Investing in AI Startups
Part 3 of 3
This blog post was written by DougL. Edits provided by friends. AI was not used to write or edit it. Browser-based technology with generative AI and ChatGPT was used to fact check.
Angels, VC, and PE with a track record full of successes typically analyze funding prospects relative to the following considerations in this order:
People
Market and business problem, and
Technology and product solution
While this framework applies to the latest wave of artificial intelligence (AI) startups, there’s still a lot to learn about their operations, business models, funding and exits. This is especially true for AI startups doing large language models (LLMs) and ChatGPT applications.
Overall, investing in startups that are part of this new wave requires revised thinking and deeper due diligence.
Here are some of the most important considerations to keep in mind when considering an investment in an AI startup.
Team: Strong, experienced, and talented people are crucial to the success of AI startups. Make sure there are many “decathletes” among the people on the team. Look for a team whose members have diverse skillsets, proven track records, and who are vision-driven.
Note: There is a divergence of opinion on the optimal size of initial AI software development engineering teams. Some say they can commence as smaller-than-SaaS teams. That’s because they use AI automation tools, and coding is easier today with ChatGPT in combination with the latest software development lifecycle (SDLC) resources and platforms. Others are of the opinion that these dev teams need to be “beefier” (hat tip to Bob Mason) because data scientists and more ML experts are needed in addition to standard, full-stack software engineers to build and deliver the product. The jury is out on this question for the time being, and the answer will undoubtedly vary based on the technology and product.
Market Demand: Make sure that there is a real market demand for the product or service being developed (i.e., product market fit or PMF) and the business problem they aim to solve. Steps to take to build this certainty include:
Identifying a clear business problem that needs to be solved,
Pinpointing the defensible solution that fully exploits AI,
Delineating a range of innovative options,
Speaking with lots of potential customers to gain a clear understanding of the needs of the target market,
Conducting market tests to validate assumptions by doing multiple product -market fit experiments and product releases,
Build a compelling, differentiated solution, and
Establish a product-led growth (“PLG”) feedback mechanism.
While these steps are applicable to traditional software, they are different and more demanding in AI product development. This is partly due to the newness of technology and the need to neutralize the ‘hype cyclone’. But it also stems from the requirement to factor in data labeling, amassing training data, and other factors.
Business model: Consider how the company will generate revenue and attain profitability. For Series A, apply the Rule of 40. (Read my blog post on it here.) In other words, look for companies with scalable business models that can generate sustainable revenue growth with attractive margins.
Technology:
Technology evaluations have returned to startup venture due diligence and investment assessments after years of being on the back burner during the SaaS and zero-interest years. Architecture, code quality and AI technology roadmap are all critical in your assessment. Seek out or invest in companies with cutting-edge technologies.
AI startups seeking funding require a retinue of specialized expertise in machine learning and data science.
Be aware that AI startups have higher research and product development costs – especially in those with LLMs. These costs are more significant than costs associated with the typical SaaS startup.
Developers have to be more involved upfront and throughout the model’s life. Expanding and evolving models is important, and safeguarding against model drift is key.
Many of the same SDLC frameworks apply but data science plays a very important role in AI startups that adds new demands to product development.
AI technology and businesses centered around ChatGPT requires extensive UI resources because ChatGPT does not have a fully formed interface. UX is important but secondary as ChatGPT provides the front-end to the text and data that is displayed.
The question that must be asked with each prospective AI startup investment is simple: Is it defensible? Defensibility refers to a company's ability to stake its claim, and then protect its technology, data, and other assets from competitors. It needs to maintain a competitive advantage – not just a competitive edge – over the long term. In other words, does it have sustainable competitive advantage?
Be sure to avoid point solutions. Those are software applications or tools that address only a specific, targeted problem in the general market or in verticals. These solutions are often focused on a single task or function, and are designed to integrate with other systems or applications within an organization's IT infrastructure. Platforms are the preferred investments among AI startups.
Intellectual property (IP): Make sure that the company has secured the necessary patents and intellectual property rights to protect its technology and innovations. It's important for AI startups to work with lawyers to determine the best IP protection strategies for their businesses. The foundation for any such strategy is for the company to secure the necessary patents and IP rights to protect its technology and innovations.
OpEx: AI applications and models use cloud providers as their primary distribution model. However, many of the larger ones (e.g., OpenAI) host their models themselves. LLMs, especially, consume huge amounts of cloud computing resources which, of course, come with costs.
Ethical considerations: As AI continues to evolve, it's essential to consider the ethical implications of the technology. Look for startups that prioritize ethical considerations and are transparent about their approach to unbiased AI, data privacy and zero-trust security.
Funding & partnerships: Look for companies with a strong network of investors, partners, and technical and business advisors. An AI startup must plan to scale, from OpEx, GTM and human capital angles, from the beginning. There is no playbook for how to build and scale AI startups. (OpenAI is exceptional because of its partnership with Microsoft.) This necessitates deeper due diligence prior to an investment and having experienced AI investors in the syndicate.
Valuations: Arriving at an AI startup’s valuation is usually quite challenging. Potential ‘curve balls’ include the unique nature and timing of the technology, the lack of established metrics for measuring success, and the inclination of founders to overvalue their companies while investors automatically half the number. Conventionally, AI investors value AI startups using a combination of market, income, cost and technology-based approaches, but in the end valuations vary widely.
Another way to invest in AI and reduce the guesswork, or analyze companies that can serve as comp’s and decrease the risk in investments in AI startups, is to take a look at leading public companies that utilize this new wave of technology. Public companies such as Alphabet, Amazon, C3.ai, Lemonade, Meta Platforms, Nvidia, Tesla, and others.
By looking closely at these and other companies heavily involved in AI, investors will gain a leg up on developing more and better metrics upon which to base their investment evaluations.