Not every good conversation at a database conference stays on databases. On the floor at RoachFest London 2026, fresh off the morning workshop he led, Felipe from Cockroach Labs and I ended up somewhere unexpected: what actually happens when AI agents take over everything.
Where Agents Are Today
Felipe’s starting point was grounded rather than breathless: agents today are still mostly in the “simple tasks” phase — surfacing information, drafting and sending emails, the kind of work that saves minutes rather than restructures a job. The open question that actually matters is when that shifts from simple tasks to genuinely autonomous decisions, and whether anyone will be ready when it does.
Automation Creep
The more interesting thread was what he called automation creep — self-driving vehicles, robotic grocery delivery, the steady accumulation of small conveniences that each individually make sense, but that collectively nudge a society toward doing progressively less for itself. None of these are dramatic on their own. Together, they describe a direction.
The WALL-E Analogy
His honest analogy landed harder than most conference soundbites: the humans in WALL-E — passive, dependent, physically atrophied, because literally everything is already handled for them by machines. It is a cartoon, but the mechanism it depicts is not implausible: convenience compounds, and nobody consciously decides to stop doing things for themselves, they just stop needing to, one automated task at a time.
“I hope we don’t get there any time soon” — his words — but he was clear that he can already see the path being laid, not as a distant hypothetical but as a visible trajectory from where the industry stands today.
Why This Perspective Is Worth Taking Seriously
What makes this worth writing up rather than filing under conference small talk is who was saying it. This is not a skeptic outside the industry — it is someone actively building the database infrastructure that tomorrow’s AI agents will run on, at a company betting its entire product roadmap on agentic workloads scaling into the trillions. A candid acknowledgment of the endpoint risk from someone with that much skin in the outcome is worth more than either the boosters or the doomsayers who have no infrastructure at stake either way.
Related Reading
- RoachFest London 2026: Distributed SQL Meets AI Resilience
- Inside Cockroach Labs’ AI Playbook: The Database Reckoning
- Production Guardrails for AI Agents
About the Author
I am Luca Berton, AI and Cloud Advisor. I work at the intersection of distributed systems, platform engineering, and enterprise AI deployments. Book a consultation.
