Using Generative AI to Create a Generative AI Strategy

GenAI consulting and teaching organizations to fish

Dan McCreary
11 min readFeb 10, 2024
We can use generative AI tools to build a first draft of an organization’s overall AI strategy and roadmap—image by the author and ChatGPT.

This article describes how you can use generative AI tools such as ChatGPT to help you build a generative AI strategy customized to your organization’s needs. This process is instructive not only to see how vulnerable your organization is but it can give you insights into how to lead change. This story exemplifies how prompt engineering skills can help you get the most value from a generative AI tool. It also shows techniques to decompose large problems into smaller parts using decomposition, a powerful way to get actionable suggestions.

We are inspired by the quote:

“Give a person a fish and they will be happy for a day. Teach a person to fish and they will be happy for a lifetime.”

This article will start with simple prompt suggestions and slowly build up to more complex prompts that generate detailed roadmaps to help your organization increase its generative AI capabilities.

A prompt refinement strategy that uses context, decomposition, and drill down to generate a detailed generative AI roadmap.

In summary, we will illustrate three steps:

  1. Step 1: Start with simple general prompts
  2. Step 2: Use the output of Step 1 to put additional context in a new prompt
  3. Step 3: For each part of Step 2, add more context and get a detailed list of actions you can take

Step 1: Starting With A Simple Prompt

Let’s begin with a simple prompt:

Create a generative AI strategy that is customized to the needs of my organization.

This prompt will generate a very generic response. It will only help you a little but might guide you to the next step. Here is a typical response. I have done some minor editing to keep the response precise for this blog.

ChatGPT (GPT-4) Response

To create a customized generative AI strategy for your organization, I’ll need more specific information about it, its goals, and its challenges. Here are some key points to consider when developing a generative AI strategy:

  1. Understanding Your Organization’s Needs and Goals:
  • What is the primary industry and focus of your organization?
  • What are the specific goals your organization aims to achieve with AI? (e.g., automating processes, enhancing customer experience, generating new content, data analysis).

2. Identifying Use Cases for AI:

  • What are the areas where AI can be most beneficial in your organization?
  • Are there specific tasks or processes that can be optimized using AI?

3. Data Management and Analysis:

  • What type of data does your organization currently have access to? (e.g., textual data, images, customer interactions)
  • How can this data be used to train AI models effectively?

4. Technology and Infrastructure:

  • What is your organization's current technological infrastructure?
  • Will you develop AI solutions in-house or collaborate with external AI vendors?

5. Compliance and Ethical Considerations:

  • Are there regulatory constraints related to data and AI in your industry?
  • How will your strategy address ethical considerations related to AI, such as bias, transparency, and accountability?

6. Skill Development and Team Composition:

  • Does your organization have employees with AI and data science skills, or is there a need for training or hiring?
  • How will you structure your team to integrate AI capabilities best?

7. Implementation Plan:

  • How will AI projects be prioritized and implemented?
  • What are the short-term and long-term milestones?

8. Monitoring and Evaluation:

  • How will you measure the success and impact of AI implementations?
  • What mechanisms will you have in place for ongoing monitoring and optimization?

By addressing these points, we can create a more tailored AI strategy. Feel free to provide more details or ask for guidance on specific aspects of the strategy.

Step 2: Adding Context to Your Prompt

You can see that giving GPT-4 a general prompt will only return with a generic result. You can’t implement this strategy directly, but it outlines the questions that need to be answered to get to the next level of detail. You can also add the sentence: Ask clarifying questions. to get specific questions that ChatGPT needs to create custom strategy outlines and roadmaps.

To illustrate the following result, I am on the school board of a school in a rural Minnesota school district that is concerned about the lack of progress in leveraging generative AI in the school district.

Sample Prompt with Context

Create a generative AI strategy plan customized to my school district's needs.
I am on the school board of a rural Minnesota school district.
Our mission is to prepare our students for jobs that will be important 10–20 years from now.
We believe that generative AI can generate highly customized lesson plans for each student.
Our schools have fast internet access, and our students each have Chromebooks.
Our school district has a very old, brittle computer system for managing student enrollment, generating student transcripts, and collecting and reporting on grades. It has no AI capabilities.
Our teachers don’t use generative AI much to generate custom lesson plans. We want to encourage more use of generative AI to help with their administrative tasks, such as generating custom lesson plans.

Sample ChatGPT Response

Given your school district’s context and objectives, here’s a customized Generative AI Strategy Plan to help leverage AI to enhance the educational experience and prepare students for future careers.

Generative AI Strategy Plan for [Your School District]

1. Vision and Objectives

  • Primary Goal: Leverage generative AI to create personalized learning experiences and streamline administrative tasks.
  • Long-term Objective: Prepare students for future job markets by integrating advanced technology into their learning journey.

2. Current Infrastructure Assessment

  • Strengths: Fast internet access and availability of Chromebooks for each student.
  • Challenges: Outdated computer system for managing student-related administrative tasks.

3. AI Implementation Areas

Educational Enhancement

  • Customized Lesson Plans: Develop AI systems that analyze students’ learning styles, progress, and interests to generate personalized lesson plans.
  • Interactive Learning Tools: Introduce AI-driven tools for interactive learning experiences (e.g., virtual labs, simulations).
  • Career Preparedness Programs: Implement AI modules that provide insights into future career paths and necessary skills.

Administrative Efficiency

  • System Upgrade: Prioritize the modernization of the current student management system with AI capabilities for better data handling and analysis.
  • Automated Administrative Tasks: Utilize AI for automating routine tasks like attendance tracking, grading, and report generation.
  • Teacher Support Tools: Offer AI-powered tools to assist teachers in curriculum planning, student assessment, and personalized feedback.

4. Training and Development

  • Teacher Training Programs: Conduct workshops and training sessions on utilizing AI tools in education.
  • Student Exposure to AI: Introduce basic AI concepts in the curriculum to familiarize students with this technology.
  • Professional Development: Encourage continuous learning for educators to stay updated with AI advancements.

5. Partnership and Collaboration

  • Tech Partnerships: Collaborate with tech companies specializing in educational AI solutions.
  • Community Engagement: Involve community stakeholders in understanding and supporting AI integration in schools.

6. Ethical and Compliance Considerations

  • Data Privacy: Ensure all AI solutions comply with data privacy laws and school policies.
  • Bias and Fairness: Monitor AI tools for biases and ensure equitable access to technology for all students.

7. Implementation Roadmap

  • Short-Term (1–2 Years): Start with pilot programs for AI in lesson planning and administrative tasks. Begin system upgrade.
  • Medium-Term (3–5 Years): Expand AI usage in educational tools and administrative systems. Integrate AI topics in student learning.
  • Long-Term (5–10 Years): Achieve full integration of AI in personalized education and administrative efficiency. Regularly update and refine AI systems.

8. Monitoring and Evaluation

  • Performance Metrics: Set benchmarks for student engagement, teacher workload reduction, and administrative efficiency.
  • Feedback Loops: Regularly gather feedback from students, teachers, and administrators to continually improve AI integration.

9. Budget and Funding

  • Budget Allocation: Allocate funds for AI software, training programs, and system upgrades.
  • Grants and Funding: Explore grants and funding opportunities specific to technology in education.

10. Risk Management

  • Technology Dependence: Balance the use of AI with traditional teaching methods to avoid over-reliance on technology.
  • System Reliability: Ensure robust support and maintenance for AI systems to minimize disruptions.

This strategy plan is a starting point and should be adapted as you progress and learn from the implementation phases. Regular review and adjustment of the strategy will be crucial to its success.

Step 3: Getting Specific: Drilling Down for Concrete Actions

The prior ChatGPT response is a good overall outline of things to consider for the district. But there still needs to be a lot of detail added. To go to the next level of detail, you can take one or two sentences from this plan and use ChatGPT to give you specific ideas. Because I have been personally involved in creating MicroSims for schools, that is one topic I will do a deep dive into.

Example of a Drill-Down Detailed Prompt

I love this suggestion:

Interactive Learning Tools: Introduce AI-driven tools for interactive learning experiences (e.g., virtual labs, simulations).

Other school districts in our area are already creating small interactive lesson-specific simulations using ChatGPT and p5.js. They are using community volunteers to generate these programs called “MicroSims”. They have already created hundreds of MicroSims and they are all freely available to use and share without fees. How can we encourage our teachers and volunteers to create and reuse MicroSims for each lesson? We need to create a strong sense of urgency to do this now and not delay this step for several years.

ChatGPT Response

Creating and integrating “MicroSims” into your school district’s curriculum is an innovative approach to enhancing the educational experience. To encourage teachers and volunteers to develop and reuse these interactive simulations and to instill a sense of urgency, consider the following multi-faceted strategy:

1. Awareness and Education Campaign

  • Informational Workshops: Organize workshops to educate teachers and volunteers about the benefits and impact of MicroSims in enhancing learning experiences.
  • Success Stories: Share case studies or success stories from other districts to illustrate the positive outcomes of using MicroSims.

2. Collaboration and Community Building

  • Collaboration Platform: Establish an online platform for educators and volunteers to collaborate, share ideas, and access existing MicroSims.
  • Community Events: Host hackathons or development marathons where teachers and volunteers come together to create MicroSims.

3. Incentivization and Recognition

  • Recognition Programs: Implement a recognition system for contributors, like ‘Teacher Innovator Awards’ or ‘Volunteer of the Month’.
  • Professional Development Credits: Offer professional development credits to teachers who actively engage in creating or using MicroSims.

4. Integration into Curriculum Planning

  • Lesson Plan Incorporation: Encourage teachers to include MicroSims in their lesson plans, demonstrating how they align with learning objectives.
  • Curriculum Committees: Involve curriculum committees in endorsing and advocating for using MicroSims.

5. Resource and Support Provision

  • Technical Support: Provide technical support for teachers and volunteers developing MicroSims.
  • Training Sessions: Offer training sessions on tools like ChatGPT and p5.js for those unfamiliar with these technologies.

6. Demonstration and Pilot Programs

  • Pilot Classes: Run pilot classes where MicroSims are used extensively, and document the outcomes to showcase their effectiveness.
  • Showcase Events: Organize events where teachers can demonstrate how they have used MicroSims in their classrooms.

7. Communication and Regular Updates

  • Regular Updates: Keep the school community updated on the progress and success stories of MicroSim implementation.
  • Feedback Channels: Establish channels for receiving feedback from teachers and students on the MicroSims.

8. Policy and Administrative Support

  • Administrative Endorsement: Seek endorsement from the school board and district leadership to emphasize the initiative’s importance.
  • Policy Integration: Work towards integrating the creation and usage of MicroSims into the school district’s educational policy.

9. Setting Clear Goals and Timelines

  • Goal Setting: Set clear, achievable goals for the number or type of MicroSims to be developed within a specific timeframe.
  • Urgency in Communication: Emphasize the urgency and importance of this initiative in all communications.

10. Building Partnerships

  • Partner with Universities or Tech Companies: Collaborate with higher education institutions or tech companies for expertise and resources.

11. Monitoring and Feedback Loop

  • Impact Assessment: Regularly assess how MicroSims are affecting student learning and engagement.
  • Iterative Improvements: Use feedback to make continuous improvements in the MicroSim development process.

By implementing these strategies, your district can rapidly scale up the development and use of MicroSims, ensuring that students benefit from these advanced learning tools as soon as possible.

Generating Assessments Forms

A screen image of an HTML form for assessing how vulnerable your organization is to disruption by competitors using generative AI. Image by the author.

ChatGPT is also very good at generating assessment forms to help you quantify your relative risk (based on your industry), as well as forms to help you do assessments of where you stand in the generative CMM models. The trick is to generate multiple-choice questions that categorize your organization. ChatGPT can generate the multiple choice questions and the HTML forms, style the forms, and generate the JavaScript code to create an overall score based on the responses. This same technique can be used to calculate risk-of-distribution vulnerability scores.

When to Bring In Consultants

The complexity staircase metaphor. At the lower level are simple concepts like educating your staff and leadership on the capabilities of generative AI. At the top are difficult questions, such as building internal knowledge graphs to predict the consequences of different strategic decisions—image by the author and ChatGPT.

I deliberately picked a simple example of a small rural school district because it does not have the complexities of many organizations with complex legacy IT infrastructure that needs to be replaced to meet the needs of AI-driven industries like healthcare and education. School strategies often focus on teacher, mentor, and volunteer professional development. However, it would be naive to think that generative AI can create the final draft of a strategy and roadmap for your organization.

In practice, many organizations don’t have the internal expertise to integrate generative AI into their products. They need a much more sophisticated generative AI plan that includes in-depth architectural tradeoff analysis (ATAM). Most school districts don’t have the budgets to build knowledge graphs that recommend lesson plans and MicroSims. They will use outside services and AI-driven learning management systems as they become more common.

If you need help assessing questions such as “Can we build our knowledge graph to accelerate recommendations?” you might need to bring in an outside consultant. These consultants should know how to use the Capability-Maturity-Models to improve your internal generative AI capabilities, create graph embeddings, and use concept indexes and vector stores. They should have experience putting together generative AI centers of excellence. There are specific tools to help organizations find the right ways to represent knowledge so that generative AI systems are cost-effective. Older brittle relational database systems have fundamental scalability limitations, and newer graph databases are well aligned to these tasks. Please let me know if I can help you with these challenges.

Conclusion: Landing Your Draft Version 1

A happy group of senior business leaders who have learned how to use ChatGPT to generate the first draft of their generative AI strategy and roadmap. But can they do it all? Image by ChatGPT and the author.

Using generative AI tools such as ChatGPT is a great way to generate ideas for your generative AI strategy and to create an outline of the first draft. Combining sound prompt engineering and decomposition skills can often create a solid first draft for your generative AI strategy. You can also use generative AI to help you create a clear sense of urgency and to lead change. You must create a clear future vision with detailed steps on the roadmap to success. You need to help everyone know their roles in these steps, and you can also use generative AI to create a sense of urgency so people take action.

Acknowledgments

I use generative AI tools dozens of times daily for all aspects of my work and home. It seemed intuitive that everyone would use GenAI to create the first draft of their GenAI strategies and roadmaps. But after talking to many people in our Applied AI Strategy meetup, it was clear that people needed a bit of guidance. I thank several AI strategy and architecture experts for encouraging me to create this article. Specifically, Arun Batchu, Matt Versaggi, Steve Peterson, and especially Michael Arulfo encouraged me to continue writing. This topic will likely come up at future events.

--

--

Dan McCreary

Distinguished Engineer that loves knowledge graphs, AI, and Systems Thinking. Fan of STEM, microcontrollers, robotics, PKGs, and the AI Racing League.