Author: Daniel Koger

  • A Prompt to Challenge Your Thinking

    Abstract blue technology-themed banner representing artificial intelligence, data, and digital information.

    Sometimes the most useful response from AI is not agreement.

    When we develop an idea, make a decision, or reach a conclusion, it is easy to focus on the information that supports what we already think.

    I use AI to help me look at the other side.

    Instead of asking:

    “Help me make the case for this.”

    Try asking:

    “What might I be missing or getting wrong?”

    AI can help you look at your idea from another perspective—not to prove you wrong, but to help you test whether your reasoning holds up.

    The goal is not to let AI make the decision for you. The goal is to challenge your thinking before you act on it.

    Sample Prompt

    Remember: the information inside square brackets [ ] is for you to replace with information specific to your situation.

    <prompt>
    I am considering the following [idea, decision, conclusion, or approach]:
    [Describe what you are considering and why.]
    Before helping me strengthen it, challenge my thinking.
    Consider whether:
    - I am making assumptions that may not be correct.
    - There are reasonable arguments or perspectives I have not considered.
    - Important information may be missing.
    - There are circumstances where my conclusion or approach may not work.
    - Additional evidence could cause me to reconsider my conclusion.
    Identify the strongest reasonable challenges to my thinking.
    For each challenge:
    1. Explain why it matters.
    2. Distinguish established facts from assumptions or speculation.
    3. Identify what additional information or evidence would help evaluate it.
    Do not challenge my thinking simply for the sake of disagreeing. Focus on reasonable concerns that could affect my decision or conclusion.
    Do not invent evidence or present speculation as fact. If something requires verification, identify it clearly.
    I will evaluate the challenges and decide whether they should change my thinking.
    </prompt>

    Why This Works

    It is easy to use AI to strengthen an idea you already have. The problem is that AI may simply become better at supporting the conclusion you gave it.

    This prompt asks AI to do something different: look for reasonable reasons your thinking may be incomplete.

    It can help you:

    • Question assumptions you may be making.
    • Consider other reasonable perspectives.
    • Identify information that may be missing.
    • Recognize circumstances where your approach may not work.
    • Determine what additional evidence could affect your conclusion.

    The goal is not to generate disagreement. It is to give your reasoning another test before you rely on it.

    Before You Use the Results

    Not every challenge AI identifies will be valid or important.

    Review the response and determine which concerns actually apply to your situation. Verify factual claims using appropriate source material or an expert, especially when the information could materially affect your decision.

    You do not need to change your conclusion simply because AI challenged it. A good challenge may strengthen your original conclusion, cause you to modify it, or show you that you need more information before deciding.

    AI can help challenge your reasoning. You remain responsible for evaluating the evidence and making the decision.

    Next in This Series

    Why Context Changes the Answer

    Our next article moves beyond individual prompts to explore techniques for getting more reliable AI results.

    Return to the Practical and Responsible AI Series Index

  • A Prompt to Find Gaps and Unanswered Questions

    Abstract blue technology-themed banner representing artificial intelligence, data, and digital information.

    You can spend a lot of time developing a plan and still miss something important.

    I have run into this myself. When you have been working on something for a while, you know what you meant. You understand the background, the assumptions you made, and the things you have already considered. That can make it difficult to recognize what you haven’t explained—or what you may have overlooked.

    It is one reason I have created specialized GPTs in the past—to give me another set of eyes.

    But you don’t need to create a GPT to get that benefit.

    Sometimes a simple prompt is enough:

    “What important questions have I not answered?”

    The goal is not for AI to fill in the missing information. It is to help you recognize what may still need your attention.

    Sample Prompt

    Remember: the information inside square brackets [ ] is for you to replace with information specific to your situation.

    <prompt>
    Review the following [plan, document, idea, or proposal].
    Identify important questions I may not have answered.
    Look for:
    - Information that is unclear or missing
    - Assumptions I may be making
    - Decisions that have not been made
    - People or groups that may be affected
    - Constraints or dependencies I may have overlooked
    - Exceptions or situations I may not have considered
    Organize the questions by category.
    For each question:
    1. Explain briefly why it matters.
    2. Identify the relevant section or statement, if available.
    3. Indicate whether the information is missing, unclear, or based on an assumption.
    Do not answer the questions by making assumptions.
    If the information is already provided, point me to it rather than identifying it as a gap.
    Do not invent problems or treat every possible question as equally important.
    I will decide which questions require further investigation or action.
    Material to review:
    [Paste your plan, document, idea, or proposal here]
    </prompt>

    Why This Works

    This prompt asks AI to look for what may be missing, rather than simply summarizing what is already there.

    It can help you:

    • See assumptions that may not be obvious to someone else.
    • Identify decisions or information that are still missing.
    • Recognize people, constraints, or situations you may have overlooked.
    • Turn potential gaps into questions you can investigate.

    Most importantly, the prompt tells AI not to fill in the gaps by making assumptions.

    Organizing the questions by category can also make the results easier to evaluate, particularly when reviewing a lengthy document or complex proposal.

    Before You Use the Results

    Review each question and decide whether it is actually relevant.

    AI may identify something that you have already considered, something that does not apply to your situation, or something that simply is not important.

    If a question does reveal a real gap, determine what information you need and verify the answer using appropriate source material or an expert.

    AI can help identify questions you may have missed. You remain responsible for deciding which questions matter and how they should be answered.

    Next in This Series

    A Prompt to Challenge Your Thinking

    Return to the Practical and Responsible AI Series Index

  • A Prompt to Turn Notes into an Action Plan

    Abstract blue technology-themed banner representing artificial intelligence, data, and digital information.

    Meeting notes can capture good ideas, decisions, and follow-up items, but they do not automatically become a useful action plan.

    A common approach is to give AI a set of notes and ask, “What should we do next?” The problem is that AI may turn discussion points into decisions, assign owners who were never assigned, or invent dates that were never agreed upon.

    A better prompt asks AI to organize what is already in the notes while keeping missing information visible.

    Sample Prompt

    Remember: the information inside square brackets [ ] is for you to replace with information specific to your situation.

    <prompt>
    Review these meeting notes and turn them into an action plan for [team or audience].
    Organize the agreed actions into a table with these columns:
    - Action
    - Owner
    - Target date
    - Dependencies or prerequisites
    - Supporting evidence from the notes
    Use only actions that were clearly agreed upon.
    Do not turn suggestions or discussion points into decisions. Do not assign an owner or date unless one appears in the notes.
    If an owner, date, or other important detail is missing, label it "Not specified."
    After the action table, provide separate sections for:
    1. Decisions that still need to be made
    2. Suggestions or ideas that were discussed but not agreed upon
    3. Questions requiring follow-up
    At the end, identify the items that need clarification before the action plan is complete.
    Do not invent details. Where possible, use a brief quotation or note reference to show what supports each action.
    Meeting notes:
    [Paste meeting notes here]
    </prompt>

    Why This Prompt Works

    • It separates agreed actions from ideas that were only discussed.
    • It preserves owners and dates when they are actually stated.
    • It makes missing information visible instead of allowing AI to fill the gaps.
    • It provides supporting evidence so team members can verify the proposed actions.
    • It turns unstructured notes into something a team can review and use.

    Before You Use the Action Plan

    • Compare the actions and decisions with the original notes.
    • Confirm that owners and dates were actually stated rather than inferred.
    • Check that an important dependency, qualification, or unresolved question was not lost.
    • Review the items marked “Not specified” with the appropriate team members.

    AI can organize the work, but the people involved remain responsible for confirming what was actually decided.

    Next in This Series

    A Prompt to Find Gaps and Unanswered Questions

    Return to the Practical and Responsible AI Series Index

    Learn More

    OpenAI Prompting Best Practices

  • Comparing Two Options: Let AI Help You Decide

    Abstract blue technology-themed banner representing artificial intelligence, data, and digital information.

    Have you ever been shopping and narrowed your choice down to two products?

    You could ask AI:

    “Which one is better?”

    But better depends on what matters to you.

    Suppose you are shopping for a night light. You have found two that you like, but you are having trouble deciding between them.

    Instead of asking AI to make the decision, tell it what matters to you and ask it to compare the products using those criteria.

    Sample Prompt

    Remember: the information inside square brackets [ ] is for you to replace with information specific to your situation.

    <prompt>
    I am considering these two night lights:
    Night Light 1: [paste product name or link]
    Night Light 2: [paste product name or link]
    Compare them based on:
    - Price
    - Brightness in lumens
    - Expected useful life
    - Design and appearance
    - Additional functionality
    Use the same criteria for both products.
    If information for either product is unavailable,
    tell me it is unavailable rather than
    making an assumption.
    Summarize the important differences,
    but do not choose one for me.
    </prompt>

    Why This Works

    The question is no longer simply, “Which one is better?”

    You have told AI what better means to you.

    The same technique works when comparing computers, appliances, software, insurance plans, contractors, or other choices.

    The criteria will change, but the principle remains the same:

    Compare both options using the same criteria.

    AI can help you understand the differences. You decide which differences matter.

    Before You Make a Decision

    AI may not have complete or current information about the products or services being compared.

    Verify important specifications, prices, and other claims against reliable sources before making a purchase or commitment.

    Next in This Series

    A Prompt to Turn Notes into an Action Plan

    Return to the Practical and Responsible AI Series Index

  • A Prompt to Rewrite Technical Content for a Business Audience

    Abstract blue technology-themed banner representing artificial intelligence, data, and digital information.

    Technical information is often written for people who already understand the subject. It may include specialized terminology, implementation details, acronyms, and assumptions that are familiar to the technical team.

    A business audience usually needs the same facts presented from a different perspective.

    They may need to understand:

    • Why the subject matters
    • How it affects the organization
    • What risks or opportunities are involved
    • What decisions need to be made
    • What actions should happen next

    Simply asking AI to “make this less technical” can create a problem. Important qualifications, constraints, and risks may disappear along with the technical language.

    A better prompt defines what may change and what must be preserved.

    In the sample prompt below, text enclosed in square brackets [ ] represents information you should replace with details specific to your request.

    For example, replace [business audience] with the intended readers, such as senior executives, project managers, or department leaders. Replace [desired length] with a limit such as 200 words, one page, or five brief paragraphs.

    Sample Prompt

    Rewrite the following technical content for [business audience].

    The revised version should help the readers understand the business purpose, organizational impact, risks, decisions, and required actions.

    Use clear business language. Replace unnecessary jargon and acronyms with plain language. When a technical term is necessary, explain it briefly.

    Preserve:

    • Facts and conclusions
    • Requirements and constraints
    • Risks and dependencies
    • Names, dates, amounts, and measurements
    • Important qualifications and exceptions
    • Decisions that have already been made

    Do not introduce benefits, conclusions, or recommendations that are not supported by the original content.

    If the source contains an unclear statement or an undefined technical term, identify it rather than guessing what it means.

    Organize the response into:

    1. Business Purpose
    2. Business Impact
    3. Risks and Dependencies
    4. Decisions Needed
    5. Recommended Next Steps
    6. Questions Requiring Clarification

    Limit the response to [desired length].

    Before completing the response, compare it with the original and confirm that simplifying the language did not change the meaning.

    AI can help translate technical language, but the subject-matter expert remains responsible for confirming that the translation is accurate.

    Example Prompt

    <prompt>
    Rewrite the following technical content for
    [business audience].
    The revised version should help readers understand
    the business purpose, organizational impact, risks,
    decisions, and required actions.
    Use clear business language. Replace unnecessary
    jargon and acronyms with plain language. Explain
    technical terms briefly when they are necessary.
    Preserve:
    - Facts and conclusions
    - Requirements and constraints
    - Risks and dependencies
    - Names, dates, amounts, and measurements
    - Important qualifications and exceptions
    - Decisions that have already been made
    Do not introduce benefits, conclusions, or
    recommendations unsupported by the original content.
    If a statement is unclear or a technical term is
    undefined, identify it rather than guessing.
    Organize the response into:
    1. Business Purpose
    2. Business Impact
    3. Risks and Dependencies
    4. Decisions Needed
    5. Recommended Next Steps
    6. Questions Requiring Clarification
    Limit the response to [desired length].
    Before completing the response, compare the result
    with the original and confirm that simplifying the
    language has not changed its meaning.
    Technical content:
    [Insert content approved for use with the AI tool]
    </prompt>

    Next in This Series

    A Prompt to Compare Two Options

    Return to the Practical and Responsible AI Series Index

  • A Prompt to Summarize a Long Document

    Abstract blue technology-themed banner representing artificial intelligence, data, and digital information.

    A long document can contain valuable information, but finding its purpose, decisions, risks, and open questions can take time.

    A common first attempt is simply to ask AI, “Summarize this document.” That may produce a useful result, but the instruction leaves an important question unanswered: useful to whom, and for what purpose?

    A better prompt tells AI what to preserve, how to organize the answer, and where uncertainty must remain visible.

    Sample Prompt

    Read the attached document and prepare a concise summary for [audience].

    Include:

    • The document’s purpose
    • Its major conclusions or findings
    • Decisions that have already been made
    • Significant risks or issues
    • Action items, owners, and dates stated in the document
    • Unresolved questions or decisions still needed

    Separate facts stated in the document from any inferences you make. Do not add information that is not supported by the document. If something is unclear or missing, say so. When possible, identify the page or section that supports each major point.

    Limit the summary to [desired length] and use clear headings and bullets.

    Why This Prompt Works

    • It identifies the audience, which helps determine vocabulary and level of detail.
    • It defines the information that should survive the compression.
    • It separates documented facts from AI inferences.
    • It asks AI to expose missing or unclear information instead of quietly filling gaps.
    • It requests source locations so important statements are easier to verify.

    Before You Use the Summary

    • Confirm that the document is appropriate to submit in the AI environment you are using.
    • Compare important conclusions, decisions, names, amounts, and dates with the original document.
    • Check whether the summary omitted an exception, qualification, or unresolved issue that matters to the reader.

    AI can shorten the reading process, but it does not remove the need to read the source when the decision matters.

    Example Prompt

    <prompt>
    Read the attached document and prepare a concise summary for an executive leadership team.
    The readers need to understand the document well enough to discuss its implications and determine whether any decisions or follow-up actions are required. Use business language and explain technical terms only when they are necessary to understand the conclusions.
    Include:
    - The document's purpose
    - Its major conclusions or findings
    - Decisions that have already been made
    - Significant risks, concerns, or dependencies
    - Action items, including any owners and dates stated in the document
    - Unresolved questions or decisions still needed
    Separate your response into the following sections:
    1. Executive Summary
    2. Major Conclusions
    3. Decisions Already Made
    4. Risks and Concerns
    5. Action Items
    6. Unresolved Questions
    Separate facts stated in the document from any inferences you make. Label each inference clearly.
    Do not add information that is not supported by the document. If information is missing, contradictory, or unclear, identify the issue instead of attempting to fill the gap.
    When possible, provide the page number or section heading that supports each major conclusion, decision, risk, and action item.
    Limit the response to approximately one page. Use concise language, clear headings, and bullets where appropriate.
    Before completing the response, verify that names, dates, amounts, decisions, and assigned responsibilities agree with the source document.
    </prompt>

    Next in This Series

    A Prompt to Rewrite Technical Content for a Business Audience

    Reference Basis

    The prompt structure reflects OpenAI guidance to provide clear instructions, relevant context, and visible boundaries between instructions and source material.

    OpenAI — Prompt Engineering

    Return to the Practical and Responsible AI Series Index

  • When Replacement Was Not the Right Solution

    Work Experience series banner introducing career stories, professional lessons, and reflections from a career in technology and business.

    One of the easiest mistakes in technology is to assume that replacing an existing system is the same as solving the customer’s problem.

    During an assignment, I traveled to Oslo, Norway, as a presales solutions architect.

    The company I was working with provided hosting services to small businesses and individual customers. Its existing hosting environment was built around Linux and other open-source technologies.

    The company had an established customer base that was well served by its existing technology. It wanted to expand its services by adding capabilities based on Microsoft technologies.

    From the technology vendor’s perspective, the opportunity initially appeared straightforward: replace the existing environment with its own products.

    But that was not what the customer actually needed.

    Its existing customers were loyal to the services and technologies they already used. Forcing those customers onto a different platform could have put much of the hosting company’s existing business at risk.

    The real question was not, “How do we eliminate Linux?”

    It was, “How do we add new capabilities without destroying what already works?”

    My architectural recommendation was to keep the existing Linux environment for the customers who wanted it and add a separate environment for customers whose needs were not currently being met.

    The client was very pleased with the recommendation. It differed from the original expectations, but insisting on a complete replacement could have jeopardized the opportunity.

    By allowing the technologies to coexist, the hosting company could protect its existing business while expanding into a new market. The technology vendor could also gain business it otherwise might not have won.

    That experience reinforced an important lesson: the responsibility of a solutions architect is not simply to promote a preferred product.

    It is to understand the customer’s business, preserve what is already working, and recommend the solution that creates real value.

    Sometimes the best solution is not replacement.

    Sometimes it is coexistence.

  • When the Lesson Plan Was Not the Real Need

    Work Experience series banner introducing career stories, professional lessons, and reflections from a career in technology and business.

    Working in technology gave me the opportunity to experience many different cultures and to see how people lived and worked in countries with very different resources.

    One experience that comes to mind was traveling to Huế, Vietnam, in 1997 to teach database design at a hospital.

    The hospital had requested the training materials in advance. When I arrived, I discovered that the doctors had already read and memorized them—and had used the books to create databases without receiving any instruction.

    At that time, computer access within the hospital was limited primarily to doctors, and they were understandably proud of the opportunity to use the technology.

    Before the hospital had an electronic database, patients were responsible for bringing their written medical records with them to every visit.

    The doctors’ effort was impressive, but the books alone had not taught them how to connect database-design principles with the actual work of the hospital. They had theory on paper, but no instructor to help them translate that theory into a practical design.

    The problem was not a lack of intelligence or effort. They had done remarkably well with the resources available to them. The problem was that reading a book was not a substitute for seeing how information moved through the hospital and understanding the work the database needed to support.

    So we set the books aside.

    For the next two weeks, we walked through the hospital with the nurses. We asked them to show us what they did, how information moved through their work, where problems occurred, and how electronic data might make their jobs easier.

    The focus shifted from teaching database theory to understanding the people, processes, and needs that the database was supposed to support.

    If we could make life easier for the nurses, we could improve the experience of the patients and, as a result, help the doctors as well.

    We did not accomplish the original lesson plan.

    But we accomplished what was needed.

    That experience reinforced a lesson I have carried throughout my career: good data design does not begin with a textbook, a tool, or even a data model.

    It begins with listening to the people who perform the work and understanding the problem that needs to be solved.

  • When Loyalty and Integrity Collide

    Work Experience series banner for Career Reflections, featuring professional experiences, leadership lessons, and insights gained throughout a career in technology and business.

    Three experiences across three decades taught me that loyalty to an employer should never require compromising personal integrity.

    Over the course of my career, I have been asked to do many things I disagreed with. That is part of working for an organization. You do not always agree with management, and you do not always get your way.

    But there is a difference between disagreeing with a decision and being asked to do something you believe is dishonest.

    Three times in my career, that distinction cost me my job.

    The first was in 1989. I was working for a technology company when management wanted us to tell customers that our profit margins were smaller than they actually were.

    I had no problem declining to disclose the company’s margins. That was confidential business information, and customers were not necessarily entitled to it. But if a customer asked me directly, I had two acceptable choices: decline to answer or tell the truth.

    Making up an answer was not one of them.

    I did not spend much time debating what to do. I simply continued doing my job honestly, despite the company’s guidelines. Seven of us ultimately lost our jobs in circumstances involving our Christian convictions about honesty.

    The second incident occurred in 2000 while I was working for a consulting company.

    A regional vice president wanted me to tell a prospective customer that I had experience with a particular database technology. There was only one problem: I had never worked with it. In fact, I had never even heard of it.

    I was perfectly willing to learn the technology. Throughout my career, I have had to learn new systems, tools, and approaches. That is part of working in technology.

    But learning something new and claiming experience you do not have are two very different things.

    I refused to misrepresent my qualifications and was fired for being disloyal.

    This time, the consequences were more personal. The regional vice president who fired me was also a friend. I was leaving behind people with whom I had worked and developed friendships over many years. Some of those friendships have continued for more than twenty-five years.

    It was a difficult departure, not because I was uncertain about telling the truth, but because of the relationships involved.

    It is also worth making a distinction. This was the decision of an individual leader, not something I would attribute to the consulting company as a whole.

    The third incident occurred in 2018.

    By then, I had spent decades working in technology, including management and leadership roles. One of the responsibilities of a manager is evaluating the performance of the people who report to him.

    The CIO began pressuring me to give one of my direct reports a negative performance review.

    The problem was that the employee had done nothing to deserve one. His work had consistently been very good, and I had no legitimate basis for giving him a poor evaluation.

    The pressure continued until I finally told the CIO that I could not, in good conscience, give the man a bad review.

    The employee was fired. A short time later, I was fired as well.

    By that point in my career, I had already experienced the consequences of refusing to compromise my integrity. I knew that standing my ground could cost me my job.

    There was no hesitation.

    I cannot speak to every consideration behind those decisions. I can only speak to what I was asked to do and why I refused.

    Looking back, these three situations involved different responsibilities.

    In the first, it was honesty toward customers. In the second, it was honesty about my own capabilities. In the third, it was honesty toward someone who had been entrusted to my leadership.

    But the underlying question was the same.

    What does loyalty require of us?

    I believe employees owe their employers loyalty. When we accept a position, we take on responsibilities to the organization, its customers, its leadership, and the people with whom we work.

    We should protect confidential information, support legitimate business decisions, and do our work to the best of our abilities.

    But loyalty is not blind obedience.

    A manager who asks an employee to deceive a customer is not protecting the company. A consulting leader who misrepresents someone’s qualifications is creating expectations that may never be met. And an executive who pressures a manager to give an undeserved negative review is undermining the trust on which effective leadership depends.

    In each case, the request may have appeared to serve an immediate business objective. But the long-term cost of dishonesty can be far greater than whatever short-term advantage it provides.

    My Christian faith has always shaped how I approach these decisions. I do not claim to have handled every situation in my career perfectly. Far from it. But there are some lines I have tried not to cross, and knowingly telling a lie is one of them.

    There is another part of these stories worth mentioning.

    After each termination, I eventually found a better job.

    That is something I am grateful for, but it was never guaranteed. When I made those decisions, I did not know what opportunity would come next, how long I might be unemployed, or what the financial consequences would be.

    Over the years, I have come to believe that loyalty and integrity should reinforce one another, not compete with one another.

    Sometimes the most loyal thing you can do for an organization is respectfully refuse to do something that would compromise its integrity.

    And sometimes that refusal comes at a personal cost.

    I have never regretted telling the truth. I would have regretted lying.

  • The Advice That Changed My Career

    My career in technology began before technology became my job.

    In 1987, I was selling parts. My largest customer was a relatively new company that provided industrial maintenance services. They ordered a considerable amount of product from me every week.

    Whenever I visited, I also found myself helping them with their computer.

    One day, the shop steward took me aside and said:

    “Dan, you are a lousy salesman, but we keep buying from you because every week, when you show up, you fix our computer. You need to change careers.”

    It was blunt advice—but it was also good advice.

    So I changed careers.

    I left a job I hated and took a chance on doing work I loved. Technology eventually became a career that took me around the world, introduced me to remarkable people, and allowed me to spend decades solving problems.

    I have loved the work ever since.

    Sometimes another person recognizes an ability in us before we recognize its value ourselves. That shop steward saw something in me that I had not yet learned to see.

    Sometimes the safe decision is not the right decision.

    And sometimes, you have to take a chance and do what you love.