Tim Nguyen From BeSmartee–AI: One Bite at a Time

Tim Nguyen

Tim Nguyen is the CEO & Co-Founder at BeSmartee, a fintech powering the digital transformation of mortgage and commercial lenders. As CEO Tim sets the direction of the company, defines its product vision, defines its culture, leads mergers & acquisition initiatives and stays close to the company’s most strategic clients and partners.

Tim comes from an entrepreneurial family background, having taught himself to code and launch multiple online businesses in high school, advising local companies ranging from $5-$100M in annual revenues and as an active investor in numerous startups.

In his personal life Tim enjoys family, writing, music, farming, no limit poker, baseball, MMA and basketball.

While AI is all the hype right now, AI as a field dates back to 1956, marked by the Dartmouth Conference that brought together researchers interested in neural networks, the theory of computation and automating intelligence.

It was at this conference that the term “artificial intelligence” was first coined by John McCarthy, who, along with Marvin Minsky, Claude Shannon and others, proposed to explore ways to make machines use language, form abstractions, and concepts to solve problems usually reserved for humans.

The point is, AI has been around, and has been deeply integrated into our daily life, often in ways that might surprise us due to its subtlety or ubiquity. For example:

Social Media Feeds. AI not only helps to personalize your social media feed but also understands your behavior to tailor content specifically for you.

Smart Assistants. Devices like Amazon’s Alexa, Google Assistant and Apple’s Siri use AI to understand natural language, learn from the commands you give, and even anticipate your needs based on your habits.

Predictive Text and Autocorrect. When typing on a smartphone or computer, AI works behind the scenes to predict the next word you might type and correct spelling errors.

Personalized Recommendations. Online platforms like Netflix, Amazon and Spotify use AI to analyze your past behavior and make personalized product or media recommendations.

Traffic Prediction and Navigation. Beyond just finding the quickest route, AI-powered navigation apps like Google Maps and Waze can predict traffic conditions hours in advance.

For the record, I believe in the AI hype and believe the market opportunity is as big as the internet and its significance as much as the atomic bomb. To this end I have made a handful of personal investments into AI startups in and out of mortgage. From this experience I wanted to share with you the three layers that matter, and where you can find your competitive advantage.

Layer 1: The AI Brain

The AI Brain is essentially the foundational technology that powers AI systems, including algorithms and neural network architectures like those used in machine learning and deep learning. These core technologies enable machines to process and analyze vast amounts of data, learn from this data and make intelligent decisions.

Building an AI Brain requires immense resources, including advanced computational hardware, access to large and diverse datasets and a deep pool of expert knowledge in machine learning, software development and data science. Given the scale and scope of the investment needed, not just in terms of financial resources but also in terms of technical and human capital, it makes sense to leave the development of the AI Brain to the Microsofts, Googles and Apples of the world. These tech giants have the infrastructure and the R&D budgets to pioneer new AI developments and drive the innovation forward.

For the rest of us, accessing these AI technologies through APIs and cloud-based services offers a more viable path forward, enabling us to leverage the power of these advanced AI systems without the prohibitive cost of building them from scratch. This approach democratizes access to powerful AI capabilities, allowing a wider range of users to implement AI solutions tailored to specific needs and challenges.

Clustering

ChatGPT, launched in November 2022, is arguably the platform that sparked the current AI revolution and as such, has given birth to too many “GPT wrappers,” essentially software wrapped around a singular Large Language Model (LLM) called GPT. But there are many other AI Brains out there, such as LLaMA, FALCON and Mistral. Contrary to popular belief there is not one AI Brain that is better than any other other. It really all depends on the task at hand.

This is where “clustering” comes into play.

Think of GPT and all the other AI Brains as a historical genius. Would you rather get advice from Einstein, Newton, Tesla, Mozart and Da Vinci, or just one of them? Of course more brains are better than one! Furthermore, each genius has a different perspective due to their background, education and experience.

If you are looking at buying an AI solution, ask the provider if they use a singular LLM, or if they use a cluster of LLMs. Also ask the provider how much specific mortgage industry knowledge it was trained on. This is what they are doing at mortgage tech vendor AskBobAI for example; using a cluster of AI Brains instead of just one, all trained on mortgage industry knowledge.

Layer 2: Training With Proprietary Data

Once you have a cluster of AI Brains, the next step involves training your solution with your proprietary data, which could come from sources like your helpdesk software, emails, internal documentation, databases and governing entities. In this layer you customize the AI Brain to deeply understand your business specifically. In mortgage speak, this would be your guidelines, your products, your historical underwriting data, your conversations and your everything else.

Initial training equips the AI with the necessary foundational knowledge, but ongoing training is crucial to keep it updated and relevant. This continuous learning process allows the AI to adapt to changes in your business environment, such as new mortgage products, processing policies or customer outcome objectives. By regularly updating the training data at least weekly, the AI can refine its predictions, improve its accuracy, and offer more precise recommendations.

This ongoing training turns the AI into a dynamic tool that not only encapsulates the current state of your business operations but also evolves alongside them. Mortgages move fast after all; therefore this iterative process is key to maintaining a competitive edge, as the AI becomes an ever-improving repository of institutional knowledge that helps streamline operations and enhances decision-making.

One very important warning: In all cases above, if you are looking into buying an AI solution, ask the provider how they protect your proprietary data. Do not let your proprietary data leak into the AI Brain. In our heavily regulated industry you will get into heaps of trouble.

“I Know Kung Fu”

Remember that scene from The Matrix where Neo gets trained with every martial art known to man? That’s what it means to train the AI with your proprietary mortgage data. Neo before being “plugged in” is synonymous to an AI Brain. He knew what he knew about martial arts, he knew how to move his limbs, he knew what hurts, but it wasn’t until Neo got trained that he really knew kung fu.

Put another way, the AI Brain is comparable to an MBA college graduate. Smart, but their skills are only foundational for the real world. By training it with your company’s data and know-how, you’ve now created a MBA with 10 years of experience at your company.

Layer 3: Delivery

Now that you have a cluster of AI Brains trained on your company’s data, the last critical step is delivery, i.e. integrating and deploying this trained AI into the day-to-day operations of your mortgage company.

In this layer, real people begin to experience the benefits of AI firsthand. Here comes the critical phase of API integrations into industry and company-specific legacy technologies such as your LOS, your POS, your servicing database and various other data management platforms. This step ensures that the AI’s insights and capabilities are readily accessible and seamlessly integrated into existing workflows that humans ultimately interact with. It’s in this layer where you truly find value in AI.

By embedding AI into operational processes, mortgage companies can measure tangible ROI through metrics like increased efficiency, reduced operational costs, and enhanced customer satisfaction. For instance, AI can automate your help desk giving answers to originators faster, or analyze advanced analytics in your point-of-sale to enhance engagement rates with borrowers.

If you are looking into buying an AI solution, ask the provider how your stakeholders ultimately interact with the AI. Is it a threat to their jobs? Or is it a tool that makes them uber-efficient? The method of delivery may be the difference between successful adoption or lost years.

Eat an Elephant

Desmond Tutu once said, “there is only one way to eat an elephant: one bite at a time.” What he meant by this is that everything in life that seems overwhelming, or even impossible, can be accomplished simply by taking it on a little at a time.

In your own mortgage business take small steps to find use cases for AI. Do not try to boil the ocean and be very weary of promises from vendors that AI is going to solve everything today. AI is still very early, many mistakes have been made and many more will be made until proven practical for core mortgage use cases. Start with existing AI platforms in more generic areas of your business such as marketing, HR and legal. Get the measurable value today, and start building a culture of embracing AI as an ally, versus viewing it as a threat. The mortgage vendors will start to emerge and you’ll eventually eat the entire elephant.

(Views expressed in this article do not necessarily reflect policies of the Mortgage Bankers Association, nor do they connote an MBA endorsement of a specific company, product or service. MBA NewsLink welcomes your submissions. Inquiries can be sent to Editor Michael Tucker or Editorial Manager Anneliese Mahoney.)