AI powered OrgChange ChatGPT now available
Announcing the OrgChange 2.0 ChatGPT with more training content and additional configuration instructions. Now available to anyone who is an OpenAI subscriber.
https://chat.openai.com/g/g-GwWbdo8Jc-orgchange-4the-new-era
People without an OpenAI subscription can post their organizational change questions at https://www.futocracy.network/spaces/13292204/feed and I will run it through this custom ChatGPT for you. Some benchmarking tests are also posted at this link showing improvements from the recent changes to the GPT configuration.
This organizational change ChatGPT is designed to help managers & leaders struggling with disengaged employees, conflicts between organization silos, and slow & disjointed decision-making. The GPT is designed for developing flexible and purpose-driven organizations with engaged employees. What worked well for traditional hierarchical organizations operating in the stable Industrial Era no longer works for the New Era that needs resilient and agile organizations capable of routine adaptation. This requires new thinking that drives the GPT content filtering for how we organize for work, lead our organizations, and implement change as a routine activity.
Technical foundation — the Futocracy OrgChange ChatBot is built on the OpenAI ChatGPT4 platform using additional training content and an enhanced custom instruction set for its configuration. An additional third layer of ChatGPT instructions will come from your personal custom instructions and how you choose to phrase your question including an interactive dialogue you can have with the AI.
Supplemental knowledge for use by the ChatGPT comes from –
· LinkedIn newsletter — New Era Organizations
· Substack articles — Management Minefields
· Medium articles — New Era Organizations
· eBook — Leading New Era Organizations
The custom configuration instructs the GPT in its use of our uploaded content and how to filter content in the OpenAI knowledgebase to skew toward post-Industrial Era thinking.
Lessons learned while developing this custom ChatGPT –
· ChatGPT sits on an OpenAI knowledge base with a user interface that has three parts. The first part is the obvious query that can be very complex in instructing the GPT in how to respond. The second part that all users have is a set of Custom Instructions and Profile that provide supplemental information to the query for how the GPT is to respond. And the third part is the document set that the user can upload with the query to supplement the OpenAI knowledge base. The OrgChange ChatBot (a custom GPT) sits on top of the user’s custom instructions and query, adding additional instructions (in the configuration) and up to 20 documents for the AI to use in responding to the query.
· The custom ChatGPT does not appear to add uploaded documents to the OpenAI knowledgebase but treats the supplemental documents as newly uploaded content to use when a query is made. Concise writing seems to integrate into the OpenAI knowledgebase better than abstract text.
· The interface for building a custom configuration is a normal ChatGPT conversation. This is an excellent way to quickly build a custom ChatGPT but starts to have problems as updated instructions are added after the initial GPT is built. The new instructions will sometimes overwrite or modify earlier instructions. As I modified the initial custom configuration with the interactive chat, the results degraded to the point that I decided to start over and thereafter modified the configuration manually.
· Help is not easy to find when trying to figure out how the custom instructions work in ChatGPT. There are five ChatGPT focused groups on LinkedIn, but all are inactive. The OpenAI FAQ seems to work for new users or advanced developers with little help for anyone building a custom GPT. So far Google searches have been best finding some answers on Reddit.
· The custom ChatGPT is improved in the 2.0 version but does not always use the Futocracy supplied knowledge content or additional instructions in the configuration. Part of this may be due to the “memory” the GPT has of previous interactions that were done for testing. The results may be better for users new to the custom ChatGPT.
· The custom ChatGPT did hallucinate once and clearly made up a response that was obviously incorrect. In researching this hallucination, I believe the GPT made an improper connection between “Dual-track Agile” and “Kotter’s Dual Operating System” and made up a new “DAC” in the connection — Dual Agile Core Leadership. In doing so, the GPT overlooked many references to Direction-Alignment-Commitment (DAC) Leadership both in the content provided by Futocracy and what is easily found with a Google search.
