AI for DevOps & SRE: From Impressive Demos to Operational Trust
Description
AI for DevOps & SRE: From Impressive Demos to Operational Trust
What would make you trust an AI agent with operational changes?
An agent can generate a valid command and still misunderstand the objective. It can mistake missing telemetry for a healthy service, or choose an action outside the scope you authorized. How do we recognize these failures before they affect real systems?
Event
Join Bucharest XDevOps Meetup for a practical, vendor-neutral session exploring the risks of AI-assisted operations and the controls that help manage them.
Schedule
We’ll work through three synthetic scenarios in isolated environments:
- The command is valid, but the objective is wrong.
- The query returns no data, but the agent draws a confident conclusion.
- The agent selects a tool outside the operator’s explicit boundaries.
For each scenario, we’ll examine the evidence, identify where the workflow breaks, and revisit it with appropriate controls: scoped permissions, validation before execution, meaningful human approval, audit trails, and independent verification of the outcome.
We’ll also discuss what these controls cannot guarantee. A permitted action can still be a poor decision. Human approval needs context. A successful tool call does not prove that the problem is resolved.
Who it’s for
DevOps engineers, SREs, platform engineers, developers, and technical leaders exploring AI-assisted operations. Familiarity with infrastructure operations is helpful; experience building AI agents is not required.
What you’ll take away
A practical checklist for evaluating an AI-assisted workflow: what it is allowed to do, what evidence supports its decisions, when it should stop or ask for approval, and how to verify what actually happened.
Proposed agenda — 90 minutes
- 0–10: Introduction — what does operational trust require?
- 10–30: Three failure scenarios — objectives, evidence, and action boundaries.
- 30–55: Practical controls — permissions, validation, approvals, audit, and outcome verification.
- 55–75: Revisit the scenarios — what the controls catch and what remains uncertain.
- 75–90: Discussion and Q&A.
Bring your questions and experiences with AI in incident response, observability, infrastructure, or delivery pipelines.
- Education
- Сourse
- Master class
- For teens
- For kids
- Lectures
- For adults


Matchmaking for Serious Relationships
More from this organizer

Gamedev Breakfast
21.11.2026

Workshop 'The Secret of Lepenski Vir'
21.11.2026

Layzie & 2much
21.11.2026

Magla Pre-Weekender with Zadig
22.11.2026

Winter Preserves Caravan at Đeram Market
22.11.2026

Voluntary Blood Donation - Šabac, Ribari
22.11.2026

7th Belgrade Half Marathon
22.11.2026

Open House Day at Enter Zemun
22.11.2026
Similar events

AI SPY — documentary film, debate and networking
Sunday, November 29th
18:00

AWS User Group KRK meetup #78 with Remitly
Thursday, November 26th
18:00

SEO Talks #5
Thursday, November 26th
18:30

Introduction to Machine Learning: Digit Recognition — Lesson 5
Monday, November 23rd
20:00

Sinergija 26 - Best Practices for FIDO2 Passkeys in Microsoft Entra ID
Wednesday, November 25th
13:30

RSD 800.00
Book Club: Popular Science
Thursday, November 26th
19:00