The big picture and the focus

• This wave of AI is specifically about building on top of Generative AI - it got started after GPT-3, and the explosive growth really started after ChatGPT (Nov 2022).

• We are only looking into building on top of LLMs. With that, at this moment we will ignore fields like image generation, speech processing, robotics, etc.

• If we look at the current Gen AI stack, there are 4 layers: infrastructure, (foundation) model development, application development tooling, and finally application. We will be only looking into applications.

• Within AI applications, there are horizontal apps (a layer on top LLMs), enterprise apps (marketing, sales, customer support, etc), and vertical apps (legal, insurance, coding, etc). We will be only looking into vertical ones.

Types of problems AI can help

• The task medium is natural language
• Many examples of the task were featured in the training data
• You don't need more than 90% accuracy

The value proposition of building on top of AI

• All the enterprise AI startups going after verticals have the right idea. Pick a market, deeply understand the workflows, build simple software to model the workflows, and use AI to augment the human judgment involved. Huge opportunity in 1,000’s of categories.

• Focusing on a specific problem. Building workflow, collaboration and deep integration's, that are specific to that problem.

• Domain specific products with deep industry context and polish.

Three Verticals Suitable for AI

Professional Services

AI is well-suited for legal services, as the core product of law is language, and large language models (LLMs) are transforming this field.

Consulting and accounting are also ready to adopt AI. The major consulting firms employ tens of thousands of consultants and accountants, a workforce that can be significantly enhanced with AI.

Financial Services

Investment professionals and wealth advisers deal with complex information, involving both proprietary data and real-time market data. Financial modeling requires both mathematical skills and human judgment, where errors can have significant financial impacts. Additionally, professionals spend a substantial amount of time on routine tasks. AI can automate these mundane workflows, saving time and reducing errors.

Healthcare

AI’s potential in healthcare is becoming increasingly tangible. Previous efforts to standardize unstructured data were limited by diverse legacy IT systems. LLMs can handle unstructured data as it is, benefiting both clinical and administrative tasks. AI can enhance diagnostics, decision-making models, and automate insurance claims, improving overall healthcare data management.

Three potential AI use cases in healthcare include:

• Transcription and documentation
• Prior authorization
• Clinical reasoning models

Some startups

Appendix A: The 2024 MAD (Machine Learning, AI & Data) Landscape

The landscape and card

Appendix B: Beyond vertical applications: Big AI deals so far in 2024