There are truly incredible things happening right now in the software industry. I’ve been building applications since the dot com era and I’m convinced we are in a time of step-function advancement. Yes, on the level of the Jet Age and the launch of the Internet itself. Perhaps even bigger than this for reasons we cannot yet see.  We will look back at the ‘20s’ as a pivotal time in history and either the peak of the 4th Industrial Revolution or just the beginning of it. While I strongly believe that this is a positive and exciting time, I also grew up with the Terminator and remember clearly that things got very, very bad when Skynet became self-aware. That is a topic for another post as I’m here to share some thoughts about AI, how we evaluate it at Massive, and provide a case study from one of our investments that is all in on AI. 

‍

Massive invests in B2B software companies as part of our strategy. These companies include enterprise applications, cybersecurity, fintech, and more. We also invest in deep tech companies and we are now seeing crossover where AI plays a big role in both of these strategies.  It became clear to us that in order to evaluate software companies and how they apply AI, we needed a framework and structured thinking so that we could answer some basic questions about the company and its products. Our framework includes the following: 

‍

AI Native - The simple and fundamental test to evaluate if a product is truly AI Native is to understand what happens to the application if the AI engine is turned off. Does the application cease to provide its core function? Would the customer reject the software or would major problems cascade so fast that it became inoperable? If yes - then this is what we’d consider an AI Native product. We are active investors in “Vertical AI” at Massive and the first criterion is passing the AI Native test. If a company and product are using AI at the core of the application in an industry or market vertical in a novel and powerful way we tend to take notice. Some examples from our portfolio include Knit, Hosta, and Operant. Foundational models (LLMs) and GenAI tools, AI APIs, and AI dev tools are all variations of AI native and are true “AI companies” in my mind. 

‍

AI First - Is the Company (note that I said Company not product or application - this will be important later) using AI within both the organization and the application they are building? Have they embraced the efficiency, margin, and technological gains that AI tools can provide? If the product is using AI but the people building, selling, and supporting it are not - this would not be considered an AI First company. However if AI is being used in customer success, content, development, QA or internally for work and collaboration - these all can make a business AI First in our minds. Beem, a consumer fintech company in our portfolio, is an example of a company that was an AI Native company before it was even a buzzy thing and has now become a powerful AI First company as well. Akshay Krishnaiah, CEO and Founder of Beem has done an impressive job leading this evolution and is a great case study. I will go deeper into this story below because we see very few AI First companies in practice although this is definitely taking off now. 

‍

AI Enhanced - Any application that uses AI to make some feature or aspect of the product better and to improve the power or usability of that feature, but clearly the application will provide its core functionality without it, or it already was in the market and now has AI features. My favorite current example of this is Superhuman - an email client that recently added AI search. I use the Superhuman AI search every day and it is honestly pretty awesome. The way that I think about search is changing and the utility and value of having contextual and intelligent filtering is a step-function change from keyword search. However, if they had to turn this off or the API to the LLM turned off, the app would still function as it did 6 months ago. Thus Superhuman is not an AI Native company. I have no idea if they are an AI First company because that would require knowledge of the tools, workflow, and culture inside the company.

‍

We use this framework to help us understand where in the company or product AI is being applied. We then dive deeper into researching and understanding if the AI leverage is unique to the company or novel in a vertical and if leveraging LLMs, ML or CV (computer vision) create some step-function improvement in the customer experience as well as the company’s prospects in a market. 

‍

How does Beem use AI for good? Lets dive into how Beem has been using AI as its an excellent example how customers reap the benefits in addition to the company itself when used for maximum impact. It is practically impossible for humans to process or efficiently analyze billions of transactions, uncorrelated data and human behaviors at consumer transaction volume. Reports or metrics only tell a small portion of the story, and in many cases (like credit scores) static algorithms that assess a person’s risk fall short. Credit scores in particular are a terrible and clunky mechanism to assess a person’s real-world underwriting profile. This is a static algorithm that factors in basic information and is not capable of looking at deeper patterns or non-obvious correlations.  

‍

Beem is an AI First and AI Native company. Their mission is to “Unlock Financial Freedom for All”. From the outset, Beem built a risk-scoring engine using machine learning (ML), which is a core technology that underpins what we all call AI today. It is foundational to LLMs and GenAI tools. So it underpins many of the things we are starting to experience and hear about today. ML enables software to learn and improve by itself rather than needing to be explicitly programmed - this is the “learning” in Large Learning Models (LLM). Rather than get lost in the technical weeds,  here are two customer stories from Beem that demonstrate the power of using AI in financial products for consumers. 

‍

• The Widow’s Story: Beem helped a widow of a war veteran who relied on VA benefits like Dependency and Indemnity Compensation (DIC) and income from Etsy sales to support her three daughters. She was rejected by every provider in the market due to her non-traditional income sources, but Beem's AI was able to recognize her unique financial situation and offer her the support she needed.
  
  
• Jake’s Journey: Jake is an HVAC technician in Northern California, who saw his credit score drop to 500 after using more than 30% of his $500 credit limit on a family event and missing a payment during a stressful move. This caused him to be denied the car lease he needed for work and even rental housing. While other services rejected him due to his low score and fluctuating income, Beem’s AI detected positive trends in his financial behavior and earnings, which were seasonally affected and contradictory to updates on his Credit Monitor. Beem’s AI was able to intervene at the right inflection point, identifying his credit problem, guiding him on how to improve it,  while also delivering early access to deposits, budgeting assistance, deals on everyday expenses, debt restructuring and ultimately helping him regain financial stability. 

Beem started building with this technology 5 years ago, and part of the company’s competitive advantage in the market is based on how far along they are today with this. They would be characterized as an AI Native company in our framework above. Beem can now instantly build AI-based composite indexes for each user that are proprietary to them. The technology looks at billions of transactions and understands broad patterns, narrow patterns and the uniqueness of each person’s financial life to better support them and their needs. The AI inside of Beem’s platform can 1) learn as the dataset grows 2) improve its ability to predict outcomes and 3) identify ways to help users and automatically make these recommendations dynamically. A platform that was not AI native at this level would have static code that followed closer to traditional “if-then” logic. This is a step-function ahead of using credit scores or bank account balances to accept or reject a customer for a financial, insurance, or health product. Customers benefit directly from gaining access to the great product set Beem has built, which helps them improve their financial health and the company benefits from smarter underwriting so that they can serve more customers with lower fraud, charge offs and other product risks. 

‍

This work started well ahead of the AI hype cycle. The in-house effort to build a risk and scoring engine has paid dividends and enabled Beem to see some of the highest usage and retention rates in the consumer fintech category while also enjoying 2x lower nonpayment rates than everyone else. If the point of all of this “intelligence” is to make applications smarter, help humans with repetitive tasks, and accelerate what we can do - Beem has proven the value and the potential. This is not theoretical -  the benefits are evident in the company’s metrics. Beem is a profitable and private consumer fintech that has not raised a ridiculous amount of capital. This would not have been possible without their leverage of AI and development of in-house tooling that was ahead of the market. 

‍

Across the rest of the fintech landscape we are seeing accelerating adoption of AI as well. It's logical - AI is very well suited for this industry that is full of long customer histories, huge datasets, non-standard transaction formatting, tons of complexity and high demand for security and privacy. Some of the obvious use cases include fraud detection & security, credit scoring, customization (of solutions/products/advice), forecasting and predictive analytics and customer engagement. Investment interest in fintech has dramatically cooled off in the last few years and I would expect that both existing companies and new ideas will deeply leverage AI in their products. In fact, if that is not already baked in the company is way behind the curve. Based on what we have seen at Beem, leveraging AI across the organization is going to be necessary to compete at any level.

Here is a quick public market data point. We are not investors and have no inside info on Dave - but market data indicates they did not become profitable until Q4 of 2023 which is a year after they went public and had burned through $600M+ of venture capital. Today they indicate they use AI-based underwriting - so perhaps this is another positive data point that LLMs applied to consumer finance is moving the needle on risk and underwriting performance. Consumers / customers win here in addition to shareholders and investors.

Beem is also an AI First company. They leverage the technology across functions - development, support, product, internal comms, research, etc. Every employee is learning and using AI to do their jobs every day to drive speed, efficiency and reduce overhead. This is both internal and external. Externally Beem uses AI for its marketing and customer acquisition engines - which has enabled them to grow organically to 3M users. The old playbook for venture-backed companies was to spend lots of money on paid acquisition early, find the leaks in the bucket over time and then build enough brand to drive organic growth later on. Beem has flipped this on its head and is leading with organic growth out of the gate.  This has been impressive to watch and we know the business now has a strong and proprietary foundation to scale up from here. 

‍

As an investor, seeing the economic benefit of AI in both the application core (AI Native) and the operational side (AI First) is truly encouraging. Greater efficiency drives higher margins and better metrics which translate directly to higher valuation multiples for the business. I would submit that the software world is now at a huge inflection point where the cost to build and operate at-scale organizations is going to drop substantially. On the capital formation side, what would have taken $30-50M to get to the “ongoing 

concern” scale could now take $15-20 or less.  On the operating side - when you add up the people, the software they need, etc companies will be able to save millions a year in costs by leveraging AI smartly.  Tomasz Tunguz just wrote about this “A Challenge to SaaS Orthodoxy” and looks at how Klara is cutting costs dramatically via AI. There are going to be some huge ripple effects that change the dynamics in both tech and venture as a result…

‍

Thanks for reading!

‍