A Step-by-Step Guide to Developing AI Co-Thinkers with ChatGPT

Author: Paolo Cervini

Integrating AI Co-Thinkers into management practices presents unprecedented opportunities for innovation. The journey from my September LinkedIn article on ManagementGPT to the recent Capgemini Invent & Thinkers50 whitepaper featuring three AI Co-Thinker prototypes has been insightful. ?Meanwhile, OpenAI's launch of the GPT Builder and, then, the GPT Store has set a new standard for developing GPTs or customized apps based on generative AI.

An AI Co-Thinker aids in navigating complex issues across strategy, innovation, and culture. It leverages additional content, methodological guidelines, and behavioral instructions – essentially, a Prompt Strategy.

Based on practical experiments, this guide helps navigate the journey of developing a valuable AI Co-Thinker, a journey structured into four main phases:

I. Initial considerations and set-up

• Existing AI solutions: Begin by exploring the GPT Store to identify AI solutions already available in your field. Understanding what's out there helps avoid reinventing the wheel and provides inspiration for your own development.

• ChatGPT performance: Test its understanding of your specific subject. Challenge ChatGPT (ideally the Plus version) with questions and scenarios to identify any existing gaps that your AI Co-Thinker could fill.

• AI Co-Thinker's development goals: Set clear, short and mid-term goals, including business impact, social contribution, branding, monetization, integration with other software.
• Target users: Understand who will most benefit from the AI Co-Thinker, ensuring the tool meets their specific needs and challenges.
• Unique value proposition: Identify what can make your AI Co-Thinker unique. This could be proprietary data, your brand, specific industry insights, or unique integration capabilities.

II. Design and development

• Flexibility vs. Structure: When using ChatGPT Builder, the challenge often lies in maintaining discipline, as its outputs can vary widely. ?While ChatGPT excels at providing creative and flexible responses, it's crucial to guide the GPT behavior to ensure more consistent outcomes, for instance by referring to a file containing very detailed prompts and rules.
• Conversational and engagement design: Craft the AI Co-Thinker for effective, engaging communication by incorporating thoughtful questions that guide the conversation. Aim also for concise, clear responses that drive towards solutions.

• Sustainability and ethics: Incorporate sustainability, ethics, and social responsibility from the start, ensuring alignment with broader values and societal issues. See my Dialogue Review article Harnessing AI for Cooperative Advantage, co-authored with Leon C. Prieto, Ph.D. , Simone Phipps, PhD , Gabriele Rosani .

• Collaborative process: Design sequences that will include and / or trigger human interactions and collaboration, enhancing transparency and collective decisions.

III. Testing and iteration

• Iterative feedback loop: Engage a selected user group for feedback, refining your AI based on real-world application and user needs. This iterative approach proved valuable for the Platform Network Effect AI Co-Thinker, tested by researchers and practitioners connected to the Boston University Platform Strategy Research Symposium.
• Exploration of AI tools: Look also beyond OpenAI's GPT Builder, assessing various AI platforms to identify those that best meet your requirements and goals. For example, MindStudio seems quite interesting.
• Direct control: Keep full access to your AI's prompts and instructions for continuous refinement based on feedback. Do not delegate them to “third” parties, as that would slow down the iteration process.
• Cross-GPTs synergies: Explore how the AI co-thinker can complement and enhance other AI systems and GPTs. See, for instance, the Apta GPT developed by Gianni Giacomelli to identify specific business subprocess that can be enhanced by generative AI and/or by predictive AI.

IV. Implementation and scaling

• Customization to company data: Tailor your AI to specific datasets and processes, mindful of security, IP, and legal considerations.
• From insights to actions: Ensure that your AI provides actionable insights and can integrate smoothly with existing systems through APIs. This enhances its practicality and utility in real-world scenarios.
• Automation of the Co-Thinker inputs: Streamline and automate the processes that feed information to your AI. This increases efficiency and ensures that your AI Co-Thinker has access to the most current and relevant data.
• Merging generative and predictive AI: Combine the creative problem-solving capabilities of generative AI with the predictive power of traditional AI models. This holistic approach will provide a more nuanced and comprehensive tool.

In summary, developing an AI Co-Thinker is a continuous journey that adapts to changing business landscapes and technologies. I recommend embracing flexibility and responsiveness to keep the Co-Thinker a valuable asset for innovative decision-making and effective management practices.

I encourage you to dive into the development of your own AI Co-Thinker and share your experiences, challenges, and successes. Together, we can lead the way in harnessing the power of AI for smarter and more ethical management practices.