Get Started >

254: Tap AI to Democratize Your Business with Jeff Berkowitz

Jeff Berkowitz, co-founder and CEO of Delve Deep Learning, is driven by his passion to help businesses clone themselves into an AI, enabling them to navigate complexities and create prosperity through innovation.

We learn about Jeff’s journey from politics and public affairs to launching Delve Deep Learning, an AI-native company revolutionizing public affairs intelligence. He explains his AI Innovation Process, which includes training AI to be your intern, testing and selecting the best tools, deciding whether to be a buyer or builder, launching for external use, and iterating on the process. He also highlights how this approach helps businesses streamline workflows, scale services, and democratize access to AI-powered public affairs intelligence.

—

Listen to the podcast here

 

Clone Yourself into an AI with Jeff Berkowitz

Good day, dear listeners, Steve Preda here with the Management Blueprint podcast. And my guest again, after four years, is Jeff Berkowitz, co-founder and CEO of Delve Deep Learning, an AI native tech company building the future of public affairs intelligence. Welcome back to the show, Jeff.

Thanks. It’s great to be here. I didn’t realize it’d been four years. Time flies when you’re having fun.

Maybe three and a half, but thereabouts. And you were representing a different company at that time because Delve Deep Learning is a startup that you guys launched since. And we will talk about that. But I’d like to start with a question asking you about your personal “Why” and what are you doing in your new business to manifest it?

Yeah, I mean, for me, my passion is really comes down to free enterprise. I’ve always really believed that when we give entrepreneurs and business leaders the opportunity to flourish, they create prosperity and opportunity for everybody and make our society better. And first sort of built on that passion in politics and government and the past, 14-15 years stumbled into entrepreneurship myself. And really both delve and now Delve Deep Learning have been focused on how do we help companies the insights they need to navigate all these complexities and pressures as they try and build opportunity and prosperity for our society.

Wow, that’s pretty big, sounds pretty lofty. So tell me a little bit about the idea. So why did you come up with the Delve Deep Learning? What triggered launching the company and how did you end up basically disrupting your own business?

We did. We have done a little bit of disruption of our own business here. But look, I think one of the key things that business leaders have to do is keep their eyes open for the inflection points in our society and technology and business. Share on X And we try and do that for our clients. What are the inflection points that you’ve got to navigate in policy debates and regulatory debates? But when we saw AI coming to the forefront, generative AI with ChatGPT 3.5, and really becoming something that was ready as a technology to leverage, it was super interesting to us. And we really dug into what does this mean for our industry, for our work, and saw an opportunity to really build on many of the things we’d love to have done with technology for a long time, but it wasn’t possible before.

Okay, so what is it? How does AI help you create a better experience for your customers or clients?

Yeah. So at Delve, we saw our research consultancy and we have a whole team of really smart research analysts that are spending all day digging through what’s happening in policy debates and what are different stakeholders saying, how is that analyzing, how does that impact companies, industries, different policy issues. And it’s very qualitative work. You can’t just throw traditional machine learning or data science at that and get meaningful results. For our client base, it was not. They couldn’t do what we do in-house, right? Because most of them can’t build out a research shop. But what we saw with AI is we can train these AI models in very much the same way that we train our research team. Many of our researchers start in our associate program. They learn how to analyze, collect all the information, synthesize it, translate it into something meaningful and actionable for our clients. And that research process and training process, you can do the same thing with AI models. So I like to joke, for 20 years, people have been showing me technology platforms they thought would blow us away. And I would look at it and say, my intern kicks this thing’s butt. And AI platforms came along with the models, generative models, and I said, this can’t beat my intern, but it can be our intern. And we can train it in that same way. And so we started to think about, well, what does that mean, and what is it ready for, and what is it not ready for, and how does it make our lives easier?

That’s super interesting. So, let’s talk about that because what you did there is you really leveraged AI. A lot of people talk about ChatGPT and leveraging AI and a lot of AI tools coming out, but you rarely talk to a person that who actually did it in a business, which is a professional service business. So we’re not talking about big tech or Silicon Valley startup, you had a professional service business and you put the business on top of it. So what is that process? And I think you called it the AI innovation process. So how can someone innovate with the use of AI, such as ChatGPT on top of a professional service firm? You develop the process for this. And the first step was, I think you already covered it, which is let’s train AI to be your intern. So what are the other steps?

Yeah, so we really talk with a lot of our clients who are trying to think through these issues themselves. What’s your short, medium, and long-term strategy for leveraging AI? It is here, right? If you step back for a minute and you think about 20 years ago, we didn’t have social media, we didn’t have an iPhone, we didn’t have any of the technologies for the digital, social, and mobile revolution that we take for granted today. Today we’re at that same inflection point with AI. So how do you get your company ready to take advantage of this inflection point? And in the short term, that means how do you use these raw materials, we like to call them, of AI models? They’re trained to have the foundation of knowledge. That’s why they’re called foundational models. But they don’t have the domain expertise, the sector-specific knowledge. So how do you imbue these models with that for your business? And then how do you translate that into something that makes a difference for either you internally, for an internal tool for becoming more effective or efficient in your work processes, or a new service offering to clients, or even replicating some of your professional services work on your client’s desktop as a SaaS, software as a service, which is what we’ve done with some of our offerings through Delve Deep Learning. So that’s your short term, and that really starts with having like an AI wrangler that you can work with internally that sort of helps you figure this out and brings the context that you have. And then you get to that medium term, which is making that strategic decision of are you gonna be a builder or are you gonna be a buyer? If you look out into the medium and long term, AI, we’re not gonna spend the rest of our lives typing away with chat bots or even voice chatting with chat bots. You’re gonna see a lot of the AI recede into agentic workflows and AI assistance working in the background in concert with you, with your clients and other stakeholders, and with each other, with other AI agents and assistants, to make things happen. And are you going to build those? Are you going to try and become the technology provider and leader in your sector and your space, or are you going to take advantage and leverage the tools that others are building? There’s not a right answer or wrong answer, but you have to make that decision because in the long term, that’s where we’re going. That's where AI is going. And if you want to be ready for it, you've got to think through what your strategy is going to look like. Share on X

It sounds like no-code programming or something.

Yeah, and there’s different layers to being a builder, right? There’s a lot of no-code coded tools out there. AI is becoming an incredibly impressive coding assistant. With our dev team, they’re able to do way more than they could before. In fact, our technical lead, who has more of a front-end background, more of a data science background, he’s able to jump in and understand what’s happening on the back end, thanks to AI coding systems kind of helping explain what’s going on, the code base explain what’s going on. I think it’s going to make it easier for more people to be builders.

Yeah, that’s fascinating. So in practice, how does it change your business? Is this just saving labor and you’re training AI to be your interns and then you don’t need those interns and you can do the same amount of work with fewer people, or is it qualitatively changing the work that you do?

There’s definitely part of it that’s efficiency and effectiveness. You won’t need as many people, which I think is important. If you look at the long term demographic trends, we are having babies anymore. There’s not a workforce of the future coming to rescue us. We’re right now in the United States. there’s 0.75 workers for every open job, right? And that trend is not going in the interest of being able to fill jobs, it’s going in the opposite direction. So you’re going to need that piece of the puzzle, but I think it’s a lot more than that. I think AI provides the opportunity to really sort of re-imagine how you do your workflows and what can you accomplish, what can you offer. Service offerings that might’ve been cost-prohibitive before are now going to become more realistic to offer. Things that you might have done as a service can become more of a product offering, as we’ve done with some of our monitoring work going on to a SaaS platform. I think you’re going to see a future where there’s AI co-pilots for folks to do their jobs, the nature of work is going to change. And so thinking through, what are those different opportunities, both internally and externally. For me, it’s really exciting. I think for many business owners, we’re sort of ready to reach more of the market to great opportunity, right? When it was our team doing the monitoring analysis on an ongoing basis for a client, that made sense for the top 10 or 20% of the market, which are real high stakes debates where large budgets are being invested. But that leaves 80 plus percent of the market that we weren’t able to service because they were never gonna pay us what it was going to cost to service it. Now we can go all the way downstream and say, hey, if you’ve got a team and they need to get better at wrapping their arms around all the information out there, we’ve built an AI that’s got our expertise, the domain expertise and analytical approach infused into it that can help you do that better. And it brings our offering to way more public affairs teams around the world.

So you’re basically democratizing public affairs intelligence because now smaller players, nonprofits or small companies can also access it and they can punch above their weight perhaps. What I like about this AI idea is that it really empowers small business, because now you no longer need the big budget. You can just be innovative and leverage these tools, and you can create something really valuable from a very small budget, and that creates a chance for people. So, love it.

Yeah, that’s a great point, Steve. I think AI gives an unfair advantage to smaller, more nimble businesses that you see a lot of headlines about how at the enterprise level with the larger businesses, AI has been more hype than hope, and they aren’t seeing the return on the investment. And I scratch my head, I’m like, how are you not seeing a return on investment? And I tell you, when you talk to small business owners and entrepreneurs who have embraced AI, they’re all seeing massive ROI on their investments in these tool sets. And I think that’s really exciting, I think, because you’re going to have, look, if you look at the chart over time, how many employees you need to reach a billion dollar company has been going down, down, down. This is only going to drive it further down to, you could build a billion dollar company with 10, 20, 50, 100 people, thanks to the AI revolution. And that’s going to be critical for the future of economic opportunity, I think.

No, I agree. This is fantastic. Okay, so how do you make this available for other people? Because it’s not just for you, but for other people. Is it that smaller customers can then have access to the kind of service that you provide as a product, or is it that you white label this product for other agencies and they can serve their clients with it, or is it both?

Yeah, so that’s a great question. Our focus has really been putting the platform in the hands of public affairs teams, whether they’re in-house at companies and associations or advocacy groups or at consulting firms. And for the consulting firms, they can sort of take two different approaches. They can use it to serve their clients and become a customer, or we’re also partnering with them. Similar sort of the managed service provider model that you see in a lot of IT solutions where we’re providing the platform and they’re providing the service to make sure that the clients are able to use it successfully.

We're really trying to be supportive of the ecosystem that exists out there and make everybody better. Share on X

Because I’m sure your inbox is very similar, but the average public affairs professional is getting overwhelmed with information and data, right? They’re spending two plus hours of their day trying to figure out what in the world is going on so they can then go do something about it for their organizations. And I don’t know about your math skills, but spending 25% of your day and leaving you only 75% of your day to do 100% of your actual job doesn’t really work very well. And this tool can really help massively reduce that two and a half hours, two plus hours that they’re spending every day and give them a better product. Now some folks for these high stakes issues, they still want us, they still want our layer of expertise on top of that. And that’s great, but if they’ve got a team that just needs to bring more sanity to their work, we’re putting the platform in their hands. And also, above that, helping folks understand how do you use AI? Where else can it go into the workflow? How can you make it more effective for your use cases?

So what’s the mechanics of someone doing that? I mean, you basically took a tool which is ChatGPT and you customize it for your knowledge base, for your space, and you essentially turn it into a highly specialized version of it.

Yeah. We’ve gone a lot farther than ChatGPT. We’ve got about four or five different AI models that are all fine-tuned, working behind the scenes of the platform. So, we jumped into the advanced AI column there. And the reason why is when you start fine-tuning models, they’re going to be good at the particular thing that you fine-tuned them for. So, how you want a model to categorize or summarize information. It’s going to be different than building a recommendation and engine for what should be relevant and important to a client. And so there’s different pieces of the puzzle that we’ve worked through behind the scenes with a series of open source and enterprise models that the way we built the platform, we can swap in as models advance and get more developed, and we do more fine-tuning on them, we’re able to swap in new models for each piece of that puzzle so that we future-proof the platform from that perspective to make sure that when the next model comes out or the next model, we don’t just all of a sudden find ourselves behind the curve.

Let me understand this. What does it take to build a customized engine that incorporates all these different AI models? Is it about creating and developing prompts? Is it about integrating these with each other? Is it both? What does it look like?

Yeah, that’s a great question. So some of it is prompt engineering and sort of making sure that you’re giving the right information to the model within the prompt, within the context window. Some of it is is also making sure that, a lot of this involves data, having the right data. One of the advantages we have is we have 10 years of working with clients and building data sets of, this is important, this isn’t important. That sort of how pattern matching is how AI works. And you need large data sets to do that. You need like 250,000 tokens or more of data to really start fine-tuning. And you need more like 1.5 million tokens of data to really start getting in and training sort of pre-training models in the way that an open AI or Anthropic does. There’s the data layer of making sure it’s got the right pattern recognition to make the right decisions. There’s giving it the context it needs within the context window. So we’ll feed a news article and it’s got the context of, this is the client, this is what they care about, this is all the relevance. And then how do those things match up? And getting the models to mimic the way one of our analysts would make that assessment and judgment. And so part of that is prompt engineering. Part of that is feeding it the right data. In the context window, part of that is getting into the guts of the models and making sure that it’s got the right pattern of recognition.

Yeah, that’s fascinating. Well, I think we’ll see more of that coming down the pike as we move forward. So if someone is maybe in a public affairs space or maybe they are a small company, maybe a pharmaceutical company, and they want to position their product or they want to lobby for it, whoever can use this, how can they learn more and how can they reach out to you?

Yeah, so they go to delvedeeper.ai, insert your email and our team will be in touch. We’ve only just started coming out of self mode in the last month or two with the new company. So, the website is a little sparse. We’ve been busy building the platform, not the marketing website. But we love to chat with folks about what we’re building on the AI front and how it can help folks navigate the different advocacy and regulatory processes, understand what’s out there in the news media and what other stakeholders are saying and do so in a really efficient way.

That’s awesome. Well, if you’re into AI and if you’re into public affairs, then definitely reach out to Jeff Berkowitz, co-founder and CEO of Delve Deep Learning. And if you enjoyed this episode, make sure you follow and subscribe for us on YouTube and follow us on LinkedIn and give us a review on Apple podcast. Thanks for coming, Jeff. And thank you for listening.

 

Important Links:

Summit OS | Stevepreda.com

PO Box 400, Rockville, VA 23146
Phone: +1.804.332.1307