Guest: Doug Kimball, Chief Marketing Officer at Ontotext
Hosts: Seth Earley, CEO at Earley Information Science
Chris Featherstone, Sr. Director of AI/Data Product/Program Management at Salesforce
Published on: September 22, 2023
In this episode, Seth Earley and Chris Featherstone speak with Doug Kimball, Chief Marketing Officer at Ontotext, about the rapidly evolving intersection of knowledge graphs, ontologies, and generative AI. They explore common misconceptions around graph technology, how organizations can connect data investments to tangible business outcomes, and the critical role of data governance when deploying large language models. Doug shares real-world examples from e-commerce personalization and supply chain management, explaining why knowledge graphs are a powerful enhancement to existing data infrastructure rather than a costly replacement.
Key Takeaways:
Quote of the Show:
"Knowledge graphs are not a rip and replace, they are an add to/enhancement of." - Doug Kimball
"If you don't have a framework in place that ensures when I ask a question, I'm getting answered from a good, accurate, trusted, verified source or sources of data, then you run the risk of garbage in, garbage out." - Doug Kimball
"A physical supply chain is also a digital supply chain - it's an information supply chain. You're moving physical goods, but with those goods you have to move data." - Seth Earley
Tune in to discover how knowledge graphs and generative AI work together to unlock hidden value in enterprise data - and why governance, not technology, is the real differentiator.
Links:
Ways to Tune In:
Thanks to our sponsors:
Podcast Transcript: Knowledge Graphs, Generative AI, and Taking Control of Your Data
Transcript Introduction
This transcript captures a conversation between Seth Earley, Chris Featherstone, and Doug Kimball about the strategic value of knowledge graphs and ontologies, how they enhance rather than replace existing data infrastructure, and why pairing them with generative AI and strong data governance is critical for organizations seeking meaningful business outcomes.
Transcript
Seth Earley: Welcome to our podcast this month. My name is Seth Earley.
Chris Featherstone: I'm Chris Featherstone.
Seth Earley: Today, really excited about our guest, who is an expert in knowledge graph technology. He's a CMO, he's very passionate about B2B retail supply chain data. He's a tech evangelist who focuses on e-commerce, customer centricity, and getting insights from data to support business management. And he's very passionate about enabling digital business. Doug Kimball, Chief Marketing Officer of Ontotext. Welcome to the show.
Doug Kimball: Thank you. So glad to be here. I've been enjoying our back and forth conversations for almost a year, and to be part of this process is pretty exciting. I have fun when I get to talk to businesses about how technology can be used to make their lives better using data. It's been a fun full circle journey to end up here.
Seth Earley: It's interesting because knowledge graphs have been around for a while. The term ontology has become a little more mainstream in the last few years - it was really something that caused executives' eyes to glaze over or to walk out of meetings. But now people are starting to realize the importance. When it comes to knowledge graphs, maybe you could talk about the misconceptions. What do people not understand about knowledge graphs and ontologies? Could you start with a working definition and talk about what people don't understand?
Doug Kimball: Sure. There's a lot that people don't understand about graph technology and knowledge graphs. One of the things we encounter a lot is: is this a rip and replace? In other words, here you come talking to me about another database structure, another architectural process - I've already got SAP, or Hadoop, etc. Do I have to make all these massive changes? I've already spent this money.
And really, knowledge graph technology and knowledge graphs are not a rip and replace. They're an add to - an enhancement of - in order to bring all that data together that they've already been using. They have all these disparate systems: SAP, IT infrastructure, Hadoop, databases, unstructured data. Nothing has to change in where they are putting things or how they store their data. It's more like inserting a better gear - like dropping a 10-speed gear into your existing engine to make it run better, faster, stronger.
When you take an ontology and layer it on top, you're taking all this great data and layering on top of that the knowledge of how to communicate about it, how to use it, how to access it, how to apply context to make better decisions. This isn't a plug-in and off you go - it does take time and it is a serious practice. But you're not giving up all the great money you spent on something to do something new and shiny. You're doing it better.
Chris Featherstone: Yeah - especially when you go to an executive or a stakeholder. Why should they even care? And once they do care, what's that investment going to look like?
Doug Kimball: It's a strategic tool. We've heard that leaders - CEOs, CIOs - have been told "we need more data." Now we've got all this data. What do we actually do with it to monetize it, to get better personalization, to reduce resource usage, to streamline to outcome-driven results? That's where knowledge graph technology comes into play. You're adding this powerful connected framework on top of all the data structures you already have. Data by itself is great. But what do you do with it?
From a C-level conversation standpoint: drive top-down. What are the results we're trying to achieve because of having better data? Understand the business problems, then drive down to the technology to make it happen.
Seth Earley: It reminds me of that old BASF commercial - "we don't make the products you buy, we make the products you buy better." It's an enabler. But sometimes there's a big challenge because it's infrastructure, it's a capability, it's foundational. It's hard to connect that directly to a business outcome. How do you talk to stakeholders and executives about what they can actually do with it?
Doug Kimball: Let me use product information management as a topic - I'll up-level that to e-commerce, and even further to customer acquisition and personalization. If I'm a business - B2B or B2C, and those lines blur so much these days - how do you provide the most customized, engaging, personalized experience for your customers? You need the right data served up to the right people at the right time, with the right connections.
That's where the power of knowledge graph technology comes in - its ability to do what we call inferences or reasoning, to start making connections among the data that aren't explicitly there. Knowledge graph technology can say, "Doug likes to shop for these kinds of things in June." Then it takes it an abstraction level deeper - it starts making inferences and says, maybe we should serve up something different to him because his shopping habits are evolving.
Seth Earley: What else needs to be in place for knowledge graphs to enable personalization? The knowledge graph is one component - what are the other pieces?
Doug Kimball: You have to have good, clean, disambiguated, deduplicated, accurate, enriched data. Just knowing that Doug Kimball purchases things online is one example. But how do you verify that this Doug Kimball is at this address - not the other Doug Kimball living in California? You've got to have that organized.
You know - customer 360, product asset 360 - all of those things come down to having the data mastered. In a relational database you have a series of things. In knowledge graphs, you can go in different directions and add different perspectives to add to the depth. Without good, clean data governed properly with a metadata layer, you can make queries faster - so what?
Seth Earley: Exactly. We built an information architecture for personalization for an organization years back. When it came down to talking about what is the difference between this segment and that segment, and what content is going to resonate more effectively - you didn't have that answer. You can build a great infrastructure, you can have great data, but if you don't have maturity in your supporting processes and real knowledge of your customer, it falls short.
That works in conjunction with things like a customer data platform, product information management system, content management system. There's usually some kind of orchestration engine as well, right?
Doug Kimball: Completely agree. A knowledge graph by itself is a really powerful database. But all data is just data until you do something with it.
Here's an example: back in 2005, Hurricane Katrina hit New Orleans, and so many people moved to other states. Retailers were struggling because they had a whole new demographic mix. People who moved to Texas suddenly wanted the spices they used to cook Louisiana food - but retailers had stocked for the old demographics. Now imagine if you'd had knowledge graph-powered solutions back then - to know that Seth moved from Louisiana and likes these things, and to start serving that content up. That's an opportunity for profit, volume, and share.
Chris Featherstone: How often are the folks you work with asking about AI?
Doug Kimball: A lot - the easy, short answer. When I first joined Ontotext in early November, machine learning was kind of the hot thing. AI as a more generalized term was the big conversation, because ideally you want a knowledge graph to power the algorithms behind machine learning to feed that properly. So AI as part of what we do with graph technologies is extremely important.
Fast forward - generative AI, of course, everybody has been exploding with this. The best indicator of the interest we've seen: we did a webinar about a month ago discussing generative AI and knowledge graphs and large language models, and had our largest attendee signup ever - over 800 people. And the actual attendee numbers were significantly higher than I expected. That tells me not only that generative AI and LLMs are hot topics, but people want to know how knowledge graphs plug in to support and train large language models. People were very engaged, and I continue to get follow-ups from myself and our salespeople.
Seth Earley: Can you speak more specifically to the connections between the knowledge graph and large language models - exactly how that's being done?
Doug Kimball: Right now it's mostly proof of concepts. We've done it ourselves - we've got an enhancement in our recent releases connecting ChatGPT to our knowledge graph. We also have what's called the Ontotext Knowledge Graph, where we're using our own corpus of knowledge to feed back, connecting it with large language model concepts to serve our own content up more intelligently.
To me, the ability of knowledge graphs to make relationships, to have that understanding of things, to find context within the data - gives me reassurance that what comes back from a large language model isn't just hallucinations. By having that context applied and verified, you have the potential to get better accuracy and more efficacy out of the large language model, as opposed to it just mining everything.
Chris Featherstone: Are there things that concern you about it - things you're advising clients on what not to do?
Doug Kimball: Data governance is an easy one. You've got all this great information out there, and data governance and compliance jumps out at me immediately. How do you ensure compliance and verification - how do you know this is the right answer? If you're just using AI for marketing language and it says something pretty and fun, it's probably okay. But if you're looking for very specific answers to a business-critical question, you can't accept poor answers driven by non-governed data.
If you don't have a framework in place that ensures when I ask a question I'm getting answered from a good, accurate, trusted, verified source or sources of data, then you run the risk of garbage in, garbage out. What are the regulatory requirements? What kind of risk factor are you willing to accept?
Seth Earley: What are the biggest mistakes organizations or the C-suite are making when it comes to AI, large language models, generative capabilities?
Doug Kimball: It goes back to asking: are they reacting to market pressures? My competitor is doing something with fill-in-the-blank technology - ChatGPT - we better do something too. There's rationale behind that because market share is important. But just jumping into it without a plan is never going to be a good result.
You need to first ask internally: what are we trying to get out of this? Document and agree on what it is you're trying to accomplish. Go back to what are the outcomes. Put that in place so you have a business-driven need being solved - not just a data-driven need. Both are important, but data supports the business outcomes.
Seth Earley: Are there any places where you're seeing specific success or traction - either with knowledge graphs or knowledge graphs and generative AI?
Doug Kimball: Travel was one that I found most interesting and relevant. Travel sites using AI to help you plan a trip - you put your parameters in, and it spits out an itinerary. I played with it and started customizing: we like these sorts of things, we don't like long walking tours, and it made some pretty interesting recommendations. Then the fun part - it made commercial recommendations for vendors you could use. It gave three taxi options, three great places to eat when you're in Crete. Now there's a "so what" - there's commerce being driven through that.
Seth Earley: Are you seeing success with conversational commerce - asking questions of product catalogs?
Doug Kimball: That's exactly what I'm seeing. We've developed what we call a demonstrator for our product information management - it's an enhancement to work with product information more efficiently. It connects ChatGPT with an AI mindset to feed and ask the right multi-level questions of your data to get the right answer. That's where I'm starting to see activities - especially with product information management types, digital asset management - to have information categorized, organized, and queryable using AI tools.
Seth Earley: What concerns you moving forward - in terms of how this is evolving, the messaging, adoption?
Doug Kimball: The biggest struggle we continue to have is awareness and understanding. I've used the term "echo chamber" - I don't want us to only sell to people who already get it. If we only sell to people who get it, we're not going to be talking to the larger business perspective or the enterprise mindset. We're not going to be helping to educate how data can be used and reused across an organization using graph technology structure.
Going back to - this is not a rip and replace, this is an enhancement perspective - those are two of my biggest challenges. Generating leads is functionally easier. Getting somebody to say "I get it, and I understand at a high level how this might help either my business units or my business challenges" - that's harder to do, because it's so potentially powerful.
Seth Earley: When you talk about supply chains, I've always said a physical supply chain is also a digital supply chain - it's an information supply chain. You're moving physical goods, but with those goods you have to move data - not just about the journey and logistics, but the product specs, end usage, all of that. It's a really complex ecosystem.
Doug Kimball: When I was at JDA - which is now Blue Yonder, the largest supply chain company in the world - I introduced a concept called the Supply Chain Grid. It was a way to explain that moving an item from point A to the finished product doesn't always take a straight line. In fact, a lot of times it can't, because disruptions happen. We visualized your supply chain as a grid with multiple connection points and opportunities. And then you manage your business with that federated approach - it gave you a lot more risk mitigation, more failover opportunities, more ways to get to the customer.
It's been neat to look back at the diagram I created then and see how it mirrors knowledge graph visualizations.
Chris Featherstone: I'd love to get your perspective on always fine-tuning your graph environments - how do you keep them relevant?
Doug Kimball: Text analytics is part of that. You've got your catalog of products and information, you run text analytics on all the different product categories and descriptions, and then you use mining - looking at how often topics show up, comparing that to search and browse or query activity. If people are searching on certain terms regularly but those terms aren't in your product descriptions, you tweak. You don't have to do that every day - it doesn't change that much. But regular tweaking and re-analyzing how you're organizing and presenting information based on data you have is important.
Progress over perfection in everything we do. I was in the Coast Guard for 10 years, and our motto was semper paratus - always ready - which we often translated to "semper paralysis." Always getting ready to get ready. I see a lot of organizations stuck there - unable to start because they have to prepare more first.
Seth Earley: So let's turn a little bit toward who you are and where you're from. Do you want to give us the world according to Doug?
Doug Kimball: My journey has been one of evolution. My master's degree is actually in counseling education - I spent 7 years as a professional counselor doing personal and career counseling. How I ended up in this role, some days I shake my head. But I think the counseling serves me. When I'm at trade shows - who are you, what's your problem, what are you challenged by, how do we help you move past that, let's put together a solution plan - it's very similar.
I worked for the Nielsen Company for almost 10 years, then migrated through jobs in different fashions - product management, product development, sales, product marketing - and had the opportunity to join Ontotext. It's been neat just to watch it all come together.
The coolest thing is I have a wife who's listened to me go on about data for a long time, and she gets it. One day we're driving, I was trying to explain knowledge graphs to her, and she said: "Knowledge graphs are like sentences, and MDM is like letters of the alphabet." I got really quiet. I said, really? And she said yes - and I expanded on it. And I thought, that is my line.
Chris Featherstone: Your parents were anthropologists, correct? What was that like?
Doug Kimball: Growing up around two PhDs with different backgrounds and interests made me really look at how people and societies communicate and evolve - how sociology plays into all that, and how it truly is about being a community. Which is like bringing data together.
Seth Earley: What do you do outside of work?
Doug Kimball: I've got 23 and 24-year-old twins - one of each, boy and girl. Hit the jackpot - they're both employed, doing well out of college. My hobby is Brazilian jiu-jitsu. I've done that coming up on 16 years now. Got my black belt almost 2 years ago. I teach two to three times a week and train when my body says it's okay - I turned 59 a couple of weeks ago. I love it because it allows me to have a bad day at work, go and get worked for an hour, and walk out feeling reset. It also keeps the mind sharp because at an elevated level, you're not just making move one, move two, move three. You're doing 3-dimensional, active chess.
Seth Earley: Do you apply your martial arts philosophy into business on a day-to-day basis?
Doug Kimball: Jiu-jitsu makes you comfortable in uncomfortable situations. I can walk into any business environment, any trade show, any place, and I might be a little nervous - new people, topics I don't know - but nobody's going to try to break my arm or choke me. There's not a 300-pound person laying on me trying to squash me. It gives you perspective.
Seth Earley: If you could go back in time to when you were graduating college, what might you say to yourself?
Doug Kimball: Be more focused. I went straight into the military after high school, and I knew I wasn't mature enough for college at that point. When I got to college the first year or two, I was not as focused as I should have been. I'm not a natural learner in the traditional sense - I like to pick things up and really hammer them into my head. Being more focused would have made a big difference.
Seth Earley: Well, Doug, it's really been a pleasure having you. I've really enjoyed our conversation. Thank you so much for spending your time with us today. We'll have your contact information in the show notes - you can be found on LinkedIn at Doug Kimball, and we'll have your Twitter and company address all in the show notes. Again, thank you so much for your time today. It's been great to chat.
Doug Kimball: Cool. Thanks.
Chris Featherstone: Awesome.