NAN128: Beyond RAG: Making Network AI Truly Agentic
Julia Santella, a Cisco solutions engineer with a software background, shares research findings on AI adoption in network automation, emphasizing that AI should augment rather than replace deterministic automation, and that critical skills for 2027 include learning to validate AI outputs rather than blindly trusting them.
Summary
Julia Santella discusses her 10-year journey at Cisco, starting as a software engineer without networking background and progressing through certifications and automation work to become a Hall of Fame speaker. She explains how a dedicated research project allowed her to investigate why network automation adoption remains low despite available tooling, with Gartner data showing two-thirds of tasks still performed manually. Her research revealed critical issues: lack of unified AI definitions in networking contexts (AI for networks vs. networking for AI vs. AI as a feature vs. AI for learning), minimal production adoption of agentic NetOps (less than 1%), and significant trust and governance concerns. Julia advocates for pairing AI's creative reasoning capabilities with deterministic automation to create guardrailed agentic systems. She demonstrates this through examples like using AI to draft configurations while relying on proven Ansible playbooks for execution, and automating network testing where AI analyzes results while deterministic scripts execute verified tests. She emphasizes that the primary risk in 2027 won't be engineers avoiding AI, but those who trust it too completely without validation. Her upcoming talk focuses on network testing as a foundation for automation, arguing engineers should apply the same testing rigor that software engineers use, starting with simple scripts that automate manual checks before deployment. Julia stresses using AI as a learning and writing tool while maintaining critical thinking, comparing it to how she uses it to overcome writer's block in emails and documentation without outsourcing all thinking to AI.
About this episode
Eric Chou welcomes Juulia Santala, a Solutions Engineer at Cisco, to discuss her transition from traditional network programmability to cutting-edge AI research. They explore how she moved beyond simple RAG to make AI truly agentic, while solving some hallucination problems in production configurations. They also cover the number one skill network engineers need by 2027,<a class="excerpt-read-more" href="https://packetpushers.net/podcasts/network-automation-nerds/nan128-beyond-rag-making-network-ai-truly-agentic/" title="ReadNAN128: Beyond RAG: Making Network AI Truly Agentic">... Read more »</a>
Key Insights
- Julia's research found that less than 1% of organizations have adopted agentic NetOps despite extensive community discussion, indicating a significant gap between hype and actual production deployment.
- The term 'AI' creates confusion in networking because it encompasses multiple distinct concepts: AI for networks, networking for AI, AI as a feature in automation, and AI as a learning tool, yet the industry lacks a unified definition similar to how 'source of truth' became standardized.
- Julia argues that the primary risk in the future won't be network engineers who avoid AI, but rather those who trust AI outputs completely without validation, making critical thinking the most important skill for 2027.
- AI is positioned as a non-deterministic reasoning machine that should be paired with deterministic automation systems; Julia illustrates this by combining AI-drafted configurations with verified Ansible playbooks to achieve both creativity and reliability.
- Gartner's research on API usage shows only 10% of developers using LLMs in production despite targeted audiences, demonstrating that adoption rates for emerging technologies remain far lower than community visibility suggests.
- Julia observed that network engineers consistently emphasized trust concerns, with many stating they would only use AI to retrieve information and not execute changes, reflecting the 'releasing a junior engineer to the network without guardrails' anxiety.
- Julia advocates for treating testing as the foundation of network automation before deployment, arguing this approach mirrors software engineering practices where junior engineers begin by writing test suites rather than production code.
- Julia uses AI daily for generating ideas and explanations but validates outputs by discussing findings with trusted network engineers rather than treating AI as a sole source of truth, demonstrating the critical thinking approach she recommends.
Topics
Transcript
Sponsored by Itential's Flow AI delivers agentic operations for infrastructure, meaning easily built AI agents that actually work the way engineers need them to. Governed, deterministic, and built for production. Add intelligent automation to your network operations without the usual AI chaos. Find out more at itential.com slash flowai. That's itential.com. That's itential.com. Hello and welcome to the Network Automation Nerds podcast where we explore the latest network automation from a practitioner's perspective. I'm your host, Eric Cho, a network engineer who loves everything about network automation. With our live coverage from AutoCon 5 in Munich, today the Network Automation Nerds podcast, I am thrilled to welcome Julia Santella. Julia is a solutions engineer at Cisco who is passionate about…
Full transcript available for MurmurCast members
Sign Up to AccessMore from The Everything Feed - All Packet Pushers Pods
TNO071: The Network Team Is Drowning. Is AI the Life Raft? (Sponsored)
Rekha Shenoy and Irfan Kimji from Backbox discuss how the exponential growth of vulnerabilities (49,000 CVEs annually) has made manual network operations unsustainable, and how AI-powered automation can help network teams manage patches and security updates at scale while maintaining human control and oversight.
HN840: How to Make a Technology Buying Decision
Sean Morgan, a research director at Deloro Group, discusses how technology buying decisions should extend beyond engineering specifications to include business alignment, ROI calculations, and understanding total cost of ownership. Engineers must shift from viewing IT as a cost center to positioning it as a business enabler by connecting technical decisions to revenue impact and organizational objectives.
IPB207: Flying Blind: Monitoring Might Not See IPv6
The IPv6 Buzz hosts discuss critical gaps in IPv6 monitoring across enterprise networks, highlighting that many monitoring platforms lack IPv6 awareness, vendor parity, and advanced analytical capabilities. They emphasize that while basic IPv6 data ingestion has improved, sophisticated features like cross-protocol event correlation, extension header analysis, and device identity tracking remain significant industry challenges.
N4N063: Link Layer Discovery Protocol
Link Layer Discovery Protocol (LLDP) is a standardized Layer 2 protocol that enables network devices to announce information about themselves to directly connected neighbors, facilitating network topology discovery and device identification in multi-vendor environments. The protocol uses Ethernet frames with special multicast destination MAC addresses to ensure frames don't propagate beyond immediate neighbors, and includes mandatory TLVs (Type-Length-Values) like chassis ID, port ID, and TTL alongside optional ones for extended information.
TCG083: Superintelligence for Everyone: Who Actually Holds the Power?
Three technology experts discuss Mark Zuckerberg's manifesto on distributed superintelligence, examining whether his promises of universal access and individual empowerment align with infrastructure realities. They conclude that while decentralized AI is theoretically safer than centralized control, the manifesto fails to account for human complexity, existing inequalities, and the enormous capital requirements that will likely concentrate power rather than distribute it.