What were the main challenges Zeus faced with large enterprise AI transformations?

Challenges included the difficulty of understanding diverse industry-specific use cases, dealing with messy and inconsistent data, integrating with legacy systems, and underestimating the ongoing maintenance costs. There was also a risk of specialized companies like Decagon or Sierra eventually outperforming their custom solutions.

What is the target market for Brendan Faulk's new company, Hercules?

Hercules is focused on AI-powered app generation, similar to what they built at Fig. They are targeting users who need to build production-ready applications easily, moving away from complex enterprise solutions towards a more accessible model.

How does Hercules differentiate itself from other AI app builders?

Hercules aims to provide a fully integrated, white-labeled experience, handling all backend, frontend, and database needs. They are focusing on empowering users to build production apps without needing to manage complex integrations or specific AI models directly.

What are the key differences between CLI-based AI tools and web-based app builders?

CLI-based tools often target engineers and can face gross margin challenges due to the demand for the best models. Web-based app builders, like Hercules, target a broader, less technical audience, focusing on ease of use and integrated services, which can lead to higher gross margins.

What is Brendan Faulk's perspective on brands in the AI tool market?

Brendan believes that brand will matter significantly in the future for AI tools, especially for those with higher price points. A strong brand can justify costs even when cheaper alternatives exist, potentially enabling higher gross margins.

What lessons did Brendan Faulk learn from his startup failures and pivots?

He emphasizes the importance of transparency and honesty in the startup world, learning from failures, and sharing these learnings. This approach can provide an information advantage and ultimately lead to better long-term strategies.

⚡️No, Don't Do Palantir for AI - Brendan Falk, Hercules (AUDIO FIXED)

Latent Space Podcast

Science & Technology3 min read36 min video

Sep 18, 2025|2,121 views|37|7

Save to Pod

Key Moments

TL;DR

Brendan Falk discusses pivoting from enterprise AI transformations to developer tools with Hercules.

Key Insights

Building enterprise AI solutions for Global 2000 companies is complex due to variable technical maturity and unique edge cases.

Traditional system integrators often fail to deliver production-ready AI solutions, leading to wasted investment for large enterprises.

The 'Palantir for AI' model requires substantial revenue per contract ($5-10M) to be economically viable, due to custom development needs.

The shift towards developer tools like Hercules is driven by a need for scalable solutions and a pivot away from complex enterprise services.

The AI developer tools market is vast, with opportunities in areas like website builders, SaaS applications, and specialized developer tools, rather than just 'AI app builders'.

Transparency and sharing learnings from pivots are valuable for the startup community, despite the prevalence of hype.

THE CHALLENGE OF ENTERPRISE AI TRANSFORMATION

Brendan Falk, founder of Zeus (now Hercules), shares his experience attempting to build an 'AI-native Palantir' for Global 2000 companies. He observed a wide gap in AI maturity among these large, often non-tech-centric organizations. Many Fortune 500 companies, despite significant revenue, lacked basic AI understanding and technical depth, relying on expensive system integrators who delivered more 'science experiments' than production-ready solutions. This created a significant market opportunity for a company that could bridge this gap.

FLAWS IN THE 'PALANTIR FOR AI' MODEL

The initial strategy for Zeus was to tackle custom AI solutions for large enterprises, aiming for $5-10 million contracts. However, Falk realized the economics were challenging. The amount of work required for custom AI agent development, including strategic advisory, build, and change management, was not proportional to contract size. Even smaller contracts demanded significant effort, making the pursuit of very large, custom projects the only viable path, which often meant targeting companies with $5 billion in revenue or more.

THE COMPLEXITY OF CUSTOM AI SOLUTIONS

Falk detailed the operational difficulties encountered when building custom AI solutions. Identifying use cases through strategic advisory was straightforward, but the build phase was plagued by unforeseen edge cases, messy data, and integration hurdles. Even seemingly simple tasks like processing RFPs or handling first notice of loss involved thousands of company-specific quirks. This manual effort, including data cleaning and integration wrestling, consumed substantial time, often replicating the challenges humans faced, thus negating some of AI's expected efficiency gains.

MANTAINENCE AND EXISTENTIAL THREATS

Beyond initial deployment, Falk underestimated the ongoing maintenance costs associated with custom AI solutions. Unlike traditional SaaS products with amortized support costs, each custom engagement required dedicated, ongoing support. Furthermore, Falk identified an 'existential threat': specialized companies like Decagon and Sierra, with immense expertise and resources in specific use cases, could easily outcompete Zeus's generalized solutions. This risk of churn and the high cost of continuous support or specialized competition made the enterprise AI services model precarious.

THE PIVOT TO DEVELOPER TOOLS: HERCULES

Recognizing these challenges, Falk and his co-founder pivoted, returning to their roots in developer tools, similar to their previous successful venture, Fig. The new company, Hercules, focuses on providing production-ready application development, abstracting away complexities like backend, frontend, and database integrations. This shift aims to leverage AI for mass-market developer enablement rather than bespoke enterprise solutions, offering a more scalable and sustainable business model.

NAVIGATING THE AI DEVELOPER TOOL LANDSCAPE

Falk analyzed the competitive landscape for AI developer tools, categorizing them into AI app builders (like Replit, Bold), website builders (Wix, Squarespace), and core developer tools (CLI-based, like Cursor). He believes general-purpose AI chatbots (like GPT-4) will encroach on prototyping tools, while Wix-like platforms will expand due to lowered barriers. Hercules aims to operate in the productized SAS and specialized tooling space, focusing on white-labeling integrations and abstracting away model choices for a less technical audience, aiming for higher gross margins than engineering-focused CLI tools.

TRANSPARENCY AND THE FUTURE OF STARTUPS

Falk emphasized the value of transparency in the startup ecosystem, sharing his candid experience and pivot to encourage others. He believes that open communication about learnings and failures, despite the prevalence of hype, provides an information advantage. This honesty helps founders navigate challenges and builds trust, suggesting a future where more founders share their journeys, fostering a healthier and more realistic startup culture.

Mentioned in This Episode

●Software & Apps

●Companies

●People Referenced

AI Transformation Pitfalls and New Venture Strategy

Practical takeaways from this episode

Do This

Focus on Global 2000 companies for large AI transformation contracts.

Develop a clear strategy for strategic advisory, custom software development, and change management.

Leverage investor connections for initial outreach to large enterprises.

When building AI tools, prioritize production-ready applications.

For AI app builders targeting non-technical users, focus on ease of use and integrated services.

Be transparent about startup learnings and pivots.

Consider the long-term implications of customer churn and maintenance costs.

Avoid This

Do not underestimate the complexity of custom AI integrations and edge cases.

Avoid focusing on small contracts or a high volume of diverse use cases without repeatability.

Do not rely solely on system integrators like Accenture for AI transformations.

Do not overlook the significant ongoing maintenance costs of custom AI solutions.

Avoid trying to be everything to everyone in the custom AI services space.

Be cautious of the 'existential threat' posed by specialized AI providers (e.g., Decagon, Sierra) in specific domains.

Do not assume AI will fully commoditize software development without considering data quality and human interaction.

Common Questions

Zeus pivoted because the model of building highly custom AI solutions for enterprises proved to be time-consuming and expensive, especially for smaller contracts. The economics didn't scale well, and the risk of specialized competitors outperforming them was high.