NA009 - The Square Table — AI, Hype, and the Future of Network Engineering

Network Auto Magic Podcast  

The Square Table — AI, Hype, and the Future of Network Engineering

Episode Overview  

In a first for Network Automagic, Steinn and Urs are joined by seven guests for a no-holds-barred roundtable on AI in the network automation world. The panel tackles what AI has actually changed day-to-day, whether it’s making engineers dumber or just shifting context, the dangers of vibe coding without validation, why AI mandates miss the point, and whether code quality still matters when velocity is the metric everyone chases. Expect sharp takes, zero nodding heads, and a debate about OSPF vs IS-IS that nobody asked for.

Episode Guests  

• Marco Martinez: Network automation specialist and infrastructure architect at Swisscom, focused on cloud-native networking, SDN, and automation — one of the new generation of network architects bred in the Swiss education ecosystem with an automation-first mindset.
  • LinkedIn: Marco Martinez
• Bart Dorlandt: Freelance network automation solutions architect and CCIE — a highly skilled engineer from the Netherlands dedicated to delivering effective customer solutions through automation.
  • LinkedIn: Bart Dorlandt
• John Howard: Head of Network Infrastructure at ProtonMail — a hands-on network architect with deep expertise in cloud, data centres, and large-scale automation, sharing his years of experience through a sharp British mind.
  • LinkedIn: John Howard
• Joseph Nicholson: Network operations engineer at NTT DATA with over 17 years of experience — a grounded engineer who has automated ISP networks for decades and advocates practical, incremental approaches.
  • LinkedIn: Joseph Nicholson
• Mark Prosser: Network operator advocate at Nokia — an advocate for a bright future in network operations, seeing beyond the challenges of network automation and championing innovative approaches.
  • LinkedIn: Mark Prosser
• William Collins: Network automation expert and industry speaker at Itential — a tech evangelist driven by deep understanding of new technology and hunger to learn, focused on building intelligent networks and educating engineers on emerging technologies.
  • LinkedIn: William Collins
• Ryan Shaw: Networking professional with experience in automation and infrastructure — a self-taught unicorn with deep knowledge of networks and programming and a hyperscaler background, bringing a practical perspective on modern network operations.
  • LinkedIn: Ryan Shaw

Listen to the show on YouTube:  

Listen to the show anywhere:  

• YouTube: @networkautomagic
• Spotify: Network AutoMagic
• Apple Podcasts: Network AutoMagic
• RSS Feed: Anchor.fm

Show notes resources:  

• xkcd: Is It Worth the Time? — The automation ROI chart referenced by John Howard
• Make It Stick — Book on the science of learning, recommended by Urs
• The Hedge Podcast — Russ White’s podcast referenced by John Howard, including the 1910 engineering textbook quote
• Network Automation Forum — The NAF Slack community where the panel stays connected

What we cover:  

AI as a Force Multiplier — Not a Replacement  

• The ROI Calculation Has Changed: John Howard argues that AI has shifted the xkcd automation break-even point — tasks that used to take too long to automate are now worth tackling
• Enabler for the Knowledgeable: Bart Dorlandt frames AI as an enabler — you still need the domain knowledge, but AI gets the job done faster when you know what you’re trying to achieve
• Prototyping vs Production: Marco Martinez draws the line — AI makes prototyping less painful, but production quality still requires human verification
• Busy Work Eliminated: William Collins reports AI has knocked off nearly 50% of his busy work — git flows, repetitive processes, administrative overhead
• Two-Hour Builds: Joseph Nicholson shares how a script that would have taken 5–10 days was built and tested in two hours with AI assistance

The Cognitive Shift — Generation to Verification  

• Cosplaying Software Engineers: John argues network automators are network engineers cosplaying as software engineers (or vice versa) — AI is leveling the playing field
• Validation Is the New Job: The panel agrees that AI shifts work from writing code to validating it — which may actually require deeper expertise than writing it in the first place
• Dissenting View on Prototyping: Ryan Shaw pushes back — AI is great for prototyping and scripting, but reviewing 300 lines of generated code for a specific pattern can slow you down considerably

AI Mandates and Workplace Pressure  

• Don’t Be a Dick: John’s advice from a company-wide AI debate — the for-or-against polarization is toxic, and people should be free to choose
• Expectation Inflation: Urs warns that when management reads “prototype in under an hour,” they reset delivery expectations — creating unhealthy pressure
• Context Switching Trap: Research shows AI-enabled engineers become less efficient by taking on too many things at once — the fatigue of context switching is real

Are We Getting Dumber?  

• Knowledge Atrophy, Not Stupidity: The panel settles on Joseph’s framing — it’s knowledge atrophy, not getting dumber. If you don’t use skills, they fade, AI or not
• Context Shifting: Ryan argues knowledge doesn’t shrink, it shifts — you have to be deliberate about what you keep sharp and what you let go
• The Next Generation: Bart raises concern about engineers who’ve never worked without AI — they build skills differently, consume information differently, and may struggle if AI becomes unavailable
• Critical Thinking Is Non-Negotiable: The panel unanimously agrees critical thinking matters more than ever when information volume explodes
• The 1910 Engineering Quote: John shares a century-old insight — “Memory is a poor reliance to the engineer. Your facts are better stored in your library room.” Engineers were never meant to memorize everything

Vibe Coding and the Ownership Problem  

• Neural Fatigue: William explains the neuroscience — AI coding locks you into “judge mode” all day, burning your executive function while the creative default mode network never fires, leading to exhaustion and disconnection from your own work
• Disposable Code Culture: John warns that when you don’t feel ownership of AI-generated code, you skip testing, skip validation, and ship things you’d never stand behind
• True Equals True: The panel mocks the pattern of AI-generated tests that only validate happy paths — “200 tests that confirm an integer is an integer”

Code Quality — A Pricing Tier, Not a Moral Virtue  

• Script vs Service: Ryan draws the distinction — a one-off migration script can be hacky, but a production service demands quality
• Bad Code Has Always Existed: Joseph and John agree — developers shipped bad code long before AI, and vendors have been patching bugs forever
• Invest the Time Savings in Quality: Marco’s closing call to action — if AI makes you 30% faster, use that time to make the output better, not just to produce more

Key Takeaways  

1. AI Is a Tool, Not a Personality: Stop projecting personas onto language models — give them clear instructions and good patterns, not ego and identity
2. Critical Thinking Is the Meta-Skill: With exponentially more information (and misinformation), the ability to evaluate and verify is the skill that separates professionals from passengers
3. Don’t Hype, Don’t Naysay — Find the Middle: William draws the parallel to cloud computing’s journey from “replace everything” to finding a balanced reality
4. Quality Eats Volume for Breakfast: If AI gives you velocity, invest it back into testing, validation, and code quality — not just more output
5. Adapt, Measure, and Make Good Choices: Ryan’s consistent message — embrace the tools, but be deliberate about what context you keep and what you drop
6. The Floor Dropped, the Ceiling Stayed: AI lowered the barrier to entry but didn’t change what good work looks like — the standard remains the standard