Design-First, Trust-Always: Summarizing what we learned at AutoCon3

Synthesizing the complete picture from all 24 presentations.

AutoCon 3 in Prague brought together 650+ network automation practitioners for what Jason Edelman called "coming home to the network automation family." Through 2 keynotes, 13 presentations, 3 sponsor updates, and 6 lightning talks, several major themes emerged that help define the current state and future direction of network automation.

I. THE FUNDAMENTAL QUESTION: WHY HASN'T AUTOMATION BEEN FULLY ADOPTED?

Perhaps unsurprisingly, this core question echoed throughout the conference, with speakers offering different but complementary answers:

Trust and User Experience (Damien Garros): "I truly believe that trust is actually one of the main issues." Automation fails when builders create "black boxes" that work for them but inspire no confidence in users.

Foundation Problems (Claudia de Luna): "We went from network automation with a little 'n' and little 'a' to Network Automation with capital letters, but really, what we've been talking about all this time is software development." Many initiatives fail because they skip foundational design work.

Tool Limitations (Josh Saul): "We haven't had good tools. We've had rough knives, we've had hammers when we needed screwdrivers."

Skills and Learning Barriers (Emre Cinar): Junior engineers face overwhelming choices and limited guidance, while the community often makes assumptions about baseline knowledge.

Enterprise Realities (Robert Blake): "Perfect is the enemy of good enough" - organizations get paralyzed trying to solve every edge case instead of making incremental progress.

II. DESIGN-FIRST PHILOSOPHY: THE FOUNDATION FOR SUCCESS

The Strategic Imperative

Claudia de Luna established the conference tone: despite years of progress, many automation initiatives "still feel fragmented and disconnected from clear business value" because they lack proper foundational design. Network automation jumped straight into operations without asking fundamental questions about requirements and design.

Christian Adell Querol provided the methodology: "We have to find a way where design matches automation." His design-driven approach addresses "the gap between what network architects envision and what gets implemented" by codifying design logic and protecting design-driven data.

Jaakko Rautanen demonstrated this through service abstraction: "When we drive a car, do we think about every detail the car has? We don't. Instead, we just think about steering wheel, pedals, and dashboard." Network operations should focus on services rather than device details.

Design Implementation Patterns

III. THE MATURATION OF AUTOMATION PRACTICES

From Scripts to Software Engineering

Antonio Balmaseda & Alberto Villegas Erce delivered perhaps the conference's most provocative message: "We were basically trying to grow unicorns... You can see a unicorn once, maybe twice in your lifetime... But you cannot grow a unicorn." Stop trying to turn network engineers into developers; embrace specialization with proper software development practices.

Shirish Basant Rai showed the evolution path: "Most organizations won't jump directly from Ansible playbooks to microservices. They'll follow a similar progression: scripts → centralized automation → orchestration → microservices."

Key Practices Emerging

The Realistic Journey

Robert Blake emphasized patience: "It's not a failure if something has to change. You've learned lessons along the way." Enterprise automation requires marathon thinking, not sprint expectations.

IV. THE TOOLS LANDSCAPE: CHOOSING AND INTEGRATING WISELY

Framework Selection Wisdom

Suhaib Saeed provided practical guidance: "If you're just starting your automation journey, Ansible's gentle learning curve might be perfect. If you're ready to embrace Python fully and need maximum performance and flexibility, Nornir could be your answer."

His decision framework emphasized matching tools to team capabilities rather than seeking universal "best" solutions.

Infrastructure-as-Code Realities

Eduardo Pozo shared his "epic journey" with Terraform, revealing hard truths: 99% of vendors have "half-baked providers," performance optimization requires architectural decisions, and "anyone trying to go into network automation thinking they're going to directly save money will see the project as a loss."

The value comes from standards enforcement, error prevention, and operational consistency; not immediate cost savings.

Specialized Domain Solutions

Matteo Colantonio transformed GARR's optical network operations from "50 clicks to 50 seconds" by building custom solutions when traditional tools fell short. His lesson: "Automation doesn't have to conform to popular tool limitations."

V. ARCHITECTURAL EVOLUTION: FROM MONOLITHS TO SERVICES

Microservices for Enterprise Scale

Shirish Basant Rai demonstrated the evolution from centralized workflows to distributed microservices architecture. The benefits: independent development and deployment, targeted scaling, conflict prevention, and resilience.

Workflow Orchestration for Complex Operations

Naveen Achyuta introduced Temporal for complex, long-running processes: "What makes Temporal compelling for network automation isn't just technical capabilities; it's the recognition that complex network operations are inherently distributed systems problems."

Service-Oriented Thinking

Multiple speakers converged on service-oriented approaches:

VI. OPERATIONAL EXCELLENCE: SCALE AND RELIABILITY

Large-Scale Operations

Luke Gollan shared lessons from migrating 7,000 services at Megaport: "You need to think about resilience, not just functionality." Their 100% automated network demonstrates that operational automation at scale requires different thinking than basic configuration management.

Day-Two Operations Challenge

NetBrain addressed the often-overlooked operational side: "We talk a lot about the build side, but when you hand it over to the run team, it was always a nightmare." Their analysis showed 50% of incidents are avoidable, and 95% of real incidents are repeats; both addressable through automation.

The Snowball Strategy

Lee Harper demonstrated incremental automation: "Like a snowball rolling downhill, your project grows and your resources grow." Starting with inventory management and basic standardization freed resources for more complex automation.

VII. SECURITY AND TRUST: THE MISSING PIECES

Security Integration by Design

John Howard addressed the critical gap: "Network automation without proper security creates systematized vulnerabilities, insecure practices get codified and replicated across infrastructure." His practical guidance using Vault/OpenBao shows that secure automation isn't significantly harder than insecure automation.

The Accidental Security Champion

Lee Harper discovered unexpected benefits: "We didn't go into this thinking of automation as a security tool," but saw security audit findings drop from 34 to 9 while growing 50% without adding staff. The lesson: operational excellence initiatives can inherently improve security posture.

Building Trust Through Design

Damien Garros outlined the fundamental challenge: "What if I give you a black box with a red button on it and ask you to press this button? I tell you it's going to fix everything. Just trust me. How many of you are going to press that button?"

His six principles for trustworthy automation:

  • Predictable: Consistent, expected behavior

  • Manageable: Easy to control and maintain

  • Transparent: Users understand what will happen

  • Simple to Use: No 50-page documentation requirements

  • Reliable: Works consistently without surprises

  • Human-Friendly: Designed for operators, not just technical correctness

VIII. THE AI REVOLUTION: HYPE VS. REALITY

The Coming Transformation

Itential (Peter Sprygada) predicted fundamental change: "When this whole AI thing started, I was like, no, there is no way AI is ever going to touch a network." But the introduction of Model Context Protocol changed everything. Organizations embracing AI-native network operations will see "significant increases in their ability to be agile, responsive, and build scalable infrastructure."

Practical AI Applications

Urs Baumann demonstrated realistic implementation: fine-tuning LLMs for network intent extraction achieved 100% JSON validity and 86% data extraction accuracy. Rather than wholesale AI adoption or rejection, Baumann advocated for thoughtful application.

Autonomous Correlation

Javier Antich pushed toward "zero-day event correlation" through semantic understanding: current rule-based approaches can only address "the head of the curve" while autonomous systems can handle novel scenarios without explicit programming.

IX. DATA AND MONITORING EVOLUTION

Beyond Screen Scraping

Bart Dorlandt showcased the evolution from reactive polling to proactive push models: "You've been looping through interfaces and neighbors with regex. Imagine doing that continuously on a service provider network." BMP enables efficient, real-time data streams that scale to service provider requirements.

Multi-Vendor Consistency

Josh Saul addressed practical SONiC challenges: "Multi-vendor SONiC is not one distribution. It's one operating system customized for different vendors." Abstraction APIs enable consistent operations across vendor implementations.

Data-Driven Decisions

Jason Edelman noted the transformation: "The vast majority of talks this week mentioned data or source of truth." The emergence of NetBox, Nautobot, and InfraHub created healthy competition while establishing data modeling as fundamental to automation success.

X. ECONOMIC AND BUSINESS MODEL DISRUPTION

Cost Structure Innovation

Henry Haller demonstrated business model disruption: "$70 access points with enterprise features" through automation and open source. "Automation doesn't just improve existing operations; it can enable entirely new economic models."

Enterprise Platform Evolution

Gluware addressed integration challenges: "There's a new voice at the table, and that voice is the Automation Builder." Rather than tool replacement, they advocate for platforms that enhance existing automation investments while providing development-grade tooling.

XI. COMMUNITY AND LEARNING CHALLENGES

The Beginner's Perspective

Emre Cinar provided crucial insights: "Some people assume everybody knows things, and the one person who doesn't is afraid to ask." His perspective is valuable precisely because it reveals barriers that veterans have forgotten exist.

His four characteristics of "consumable" tools:

  • Minimal Dependencies: Fighting installation issues shouldn't be the learning bottleneck

  • Documentation: Real-world examples, not just theoretical guidance

  • Supportive Community: Safe spaces to ask basic questions

  • Real-World Examples: Understanding how tools apply to actual problems

Skills Evolution

Jason Edelman outlined evolving baseline skills: "Linux, Git, a little Python, making API calls" become table stakes for network engineers, not optional advanced skills.

Historical Perspective

The community has evolved from individual scripts to collaborative frameworks. Key enablers include community projects (NetMiko, NAPALM, Ansible), vendor support (Cisco DevNet), and educational resources. Container Lab emerged as the standout recent project for accelerating learning.

XII. SYNTHESIS: THE PATH FORWARD

Success Patterns That Emerged

  1. Start with foundations and design, not tools (Claudia de Luna)

  2. Build for trust and transparency, not just functionality (Damien Garros)

  3. Embrace team specialization, over individual transformation (Antonio & Alberto)

  4. Think in services and abstractions, not devices (Jaakko Rautanen)

  5. Plan for the marathon, not sprints (Robert Blake)

  6. Integrate security by design (John Howard)

  7. Use the snowball strategy for incremental progress (Lee Harper)

  8. Choose tools for team reality, not theoretical ideals (Suhaib Saeed)

Critical Mindset Shifts

  • From automation to autonomy: Systems that adapt rather than just execute (Javier Antich)

  • From device to service focus: Business value over technical capability (Luke Gollan, Jaakko Rautanen)

  • From builders to users: Designing for operators, not just implementers (Damien Garros)

  • From tools to trust: Reliability and confidence over functionality (Damien Garros)

  • From individual to community: Shared solutions and knowledge transfer (Jason Edelman)

  • From scripts to software: Professional development practices and team approaches (Antonio & Alberto)

The Real Challenge

Multiple speakers converged on the same insight: the technical problems are largely solved. The remaining challenges are human, organizational, and cultural:

  • Trust: Making automation systems that inspire confidence (Damien Garros)

  • Foundations: Starting with design rather than rushing to tools (Claudia de Luna)

  • Learning curves: Making automation accessible to practitioners at all levels (Emre Cinar)

  • Enterprise constraints: Working within real-world organizational limitations (Robert Blake)

  • Team dynamics: Balancing specialization with collaboration (Antonio & Alberto)

XIII. LOOKING FORWARD: THE NEXT PHASE

Technical Evolution

  • AI integration for autonomous operations and intelligent analysis

  • Service-oriented architectures replacing device-centric thinking

  • Workflow orchestration for complex, multi-step processes

  • Semantic understanding for handling novel scenarios

  • Push-based monitoring replacing polling architectures

Industry Maturation

  • Professional practices becoming standard across the industry

  • Security integration from design rather than afterthought

  • Trust and reliability as key differentiators between solutions

  • Community growth and knowledge sharing at scale

  • Specialization in teams and tooling approaches

The Acceleration Point

As Jason Edelman noted, "The foundational work is done; tools exist, communities are established, and early adopters have demonstrated value. But mass adoption remains elusive. The challenge isn't technical anymore. It's cultural, organizational, and educational."

The next phase requires wisdom about what to automate, how to build trust, and when to embrace emerging technologies like AI while maintaining operational reliability.

XIV. THE CONFERENCE'S GREATEST INSIGHT

AutoCon 3 demonstrated that network automation has evolved from experimental scripts to business-critical infrastructure. But perhaps the conference's most important insight came from recognizing that automation success depends as much on psychology, trust, and user experience as it does on technical correctness.

Building systems that people confidently use matters more than building systems that perfectly work in isolation.

As Damien Garros concluded: automation isn't just about making things work; it's about making things work in ways that others will trust and adopt. That trust, more than any technical advancement, may determine automation's ultimate success in transforming network operations.

And as Claudia de Luna reminded us in her opening keynote: "It begins with a dream. We have to dream that there is a better way. Once we dream that, we have to believe we can go do it. Once we have that belief, it's the fun part; we have to go do it and make it happen."

The future of network automation won't be determined by better tools alone, but by our collective wisdom in applying those tools to create systems that people trust, understand, and enthusiastically adopt.


Complete AC3 Speaker Index

Keynotes:

  1. Claudia de Luna - "Start at the Beginning: Why Network Automation Needs Better Foundations"

  2. Jason Edelman - "The Long Arc of Network Automation: Lessons from a 25-Year Journey"

General Session:

  1. Shirish Basant Rai - "From Scripts to Microservices: The Evolution of Enterprise Network Automation"

  2. Emre Cinar - "The Fresh Eyes We Need: What Network Automation Looks Like to a Beginner"

  3. Robert Blake - "The Long Game: Why Enterprise Campus Automation is a Marathon, Not a Sprint"

  4. Antonio Balmaseda & Alberto Villegas Erce - "Stop Trying to Grow Unicorns: Why Network Automation Needs Software Developers"

  5. Luke Gollan - "The Art of Network Surgery: Migrating 7,000 Services Without Breaking Anything"

  6. Matteo Colantonio - "From 50 Clicks to 50 Seconds: Automating Optical Networks Beyond Traditional Tools"

  7. Naveen Achyuta - "Beyond Scripts: How Temporal Transforms Complex Network Workflows"

  8. Eduardo Pozo - "The Network Automation Epic: Terraform's Journey from Cloud to Campus"

  9. Lee Harper - "The Security Side Effect: How Network Automation Accidentally Transformed Our Risk Posture"

  10. Javier Antich - "The Correlation Conundrum: Why Current Approaches Fall Short and What Comes Next"

  11. Jaakko Rautanen - "The Car Analogy: Why Network Abstraction Matters More Than You Think"

  12. Christian Adell Querol - "Beyond Data: Why Design-Driven Automation is the Missing Link"

  13. Damien Garros - "The Black Box Problem: Why Trust is the Missing Piece in Network Automation"

Sponsor Presentations: 

  1. Itential - "The AI Revolution We Didn't See Coming: Why Network Automation Is About to Change Forever"

  2. Gluware - "Beyond the Island of Automation: Bridging Complex Networks with Purpose-Built Platforms"

  3. NetBrain - "Beyond Building Networks: The Day Two Automation Challenge"

Lightning Talks:

  1. Suhaib Saeed - "When to Choose Nornir Over Ansible (and Vice Versa)"

  2. Josh Saul - "Taming Multi-Vendor SONiC with Abstraction APIs"

  3. Henry Haller - "Disrupting Enterprise Wi-Fi with Automation and Open Source"

  4. Bart Dorlandt - "Escaping the Screen Scraping Trap with BGP Monitoring Protocol"

  5. John Howard - "Security's Missing Link in Network Automation"

  6. Urs Baumann - "Fine-Tuning LLMs for Network Configuration Intelligence"

Chris Grundemann

Executive advisor. Specializing in network infrastructure strategy and how to leverage your network to the greatest possible business advantage through technological and cultural transformation.

https://www.khadgaconsulting.com/
Next
Next

Fine-Tuning LLMs for Network Configuration Intelligence