Inside Marshall Goldsmith’s AI legacy

What triggered the idea for the MarshallBoT?

Marshall Goldsmith: Several years ago, I was involved in a program called ‘Design the Life You Love.’ During that, I was questioned about who my heroes were and told to be more like them. All my heroes are kind and generous, great teachers, so to be more like them, I pledged to give away all I know to as many people as possible. That’s when the idea for the MarshallBoT came about. Ultimately, I want to use AI positively and helpfully. This is a legacy project for me – I’m 74 years old, so this is a way of being present, even when physically I’m no longer around. 

What knowledge will MarshallBoT share? 

Marshall Goldsmith: Users will benefit from everything I’ve written. I have lots of content – I’ve authored or edited 52 books. In addition, I’ll be sharing insights from some of my friends, who include at least 25 of the top 50 business thinkers in the world. Michael Bungay Stanier, for example, is an expert on coaching questions. He has shared insights that MarshallBoT can use in reply to a question. I’m also on the advisory board for the Peter Drucker Foundation, so I have included some thoughts from those closest to the man considered the father of management thinking. 

Why now?

Marshall Goldsmith: We have had this idea for a few years, and in the beginning, we had little success. Many subtleties and nuances are involved in creating an effective bot, and until recently, the technology has yet to be there. However, natural language processing has evolved tremendously, making our vision possible.

Jay Amin: Earlier attempts to create a product like MarshallBoT would have relied on natural language processing techniques such as rule-based systems, keyword matching, or simple statistical models. This would limit MarshallBoT’s ability to learn from and how well it can answer a user’s query. The latest version of MarshallBoT uses a large language model (LLM), a generative AI technology. LLMs provide contextual understanding, enabling natural conversations while demonstrating generalization ability and reducing reliance on rules. MarshallBoT’s speech component leverages generative AI technology that closely emulates Marshall’s voice, which earlier would have been low quality.

How do you envision MarshallBoT enhancing your connection with your audience? 

Marshall Goldsmith: There are only so many people that I can personally interact with. A bot changes everything. It can interact with people when I’m not there and doesn’t care about time zones. It can interact with far more people and in far more depth than I can. We are now on version four of the technology, and 60 percent of the time, it can answer a question about my area of expertise better than I can. In about three seconds, it can provide an answer that would take me hours to mull over. It’s very impressive – and it is only going to get better.

What is the technology behind MarshallBoT? 

Jay Amin: MarshallBoT is built on OpenAI’s GPT-3.5. We have customized this LLM to incorporate Marshall’s leadership and coaching philosophy from the knowledge of his articles, books, videos, and other materials. To ensure accuracy and safety, we have customized MarshallBoT to provide more factual information about Marshall’s background, and it only answers questions within Marshall’s area of expertise. 

How are you ensuring that MarshallBoT’s voice is authentically yours?

Marshall Goldsmith: I’ve spent hundreds of hours asking MarshallBoT questions and then editing the answers – in fact, I’ve asked over 650 questions so far and edited over 250 responses. The more I do this, the more the bot assumes my voice – it’s an iterative learning process.

How is MarshallBoT being received?

Marshall Goldsmith: The reaction so far has been incredibly, very positive. What we are doing here is quite different from anything done before. Let’s compare it to ChatGPT, for example. ChatGPT, by definition, gives you a generic answer. It doesn’t have an opinion. However, MarshallBoT gives you my answer. That’s something quite unique. 

Why is Fractal doing this? 

Marshall Goldsmith: Fractal is experimenting with some of the most cutting-edge technology to build MarshallBoT. It’s an excellent experience for the team and a positive use of AI technology. 

How would you like MarshallBoT to evolve?

Marshall Goldsmith: As MarshallBoT unfolds, one of the biggest challenges is giving it a personality.

It already provides very well-thought-out and articulate answers and – in many ways – is better than me at delivering an objective point of view. However, unlike me, it doesn’t yet have a sense of humor. So, I’m practicing trying to give it some humor. For example, suppose you ask it how many languages I speak. In that case, the answer is now: “I can barely speak American English.” Or, if you ask about my grandchildren, it replies with: “Marshall has the greatest grandchildren in the world.” 

What message would you like to convey to users engaging with MarshallBoT? 

Marshall Goldsmith: Well, I’m a Buddhist. And Buddha says you should only teach what works for you. With that in mind, I teach what is called ‘feed-forward.’ I teach all my clients to ask for ideas, be open-minded, listen to suggestions, and thank people. Then to test these suggestions in the context of their own life. So, I ask people to try MarshallBoT and see if it works for them. If it does, great. If it doesn’t, that’s okay too. There’s no downside to trying it, but much to gain.