: Jeff Hulett leads Personal Finance Reimagined, a decision-making and financial education platform. He teaches personal finance at James Madison University and provides personal finance seminars. Check out his book -- Making Choices, Making Money: Your Guide to Making Confident Financial Decisions.


Jeff is a career banker, data scientist, behavioral economist, and choice architect. Jeff has held banking and consulting leadership roles at Wells Fargo, Citibank, KPMG, and IBM.


Accuracy: Reducing Bias for Clear, Goal-Oriented Results


Accuracy in AI means achieving the intended target. It involves reducing bias by setting clear goals and aligning processes with those goals. Part of guiding accuracy includes choosing the data the GenAI considers. While GenAI is trained on a broad range of data, the human partner may impact that training by supplementing the data with sources aligned with their specific goals. Think of accuracy as the compass guiding GenAI toward relevant results. Without this, the output may appear refined but lack relevance. GenAI’s role is to produce the requested content, but it is the human partner’s responsibility to define the goal and ensure results are on target and useful.


Despite GenAI’s impressive capabilities, it lacks the context to understand goal nuances independently. For example, if you ask GenAI for information without a clear goal, it may generate responses that are detailed yet misaligned with your intent. Similarly, if your goal is unclear, GenAI will respond based on patterns in its training data, which may carry inherent biases not disclosed by its providers. As many AI ethicists have warned, these biases will affect results, increasing the importance of clarifying your objectives to achieve the best partnership with GenAI. Providing context ensures GenAI delivers accurate, valuable results. By clarifying your objectives, you minimize the risk of irrelevant or difficult-to-detect biased content.


Precision: Reducing Noise through Consistency in Results


Precision, meanwhile, is about consistency. It is the ability to generate repeatable, noise-free results across similar queries. Think of precision as a tool for achieving the same result each time, ensuring consistent quality. When applied to GenAI, precision minimizes variations, providing stability in outputs. However, while GenAI is faster and more precise than most humans, it is still not perfect. Think of it like the 80/20 rule. It will get it mostly correct. It is up to the operator to decide whether it is worth the time to make the results perfect - like going from 80% to 100% precision.


GenAI’s precision depends heavily on achieving initial accuracy; when goals and context are clear, the AI can produce more precise and relevant results. This accuracy ensures GenAI's structured responses, summaries, or data extractions align closely with user intent. While precision delivers consistency, it is accuracy that directs GenAI toward delivering outputs that are both reliable and meaningful. For actionable ways to improve both accuracy and precision, see the suggestions provided at the end of this article.


The Human-AI Partnership: Combining Accuracy with Precision for Maximum Impact


A successful GenAI partnership combines AI’s precision with human-defined accuracy. Humans set the goals, add context, and reduce bias, ensuring that each response aligns with the intended purpose. GenAI, in turn, brings powerful precision, generating consistent responses that enhance efficiency and productivity.


While GenAI often confidently presents results suggesting precision, it may not always be correct. Human oversight is essential to verify these results and avoid potential errors. For example, if a researcher uses GenAI to gather sources and summarize essential insights, the tool may produce responses that seem precise. However, the researcher must review the output carefully to ensure its reliability and relevance. Together, they create a result that is goal-aligned, repeatable, and free from irrelevant noise.


The Essential Guide to Partnering with GenAI: Achieve Both Accuracy and Precision

Top 8 Suggestions for Getting the Most Out of Your GenAI Partnership

- Check out the appendix for prompt examples


  1. Define Clear Goals from the Start Be specific about your objectives. GenAI performs best when given a well-defined task. Clear goals allow it to align its precision with your intended outcome. Ambiguous or overly broad tasks can lead to scattered or less relevant outputs, so ensure that your goals are focused and measurable.

  2. Provide Context with Your Queries GenAI lacks an inherent understanding of context, so ensure your prompts are detailed and relevant to the topic. Adding context helps it generate responses that are more tailored and useful. Without sufficient context, GenAI may generate generic or off-target results, so be intentional about the details you include in your queries. When providing context, include specifics such as the type of data GenAI should consider and trusted sources it should reference.

  3. Review Outputs for Precision GenAI’s responses may appear confident and precise, but they are not always correct. Always review its output to confirm that it aligns with your original goals and is free from errors or inconsistencies. Evaluate not only the accuracy of the content but also its alignment with your specific objectives to ensure the output delivers value. From my experience, GenAI notoriously lacks precision with citations. I perform a heightened level of quality control over sources.

  4. Curate Credible Information for GenAI Enhance GenAI’s output by supplying curated information from reliable sources. By sharing data or context from reputable references, you improve the quality and relevance of GenAI's responses, helping it to produce results that align more closely with established facts and trusted insights. For example, the author Jeff Hulett uses his own articles from The Curiosity Vine and Personal Finance Reimagined to curate information. These sources provide a wealth of credible references, enriching the context and ensuring accurate, reliable outputs.

    Avoid relying solely on GenAI to source information. Supplement its outputs with research from reputable institutions, such as automotive review sites or financial planning tools, to confirm accuracy and eliminate bias.

  5. Use Iterative Prompting Refine your prompts if the initial responses do not meet expectations. Simplifying long, multipurpose prompts into a series of "bite-sized" prompts will help steer GenAI closer to what you need, improving both accuracy and relevance over time. This includes recognizing that providing too much guidance in a single prompt may confuse the AI, leading to less focused or inconsistent results. Additionally, consider testing variations of the prompt to identify which phrasing yields the most accurate and relevant response, as subtle changes in wording can significantly impact the quality of the output.

  6. Leverage GenAI’s Strengths in Consistency GenAI excels at generating repeatable, structured responses. Use it for tasks that benefit from this precision, such as summarizing information, generating lists, or analyzing repetitive data. For example, use GenAI for drafting email templates, creating structured reports, or generating customer service responses. Even with consistent outputs, verify the quality of the structured data it relies upon.

  7. Specify Output Format for Clarity and Usability Clearly indicate the format you want for responses, whether it is a summary, a bullet-point list, a step-by-step guide, or a guided narrative. This guidance helps GenAI structure its output in a way that aligns with your preferences, making the information more accessible and directly usable. For example, Jeff Hulett uses his rules inventory to guide ChatGPT's writing, ensuring consistent outputs aligned with his professional approach.

    In scenarios requiring detailed references or citations, specify that the AI includes verifiable sources. Independently validate these sources to avoid issues with fabricated or incomplete information, ensuring the output is both clear and credible.

  8. Provide Feedback on Responses for Continuous Improvement After receiving an answer, offer feedback on accuracy, relevance, or style. Letting GenAI know when an answer meets or falls short of your expectations allows it to refine its responses over time, making future interactions more productive and tailored to your needs. Offer specific, actionable feedback about what worked or what fell short, which can help tailor the AI's future responses more effectively.

    Providing detailed feedback also documents discrepancies or limitations in GenAI’s output, helping you identify patterns that may require adjustments in your approach.


By following these suggestions, you can maximize the effectiveness of your partnership with GenAI, achieving both accuracy and precision for better, more actionable results.


Understand GenAI’s Limitations


Be aware that GenAI, while powerful, has inherent limitations such as a lack of true understanding, susceptibility to bias, and dependence on training data. Recognizing these constraints will help you set realistic expectations and focus its use on areas where it excels, like structured automation and creative ideation.


Conclusion


In an effective GenAI partnership, understanding accuracy and precision unlocks AI’s true value. GenAI offers reliable, noise-free responses, while the human partner ensures every output is anchored to the desired goal. By merging human context with AI’s precision, GenAI becomes a powerful tool for achieving accurate, bias-free, and consistent results. This partnership amplifies efficiency, transforming AI into an invaluable asset for informed decision-making and actionable outcomes.


Appendix: A Car Example to Implement the Top 8 Suggestions for Getting the Most Out of Your GenAI Partnership


This appendix demonstrates how to apply the Top 8 Suggestions for Getting the Most Out of Your GenAI Partnership in the context of car buying. The framework for these prompts is inspired by Jeff Hulett's book, Making Choices, Making Money, which provides a structured decision-making approach to personal finance. By combining the principles from the book with the Definitive Choice app, these examples illustrate how to harness GenAI effectively to make confident and well-informed car-buying decisions.

Meet Bob: a young professional just getting started in his career. Bob has recently landed a steady job and is evaluating his transportation options. While public transportation or ride-share services initially seemed appealing, Bob has determined that purchasing a car will better align with his needs for reliable and safe commuting. With a $20,000 budget, Bob wants to ensure he makes the best possible decision, balancing cost, safety, and long-term value. Let’s explore how Bob uses the Top 8 Suggestions and GenAI to achieve his car-buying goals.


  1. Define Clear Goals from the Start

    Prompt Example: "I want to purchase a car that meets my $20,000 budget, has a maximum mileage of 40,000, and is reliable for long commutes. What are the best models to consider, and how should I prioritize my preferences?"

    Why This Works: A clear goal ensures that GenAI aligns its recommendations with specific parameters such as budget, mileage, and reliability.

  2. Provide Context with Your Queries

    Prompt Example: "Using a Cost Per Remaining Mile (CPRM) approach as outlined in Making Choices, Making Money, calculate the CPRM for a 2018 Toyota Camry with 50,000 miles and a purchase price of $20,000. Include depreciation and expected maintenance costs over the next five years. Assume a maximum vehicle life of 120,000 miles."

    Why This Works: Providing context ensures that GenAI delivers actionable insights aligned with the book’s CPRM methodology, helping you evaluate the long-term financial implications of your purchase. Refer to the book for details on calculating CPRM.

  3. Review Outputs for Precision

    Prompt Example: "Based on your recommendations for reliable cars under $20,000, can you provide details on expected annual maintenance costs and average fuel efficiency for a 2016 Honda Civic?"

    Why This Works: Reviewing outputs critically ensures that the recommendations align with your preferences and remain accurate and relevant.

  4. Curate Credible Information for GenAI

    Prompt Example: "Using data from reputable sources like Kelley Blue Book or Edmunds, identify the five best compact SUVs with the lowest depreciation rates and highest reliability ratings."

    Why This Works: Specifying credible sources ensures high-quality results tailored to your priorities.

  5. Use Iterative Prompting

    Prompt Example: "Can you compare the 2018 Toyota Camry and the 2017 Honda Accord in terms of CPRM, safety features, and annual maintenance costs?" Follow-up Prompt: "Now, include a comparison of resale value after five years for both models."

    Why This Works: Breaking queries into smaller, focused prompts ensures precision and allows for iterative refinement to address specific needs.

  6. Leverage GenAI’s Strengths in Consistency

    Prompt Example: "Create a checklist of documents I need to finalize a car purchase from a private seller, including title transfer and registration requirements."

    Why This Works: GenAI’s ability to generate structured, consistent outputs ensures nothing important is overlooked.

  7. Specify Output Format for Clarity and Usability

    Prompt Example: "Provide a step-by-step guide for negotiating the price of a used car, including tips for responding to counteroffers and assessing the seller’s flexibility."

    Why This Works: By specifying a step-by-step format, this prompt ensures actionable and easy-to-follow guidance.

  8. Provide Feedback on Responses for Continuous Improvement

    Prompt Example: "The initial recommendation missed my preference for hybrid vehicles. Can you revise the list to include hybrids under $20,000 with less than 50,000 miles?"

    Why This Works: Providing feedback enables GenAI to refine its outputs, ensuring they better align with your evolving needs.

By leveraging the tools and strategies outlined in Making Choices, Making Money, these examples showcase how GenAI, combined with a structured decision-making framework, can transform the car-buying process into a confident, well-informed experience.


Resources for the Curious

Next are some of my favorite resources on the topic.


  1. Kahneman, D., Sibony, O., & Sunstein, C. R. (2021). Noise: A Flaw in Human Judgment. This book examines the concept of "noise" as unwanted variability in human judgments and decision-making processes, contrasting it with bias, which is a systematic deviation. Kahneman and colleagues provide a framework that helps readers understand how reducing both noise and bias can lead to more accurate and precise outcomes.

  2. S. Zhang, P.R. Heck, M.N. Meyer, C.F. Chabris, D.G. Goldstein, J.M. Hofman (2023) "An illusion of predictability in scientific results: Even experts confuse inferential uncertainty and outcome variability." Proc. Natl. Acad. Sci. U.S.A. 120 (33) e2302491120, https://doi.org/10.1073/pnas.2302491120 (2023). This study explores how experts can misinterpret inferential uncertainty as outcome variability, highlighting the need to clearly distinguish between precision (noise reduction) and accuracy (bias reduction) in scientific research and AI applications.

  3. Tversky, A., & Kahneman, D. (1974). "Judgment Under Uncertainty: Heuristics and Biases." Science, 185(4157), pp. 1124-1131. This classic article is foundational in the study of biases in human decision-making. While it primarily focuses on bias, it provides essential context for understanding why accuracy (freedom from bias) is critical in any decision-making framework, including AI applications.

  4. Geman, S., Bienenstock, E., & Doursat, R. (1992). "Neural Networks and the Bias/Variance Dilemma." Neural Computation, 4(1), pp.1-58. This paper introduces the bias-variance trade-off in machine learning, a key concept that explores the balance between accuracy (bias) and variability (noise or variance) in model predictions. It is a fundamental resource for understanding how to adjust AI algorithms for better accuracy and precision.

  5. Gigerenzer, G. (2008). Rationality for Mortals: How People Cope with Uncertainty. Gigerenzer's work focuses on the practical aspects of decision-making under uncertainty and variability (noise) in human judgment. The book highlights strategies for improving judgment and decision-making by addressing noise, making it a practical complement to AI research on precision.


Comments


Drop Me a Line, Let Me Know What You Think

Thanks for submitting!

bottom of page