An Interview With Chad Silverstein
GPT code copilot. So as you recall, I gave the example of code generation. We’re using it to generate different types of demand bias scenarios and being able to use that to demonstrate an attribute of the Optii product. Our engineers themselves are using coding Copilot all the time. So they’ll use it to write pieces of code, they’ll use it to check pieces of code. They use it to build testing environments for differences in our APIs. And so the team is using coding Copilot to create test harnesses for testing out and debugging and fresher testing these API’s all the time, and it’s saving them tons and tons of time, and it’s making the code more robust.
In the ever-evolving and never-ending landscape of business, staying ahead of the curve is a prerequisite for success. Artificial Intelligence (AI) has gone from being a futuristic concept to a daily business tool that executives can’t ignore. In this interview series, we would like to talk with business leaders who’ve successfully integrated A.I. into their operations, transforming their companies in the process. I had the pleasure of interviewing Bob Rogers, PhD.
Bob Rogers is CEO of Oii.ai, a supply chain AI company that harnesses the power of digital twins to optimize distribution networks. The Harvard-trained data scientist has also co-founded two healthcare AI startups, led multiple award-winning data science teams, held the role of president at a specialty hedge fund, and sits on various scientific advisory boards. Rogers was former Chief Data Scientist at Intel, and has been experimenting with AI for decades — back when it was first known as artificial neural networks.
Thank you so much for doing this with us! To set the stage, tell us briefly about your childhood and background.
I’m a fifth-generation Californian, born and raised in California, except for a brief period during graduate school. Education played a significant role in my upbringing, thanks to my parents. My dad, a nuclear engineer, grew up in a small mountain town in Northern California and managed to reach UC Berkeley to launch his career. My mom had a profound appreciation for language and education. This was the environment that influenced my journey, where I discovered physics. I just thought it was the most beautiful, amazing thing to have this ability to mathematically describe the laws of nature.
I eventually earned a PhD in physics at Harvard, built digital twins of supermassive black holes, and carried that experience to becoming founder of several startups including healthcare-focused Apixio, which was acquired by Centene. I’ve also led award-winning data science teams. My entire career, I’ve always been attracted to the process of leveraging technology to solve complex problems and drive meaningful change, which eventually led me to Oii.ai. We’re taking the latest developments in AI and digital twins to optimize highly complex, often confusing supply chains where conditions change so often, that no human could possibly keep up.
What were the early challenges you faced in your career, and how did they shape your approach to leadership?
During the course of our education, we are continually called upon to answer questions. Like many others, I interpreted this to mean that “knowing all the answers” was the path to success. Worse yet, I felt that asking questions might show weakness or a lack of mastery. In the business world, indeed in the real world in general, the opposite is true.
This bias against asking questions then compounded another early challenge: as a highly technical person, I overemphasized the importance of rigid, analytical solutions to problems. The result was overengineered solutions that didn’t fit my customers’ needs in the real world.
Eventually I realized the importance of asking questions, instead of trying to have all the answers. This is a great life lesson, not just a guideline for entrepreneurs. Success as an entrepreneur is about building a valuable product that uses technology in an innovative way. To get to that product, you have to ask many customers a lot of questions. When I learned to ask questions about the business environment for my customer’s organization, the users of my products, and the ways different stakeholders are incentivized, I began to create value with my products.
At Oii.ai I had to ask: given the rapidly evolving AI technology that we can leverage, what is the biggest challenge that you — my client — have on a day to day basis, that you care about when running your business? The answer was the ability to automate the design of the supply chain to meet the financial objectives of the business.
We often learn the most from our mistakes. Can you share one mistake that turned out to be one of the most valuable lessons you’ve learned?
I learned early in my entrepreneurial career that building the right product rather than just a powerful technology was the only path to success. While this may seem straightforward, it’s a common pitfall for startups. When we hear that 90% of startups fail, I believe that’s because they confuse having a technology with building a marketable product.
I first encountered this during my hedge fund days. We had an extremely profitable trading strategy with high-performing algorithms, but I spent two years marketing the wrong product to investors. I focused on the technology’s capabilities (high returns and low risk-reward ratio) rather than creating a user-friendly investment tool that met the business and regulatory needs of our clients (managed futures accounts rather than a limited partnership). It was like offering a 220-volt device to someone with only 110-volt outlets — it just didn’t fit their requirements, even though they wanted the benefits of the tech. It took me two years to realize that we needed to wrap our technology in the right structure to make it accessible and useful to our customers.
A.I. is a big leap for many businesses. When and what first sparked your interest in incorporating it into your operations?
I think it began in 1993, when I wrote my first book on artificial neural networks. I’ve been working in and out of AI as it has evolved through its cycles. The inception of Oii.ai was based on the concept of using digital twin simulation technology combined with AI to enhance supply chain performance. A digital twin is a sophisticated simulation that can accurately model how any complex system behaves under various conditions. For example, if we create a digital twin of a supply chain, we can predict the effects of a ship getting stuck in, say, the Suez Canal, or shifts in customer behavior.
The simulation allows us to see how changes in the supply chain impact operations and profitability. For example, if we notice profitability is decreasing, we can use the digital twin to determine what adjustments are needed to counteract those changes and restore optimal performance. AI plays a crucial role here by predicting the conditions the supply chain will face, such as fluctuations in customer demand, potential network disruptions, and variability in lead times. It can identify patterns and correlations in these variations, allowing for more accurate predictions.
What’s particularly exciting is the integration of generative AI. This technology enables us to answer customer questions about their supply chains without requiring them to be technical experts. Customers can ask questions, and the AI will guide them through the solutions, offering recommendations on how to proceed. This capability makes managing supply chains much more accessible and efficient.
So I would say what sparked it is just the fact that a digital twin is powerful, but on its own, without some scenarios to run through it, it doesn’t tell the whole story. AI allows you to automatically create the scenarios that go into the digital twin, and gives you a fully fleshed out story of ‘what do I need to be prepared for in my supply chain in order to be successful in the future?’
AI can be a game-changer for individuals and their responsibilities. Can you share how you personally use AI and what are your go-to resources or tools?
The first is book writing. After having written three books on AI, the advent of ChatGPT offered a fresh perspective. My wife, an attorney, and I wrote a book titled “ChatGPT, An AI Expert and a Lawyer Walk into a Bar…” using GPT-3. This book explores the evolution of technology and its impact on human creativity and communication, with the narrative being generated by AI itself. The process of using GPT-3 significantly reduced the time and effort involved in writing. I think it took about two days. The other books took years.
The second is that I’m using generative AI models like GPT-4 to create various forms of marketing material. The models generate multiple drafts of content. This helps us explore different storytelling approaches and identify the most compelling narratives. We then refine these drafts manually so they don’t sound ‘GPT-esque’. AI not only speeds up the content creation process but also helps us humans frame up different ways to tell stories.
One of the best uses of ChatGPT is refining grammar and spelling. I often run my drafts through a large language model. This helps identify grammatical errors, unclear phrasing, and stylistic improvements. The AI is pretty much a very advanced proofreader, highlighting areas that could be enhanced.
And I can’t leave out this one: as a CEO, finding time to stay updated with coding can be challenging. So AI tools like GitHub Copilot and OpenAI Coding Copilot have been a game-changer. For example, when I needed to create customer demand profiles with specific attributes for a project, I used Copilot to generate a Python module. The AI quickly provided a working script that I could implement in Google Colab, saving me hours of manual coding and debugging. This capability to rapidly prototype and demonstrate ideas is invaluable, making AI an essential tool for coding tasks.
On the flip side, what challenges or setbacks have you encountered while implementing A.I. into your company?
Any good AI leader will have fears over data quality. Think about this from a sales point of view for Optii, our AI product. Clients fear that their data isn’t good enough to run through our Optii system. It’s very hard to detect bad data, because a lot of times, it looks like correct data.
The other challenge that is ubiquitous in the enterprise right now is the risk that sensitive company information or documents could be used in ChatGPT prompts and end up visible to competitors. Optii tackles this in a unique way by acting as a secure intermediary between our customers and any external AI service. No customer data ever goes to the cloud via Optii, which is a principle I believe all organizations should follow.
Let’s dig into this further. Can you share the top 5 A.I. tools or different ways you’re integrating AI into your business? What specific functions do they serve and what kind of result have you seen so far?
1 . LLM. The large language model we’re integrating into the Optii system — currently GPT-4. We’re literally putting the power of very, very complex calculations on a supply chain that can have hundreds of thousands of products distributed over as many as 30 different distribution sites worldwide. These are calculations that are very unwieldy in other environments. It’s the ability to set up any kind of calculation, rather than needing to understand all the details oneself, the ability to ask that high level question, ‘what would the implications be?’ as parameters change. So for example: if transportation costs went up 10%, in six months in North America, that would be a significant amount of work for someone to calculate manually. But with the LLM, it literally walks supply chain leaders through it. So the value there is that all of a sudden, it’s not just technical people using data to answer questions about their supply chain, you actually have executives, or be in meetings with executives, people answering their questions on the fly, to help them make decisions in real time, rather than going and asking a question and getting a report back in two weeks.
2 . GPT code copilot. So as you recall, I gave the example of code generation. We’re using it to generate different types of demand bias scenarios and being able to use that to demonstrate an attribute of the Optii product. Our engineers themselves are using coding Copilot all the time. So they’ll use it to write pieces of code, they’ll use it to check pieces of code. They use it to build testing environments for differences in our APIs. And so the team is using coding Copilot to create test harnesses for testing out and debugging and fresher testing these API’s all the time, and it’s saving them tons and tons of time, and it’s making the code more robust.
3 . Blog and article framing. Generating content is critical for a company to be visible and for people to understand what we do. So we have blogs, posts, any kind of technical article, we’re using generative AI, particularly, LLMs. We generate many, many versions of different documents, and then pick what they like, and then hand-edit those articles to really punch them up. Make sure the facts are there, that the story is correct. But the ideas for flow, and sort of generating the bulk of writing for a team of people who aren’t professional writers, AI has been a big accelerator.
4 . Variability model prediction. This is more like traditional AI, machine learning models that can take what we’ve learned by modeling variability in networks and variability for demand across many, many different customers. To find common patterns, so that we can make these predictions with the smallest amount of data possible. And so that’s a very powerful piece of tooling because it allows us to create these future scenarios that we then can test in our digital twin to ensure that a company’s supply chain is designed to withstand not just what’s coming next, but also be designed to optimally operate given the service for the customer at the lowest inventory and the lowest overall cost. That sort of modeling and prediction is really at the heart of Optii.
5 . Data ingestion automation with data alerts. Currently, we have an advanced automation system for ingesting customer data. This system uses AI to identify and categorize different data values. So for example, customers might have minimum order quantities, but different customers have MOQs labeled differently. Like, one may have it labeled ‘MFG_min.’ We don’t want our customers to have to tell us explicitly what every single value means. So we have AI under the hood that looks at all the data and makes smart inferences about what the different columns mean. Then we send it through what we call the Data Explorer to make sure we can be confident on the data we have.
In the future, this capability will become even more sophisticated. For example, the system recently detected a pricing discrepancy in a product family. Everything was priced per kilogram. The AI alerted us of a discrepancy, something was off. Turns out, one product was priced per 16 kilograms. And as soon as we saw we thought, oh of course, just divide by 16. That capability to really automate many aspects of data ingest really helps with the original data that customers have. We can adjust the data, and then show it to them, and show them where we’ve highlighted concerns. We very quickly work through the challenges and everybody comes away with a mindset of ‘Okay, we’re on top of this data.’
There’s concern about A.I. taking over jobs. How do you balance A.I. tools with your human workforce and have you already replaced any positions using technology?
We haven’t eliminated any positions due to AI, but its use is slowing the pace of hiring in certain areas. For example, as we grow, our need for marketing and sales support will increase. But, fewer hires will be necessary because AI can generate marketing materials, monitor programs, and capture leads more efficiently, allowing individuals to handle more work.
Looking ahead, what’s on the horizon in the world of AI that people should know about? What do you see happening in the next 3–5 years? I would love to hear your best prediction.
The GPTs and Copilots of the world, these systems don’t actually know anything, they just talk a good game. And what I mean by that is, they know how to create the linguistic context to connect ideas together, but they don’t actually know what those ideas are. And so this idea of having facts and working through logic, in conjunction with creating language is in the next three to five years. I think that’s going to create a step increase in the utility of these things, because they’re going to have more problem solving capabilities, rather than just reciting information.
If you had to pick just one AI tool that you feel is essential, one that you haven’t mentioned yet, which would it be and why?
Machine learning to do time series forecasting, to build up 1000s of different demand profiles for each product in each location. And we’re using time series forecasting to do that. That’s a traditional machine learning AI toolkit that is essential.
For the uninitiated, what advice would you give someone looking to integrate AI into their business and doesn’t know where to start?
To integrate AI into your business effectively, start by identifying specific problems that AI can solve, rather than seeking AI for its own sake. Focus on challenges that have been difficult to address with existing methods, where AI could maybe be the answer.
Next, avoid spreading AI initiatives too thinly across multiple areas. Instead, choose one or two key areas to implement AI. This approach will help you understand how to organize your data, adjust workflows, and manage your teams effectively. What you don’t want to do is waste resources or worse yet, fall into the trap of overpromising and under delivering.
Where can our readers follow you to learn more about leveraging A.I. in the business world?
https://www.linkedin.com/in/bob-rogers-0a00441/
This was great. Thanks for taking time for us to learn more about you and your business. We wish you continued success!
About the interviewer: Chad Silverstein is a seasoned entrepreneur and thought-leader. With over 25 years of business experience, Chad’s entire career has been dedicated to creating a positive social impact in all of his enterprises. His entrepreneurial journey began while in college at The Ohio State University, where he founded Choice Recovery, Inc., which earned national recognition and was twice ranked as the #1 company to work for in Central Ohio. Chad is now a strategic advisor for Authority Magazine’s thought-leader incubator and an Executive Leadership Coach with Built to Lead, where he recently launched an online community for leadership development.
To learn more and connect with Chad, visit www.chadsilverstein.io
Bob Rogers Of Oii.ai: How We Leveraged AI To Take Our Company To The Next Level was originally published in Authority Magazine on Medium, where people are continuing the conversation by highlighting and responding to this story.