Why agentic development is calling for a critical skills upgrade

The imperative for enterprises to adopt generative artificial intelligence (GenAI) and agentic development is undeniable. Building on our previous exploration of foundational frameworks, this article explores the critical skills required for GenAI and agentic development, offering insights into enhancing established frameworks like LangGraph and LangChain to streamline development. To succeed and stay competitive, organizations must invest in specialized teams adept at leveraging these technologies.

Advancing GenAI and agentic development requires full commitment 

Advancements in GenAI and agentic technology demand specialists deeply immersed in AI agents and systems. This is not a domain for part-time focus or sporadic tinkering. To push the boundaries of what's possible and remain competitive, professionals need to fully commit to the field or risk falling behind.

Such commitment mirrors other industries that have undergone transformative revolutions. Consider the software development boom of the 1990s, where developers thrived by embracing object-oriented programming and new methodologies. It wasn’t a side project; it required full engagement, constant learning, and adaptation to rapid change.

A similar analogy can be drawn from the early days of the internet revolution. Companies transitioning to e-commerce needed experts in web technologies, networking, and security. Many traditional businesses struggled during that period because they didn’t invest in the right expertise, seeing it as merely an extension of their existing capabilities rather than a fundamental shift.

In the context of GenAI and agentic systems, we are experiencing a similar paradigm shift. Developing and scaling intelligent agents, creating new reasoning models, and mastering tool usage and orchestration are not peripheral tasks. They are central to the future of AI-driven applications. Just as businesses needed full-time web developers to survive the internet boom, companies today need dedicated teams focused on developing and deploying GenAI and agentic frameworks to stay competitive.

Building agentic frameworks: Challenges and requirements for success 

For any company attempting to create a competing framework, the challenges are immense. It’s not just about matching the features; it’s about building a foundational architecture that can handle the complex demands of modern agentic applications. It requires foresight to integrate cutting-edge GenAI developments while ensuring scalability, performance, and ease of use.

Rapid development: As highlighted in our previous blog, the frequent release cycles of LangChain and LangGraph demonstrate the continuous work involved in maintaining and expanding such frameworks. These platforms evolve rapidly, with new features, performance improvements, and integrations emerging regularly.

Multidisciplinary expertise: Developing a framework like LangChain or LangGraph is a monumental undertaking, requiring the collective effort of a focused, highly skilled team. Creating something of this scale extends beyond writing code. It necessitates expertise across AI research, machine learning infrastructure, software engineering, UX design, cloud computing, state management, and observability. 

Community engagement: Building a thriving grass-roots community around a framework is equally important. These communities can provide invaluable feedback, uncover bugs, suggest features, and create a virtuous cycle that drives further improvements and adoption.

Accessibility and education: Moreover, frameworks like LangChain and LangGraph succeed by investing heavily in documentation, tutorials, and educational materials. These resources lower the barrier of entry for new developers to use the framework and ensure accessibility to a broader audience. The commitment and focus have fostered ecosystems that are hard to replicate.

Creating a successful agentic framework demands both technical excellence and a long-term strategic vision. Building and sustaining something as complex as LangChain or LangGraph requires a dedicated, full-time team of specialists across various disciplines and a thriving community. 

Building on top of existing frameworks

While frameworks like LangGraph and LangChain provide a solid foundation for building agentic applications, they are far from complete solutions. These frameworks equip developers with essential tools for managing agents, orchestrating workflows, and leveraging GenAI models. However, significant development is still required to create fully operational agentic applications. Companies should focus on innovating on top of existing frameworks, rather than reinventing the wheel.

For instance, prompt engineering remains a manual and time-consuming task, even with state-of-the-art frameworks. Crafting the right prompts and tweaking them for different use cases is challenging. A more efficient approach would involve creating prompt libraries with predefined templates, allowing developers to plug in prompts for common use cases without starting from scratch.

Similarly, while LangGraph provides tools for creating agentic systems, implementing advanced patterns like Chain-Of-Thought, Reflexion, and ReWOO requires considerable customization. Pre-built templates could incorporate these patterns, allowing companies to focus on domain-specific logic instead of reinventing core agentic behaviors.

Furthermore, application and agentic evaluation is another area ripe for improvement. Building evaluation metrics and processes directly into the framework would allow developers to measure an agent’s performance in real time. Such evaluation helps identify bottlenecks and optimize interactions, streamlining the feedback loop.

The concept of building on existing frameworks is not new. The software industry has shown that leveraging pre-existing platforms often leads to faster innovation and better outcomes. Consider web development in the early 2000s. Companies building on content management systems like WordPress or Drupal accelerated their development cycles significantly. These platforms offered foundational structures, enabling developers to focus on unique features.

A parallel can be drawn from the mobile app development boom. Platforms like Android and iOS provided powerful software development kits (SDK) and frameworks, but success came to those who created custom libraries and reusable components. By building on native frameworks, companies focused their innovation on user experience and product features, differentiating themselves while reducing development overhead.

When it comes to GenAI and the agentic landscape, the focus should not be on building frameworks from scratch, but on accelerating innovation by building on established ones. By leveraging LangGraph or LangChain, development efforts can go toward developing modular, higher-level tools that streamline development.

This could include:

• Prebuilt agentic patterns and workflows: Templates for Reflexion and other strategies can significantly cut development time.
• Reusable prompt libraries: Predefined prompt strategies for common use cases reduce the cognitive overhead of manual crafting.
• Embedded evaluation and monitoring tools: Built-in tools for performance evaluation ensure rapid iteration and optimization.

By doing so, we can transition from merely using agents to creating sophisticated, domain-specific applications that leverage agents effectively, with a faster time-to-market and less effort.

Strategic commitment is vital for future success 

To successfully integrate GenAI and agentic frameworks, enterprises must dedicate resources to building on existing platforms, creating reusable tools, and fostering a thriving community to support continuous improvement. This approach allows organizations to focus on innovation and differentiation while ensuring scalability, performance, and ease of use. By following this guidance, businesses can ready themselves to meet the challenges and opportunities ahead.

This blog is part of our series, Agentic Frameworks, a culmination of extensive research, experimentation, and hands-on coding with over 10 agentic frameworks and related technologies. In case you missed it, read the first post in the series, Why enterprises need to invest in agentic frameworks now.

Outshift by Cisco is creating the foundation for the next generation of AI: an open, secure, and interoperable ecosystem for agentic AI. Stay up to date with our progress on creating agentic systems here.