Back
Strategy

An angel's guide to investing in AI

As AI companies continue to raise capital at unprecedented levels, some investors are left wondering how they can get started. This guide will show you how to think about investing in AI, where to invest, and what you should do before writing your first check.
Published Jun 11, 2025
7 min read

There’s a lot of promise around the impact that AI will have on the world. By changing how society works, plays, and engages with technology, investors are more sure than ever that there will also be massive economic reward by investing in this meaningful technology. 

Data shows that 58% of global venture capital dollars went to AI and machine learning startups in Q1. In the MENA region, AI is the fastest-growing tech vertical for venture capital funding. The investor demand is real, but navigating the landscape of AI deals can be confusing to angels who want a seat at the table. 

In this guide, we’ll walk through the key areas of AI investing and show angel investors how they can be involved. 

Key areas for early-stage AI investing

Artificial intelligence touches a wide range of software, hardware, and data. It’s an ecosystem of interconnected layers, each one being critical to the social and economic success of the industry. As an investor, you can think of AI companies as categorized into three groups: foundation models, infrastructure, and applications

Each layer presents different types of investment opportunities, risk profiles, and business models. Below, we’ll walk through each area of AI investing, explain what it means, and offer key examples of each. 

Foundational models

Foundation models are large, versatile AI systems trained on massive datasets to perform a broad range of tasks, like understanding language, generating images, or recognizing patterns across multiple inputs (text, audio, video). They’re called “foundation” models because they serve as the base layer for a wide range of downstream applications.

Rather than being built for a single purpose, these models are designed to be general-purpose and adaptable. Developers can fine-tune them for specific use cases, like legal document analysis, medical imaging, or customer service chatbots, without needing to train an entirely new model from scratch.

Here’s how they break down:

  • Large language models (LLMs): General-purpose AI systems trained on a vast amount of text to understand, generate, and interact using human language. These are capital-intensive to build but serve as the foundation for many downstream applications. 
    • Example: Pronoia, an Arabic-first LLM designed to power regional consumer and enterprise tools.
  • Multimodal models: AI models that process and generate across multiple data types, like text, images, and audio, offering broader capabilities than text-only LLMs. They require even more computation and data, and are still early-stage outside of English-language ecosystems. 
  • Verticalized foundation models: Specialized versions of LLMs tailored for specific industries such as healthcare, legal, or finance. These models often have clearer go-to-market paths and stronger defensibility due to domain expertise and data advantages. Most are built on top of general-purpose models like GPT-4 or Claude, with added fine-tuning for industry relevance.

AI infrastructure

AI infrastructure refers to the tools, systems, and backend technologies (both hardware and software) that enable the development, deployment, and scaling of the foundational models above. You can think of AI infrastructure as the support system that enables models to function effectively.

Investors can think of AI infrastructure in the following buckets:

  • AI chips and hardware: Specialized processors, like GPUs, TPUs, and custom accelerators, designed to handle the intensive computation required to train and run large models such as LLMs and multimodal systems. These businesses are capital-heavy and often raise from institutional investors, but angels can participate indirectly through pooled vehicles like SPVs.
  • Data infrastructure and labeling: Tools and services that prepare, structure, and annotate the massive datasets used to train AI models. High-quality data is essential for model accuracy, yet this part of the stack is often underappreciated.
  • Model hosting and inference APIs: Platforms that simplify the process of deploying and scaling AI models, allowing developers to plug into advanced models without managing their own infrastructure.
  • Synthetic data generation: Tools that create artificial datasets to supplement or replace real-world data, particularly useful in regulated industries or where data is scarce. These solutions are gaining traction in sectors like healthcare, finance, and defense, where privacy concerns limit access to real data but model performance still matters.

Application layer

The application layer is where AI becomes visible and useful to end users. It includes the products, tools, and software experiences built on top of foundation models and infrastructure, turning raw AI capabilities into real-world solutions for businesses and consumers.

Whereas foundation models provide general intelligence, and infrastructure enables access and scale, the application layer delivers specialized functionality through targeted use cases—everything from AI-powered legal assistants to personalized fitness apps or autonomous customer support agents.

There are various ways for investors to get involved in investing across AI applications, including:

  • Agentic workflows: AI systems designed to autonomously perform multi-step tasks across complex workflows, acting more like digital employees than simple tools. While still early in adoption, these agents show strong potential when embedded in high-value domains like compliance, operations, or financial analysis. The winners will likely be those with deep integrations and clear use cases. 
    • Example: Beam AI, a leader in generative AI agents to help organizations automate repetitive manual tasks, boost productivity, and enable teams to focus on the work that really matters. 
  • Enterprise SaaS: Cloud-based software platforms built for businesses, where AI capabilities are embedded directly into the core functionality (automating workflows, enhancing decision-making, or personalizing user experiences). These products often benefit from recurring revenue and sticky customer relationships, especially when tied to operationally critical tasks.
    • Example: Lucidya, a customer experience management (CXM) platform that utilizes AI to provide real-time insights and interactions for organizations.
  • Vertical software: Industry-specific software solutions tailored for sectors like legal, real estate, or logistics. These tools gain defensibility through proprietary data, compliance alignment, and tight integration into daily workflows, offering more predictable monetization paths than broad horizontal tools.
    • Example: Razi, an AI-powered assistant developed by LocAI that streamlines clinical workflows and supports decision-making in the healthcare sector.
  • Developers: AI-powered products that help software engineers write, debug, or deploy code faster and with fewer errors. Investors should look for strong usage metrics, community traction, and embeddedness in developer workflows.
    • Example: Tabnine, an AI-powered coding assistant that integrates with popular Integrated Development Environments (IDEs) to provide real-time code completions, suggestions, and a chat interface to assist developers throughout the software development lifecycle
  • Consumer AI tools: Products built for end users, often focused on creativity, productivity, or entertainment. While user growth can be rapid, monetization is more challenging and often hinges on retention, network effects, or niche virality. Many consumer AI startups burn quickly without a clear path to recurring revenue.
    • Example: Canva, an AI-powered design assistant that helps users quickly create, edit, and enhance visual content.

Building a thesis for AI investing

Why having an investment thesis matters

It's tempting to jump at every exciting opportunity you encounter, especially with a rapidly evolving industry like AI, but seasoned investors know the importance of having a clear investment thesis. A thesis is essentially your strategy; it clarifies what deals you're looking for, streamlines your decision-making when vetting new opportunities, and helps you avoid distractions that might be more hype than substance.

Your investment thesis also helps you build expertise, strengthen your network, and communicate clearly about your interests. This can lead to better targeted deal flow, increasing your chances of finding the right companies more quickly.

Defining your AI investment focus

The first step to building your thesis is defining where you want to play within the AI landscape. Think of this as your sweet spot. It could be industry-specific, like healthcare, finance, or logistics, or technology-specific, such as focusing exclusively on infrastructure like data labeling tools, AI chips, or foundational language models.

For instance, you might decide, "I'm going to invest primarily in AI startups in healthcare because I understand the challenges hospitals face and see huge potential in AI-driven diagnostics and treatment." Alternatively, your thesis might focus on geography: "I specialize in supporting AI startups serving Arabic-speaking consumers because that's a rapidly growing market with little competition."

By clearly defining your area of focus, you make it easier for yourself to spot promising opportunities and quickly pass on deals that don't align with your expertise or interest.

Setting clear evaluation criteria

Once you’ve defined your area of interest, it's essential to outline your evaluation criteria clearly. What does an ideal startup look like for you? Every investor will have slightly different priorities, but here are some common factors to consider:

  • Team expertise: You're investing in people as much as technology. Look for teams that deeply understand their domain, have highly relevant and technical backgrounds, and a proven ability to execute.
  • Data strategy: Particularly in AI, data is critical; look for startups with unique access to high-quality or proprietary data.
  • Market opportunity: Is there a genuine need or demand? Assess how clearly the startup defines its target customer and how effectively it's solving that customer’s problems. AI is likely to expand the addressable market of any target audience, so remember to consider how the market could grow in the future.
  • Scalability: Can the company scale efficiently? Pay attention to their business model, unit economics, and how they plan to expand.
  • Competitive landscape: With all of the new AI tools launching right now, it may seem like a crowded market. Competition is okay, but you’ll want to make sure the company you’re conducting due diligence on has a defensible moat that will remain even as market conditions change in the future. 

Creating a checklist or scoring rubric based on these criteria helps streamline your evaluation process and ensures consistency.

The ultimate goal of your due diligence on a potential investment is to understand the specific application of AI in a company. The use cases of AI are rapidly changing and evolving, making it challenging to keep up. If you still have questions remaining or don’t fully understand how AI is integrated, it’s important to continue asking questions until you fully understand the role that AI is playing in the company you’re evaluating.

How to refine and evolve your thesis

As you review more AI companies and learn from your investments (all while the AI landscape evolves), you'll likely refine and evolve your thesis. This flexibility allows you to respond effectively to new learnings or shifting market trends.

Continuously adjusting your investment thesis as the market and your experience evolve is what makes a good investor. Your original assumptions might not hold true anymore if the market has changed or you’ve learned new information about the landscape of deals. Staying adaptable ensures that your investing stays aligned with both your interests and market realities.

Communicating your thesis clearly

Finally, having a clear thesis is most beneficial when you communicate it effectively. Let other investors and founders know precisely what you're looking for. 

Being able to articulate your thesis succinctly, such as "I invest in early-stage consumer application AI companies that solve payment challenges for SMBs in MENA," will help attract targeted deal flow. It also helps other investors and startup founders easily remember your focus, increasing your visibility within your chosen niche.

The clearer your thesis, the stronger your brand as an investor. Over time, you'll build a valuable reputation as a go-to expert within your area of focus, opening doors to the most promising startups in the AI ecosystem.

Leverage your networks to enhance deal flow and diligence 

When investing in emerging and technically complex sectors, relationships are a powerful edge. Whether you can collaborate with VCs, angel networks, or syndicates with an AI focus, being part of the right community gives you access to better deal flow and helpful perspectives. Instead of evaluating startups alone, you can tap into the experience of operators, engineers, and AI specialists who can help evaluate a company’s technology and business model.

How Zest can help

Zest is digitizing private market transactions, building tools to streamline how entrepreneurs, funds, and investors transact. Our platform is designed to save you time and reduce administrative costs, simplifying the end-to-end transaction process.

Get started with Zest today.

home-lightc-1home-lightc-1home-lightc-1

Launch a Zest vehicle to suit your needs