Welcome to Part 3 of our series on Agentic Workflows and AI Agents. In Parts 1 and 2, we explored how AI agents are reshaping workflows and the design patterns guiding their development.
Today, in our final part, we’ll dive into the practical blueprints businesses need to unlock the full potential of AI, and how this transformation is also driving significant success across our portfolio companies—from breakthrough innovations to impressive market validation.
Are AI Agents the Future of Work?
Does OpenAI’s new o1 model mean AI agents will replace all white-collar jobs? While AI's capabilities are rapidly evolving, this isn’t about replacing humans—it’s about rethinking the way we work.
AI agents are already automating structured, repetitive tasks, freeing up employees for more creative and strategic roles. Businesses that successfully implement AI agents are redefining entire roles and organizational structures.
For example, Tembo is making it easier to build AI applications on PostgreSQL with its platform, Tembo AI, which integrates embedding and chat models to streamline development and reduce the manual effort needed for database operations.
To make the most of these advancements, businesses must adopt AI-native approaches, rethinking workflows from the ground up.
The Power of AI-Native Workflows
AI-native workflows go beyond simply adding AI to existing processes.
They involve creating entirely new systems where AI drives every step, making operations more autonomous.
Our portfolio companies are at the forefront of this transformation by embracing AI-native approaches that redefine how work gets done.
Meanwhile, Datalogz is transforming business intelligence (BI) by providing AI-driven optimizations that streamline workflows for Fortune 500 companies.
By making data analysis faster and more actionable, Datalogz enables businesses to extract more value from their data operations with minimal manual effort.
The diagram below (credit to Kelvin Mu) represents one of the ways Defined has been inspired to think about capturing value in the AI agent ecosystem.
We are currently focusing our investments and discovery in the “mixture of agents layer”.
Companies such as Quandri and Arcanna AI exist within the vertical agent layer, focusing on brokerage and cybersecurity applications, while Inverted AI operates in the embodied agents layers, with applications in autonomous driving systems.
Zyphra spans both the vertical agent and embodied agents layers, pushing the boundaries of on-device AI models and foundational AI systems.
The AI Scientist: A Glimpse into the Future
AI-native approaches are best illustrated by projects like Sakana AI’s AI Scientist, which represents the next frontier of AI development.
This system can independently conduct scientific research, from hypothesis generation to experiment design and data analysis.
Similarly, in our portfolio, companies are taking bold steps to apply AI in groundbreaking ways. ViewsML, for instance, is using AI to revolutionize immunohistochemistry.
By partnering with two of the world’s top 10 pharmaceutical companies, the company is validating its product in real-world, high-stakes environments, bringing AI-driven precision to medical diagnostics.
The Role of Code Generation in AI Agent Development
Code generation is another essential piece in the rise of AI-native applications, enabling the rapid development of AI-driven platforms.
AI-powered code editors like Cursor and Replit’s AI agent are transforming the development process by providing context-aware code suggestions and natural language coding—allowing developers to describe features in plain English, and instantly receive generated code.
This shift is echoed in companies like Arcanna AI, which is in advanced discussions with consulting firms to co-sell its agentic cybersecurity solutions.
By leveraging AI-driven automation in threat detection, Arcanna AI is demonstrating the value of integrating code generation and AI-native approaches into critical, security-focused workflows.
The Path Forward
As businesses embrace AI agents and agentic workflows, it’s essential to approach this transformation gradually.
The gradual integration of AI agents will likely follow a "crawl, walk, run" trajectory. Initially, AI will handle simpler, more routine tasks before progressing to more complex workflows as systems evolve and trust in AI capabilities grows.
For now, many deployments still operate within a human-in-the-loop framework, where AI agents function as advisors or assistants, with humans retaining a significant role in decision-making and oversight.
As AI continues to mature, these systems may evolve into proactive execution, where AI agents can independently recognize the need for workflows and execute them with minimal human intervention.
By adopting AI-native approaches and integrating these advanced systems, businesses can harness the power of AI to reshape industries, drive innovation, and unlock new possibilities.
If you are a founder or know of people working to increase automation with agentic workflows and further the capability ladder of AI Agents, then please reach out to connect.