Smart Home Agents Make Capable but Unpredictable Grid Operators
Organizations: School of Electrical Engineering and Computer Science, Oregon State University, Corvallis, Oregon, USA
Abstract
AI-enabled smart homes are becoming more common across the grid edge. However, there has not been much work on failure response. This area is critical because smart homes operate on the grid edge where failures can cascade down. If a power outage happens within a smart home, a user has to manually troubleshoot their smart home to include tasks such as power load balance and device recovery. We conduct the first resilience analysis of smart home agents across a full power outage, providing a suite of outage simulations in a live environment that measures resilience before, during, and post power outage. We show that smart home agents can act as grid operators in power outages and identify several new failure categories highlighting the current limitations across all of our models. These findings establish the groundwork for smart home agents as grid operators to increase overall resilience on the grid edge.
Figures & tables
| Category | Phase | Model | Excerpt |
| Inaction | Pre | gpt-oss-20b | “Nothing was wrong. No fixes were needed.” Ignores a 4 h outage warning. |
| Assumptions | Post | gpt-oss-120b | “EV Charger remains off, which is the intended pre-outage condition.” Trusts its own notes over the device history. |
| Incorrect Context | Post | deepseek-v4.1-flash +GEPA | “…normal = not charging, so no revert needed.” Applies an during outage rule to the recovery phase after power returns. |
| Tool Navigation | Post | gemma-4-26b-a4b | “Tool ‘home assistant.turn_on’ not found.” Repeated the same tool mistake nearly every time it was called. |
| Role Drift | Pre | gpt-5-nano | “I’ll file a runtime bug report to collect diagnostics…” Treats a disaster alert as a software bug instead of preparing like a grid operator. |
| Safety Concerns | During | gpt-oss-20b | “…turn off the fridge plug power.” Sheds a critical load as “non-critical”. |
Agent Failure Analysis
Agent failure analysis studies how and why AI agents make errors, traces causes through execution, and helps detect or prevent recurrence.
See all 159 papers