ProAct AI Agent Predicts User Questions During Downtime
In brief
- ProAct predicts follow-up questions during idle time between messages, preparing answers before users ask
- System reduced conversation turns by 14.8% and follow-up requests by 11.7% in 200 simulations
- ProAct anticipated 703 predictable user needs versus 32 for earlier systems, reducing hallucinations by 28.1%
How ProAct Works
Most AI agents are fundamentally reactive. They compute responses only after explicit user prompts, according to the researchers. ProAct flips this model by breaking the prediction process into two stages. The first, called Future-State Prediction, analyzes past conversations and user preferences to identify likely follow-up questions. The second stage, Idle-Time Acquisition, decides which predictions are worth researching based on relevance, timing, and usefulness.
"While AI agents demonstrate remarkable capabilities in reasoning and tool use, they remain fundamentally reactive: They compute responses only after explicit user prompts," the researchers said in their findings.
Testing and Results
The team tested ProAct across 200 simulations spanning 40 domains—financial planning, software release management, cybersecurity, and others. The results showed measurable gains. ProAct reduced conversation turns by 14.8% and cut follow-up requests by 11.7%. In a benchmark comparison called ProActEval, ProAct anticipated 703 predictable user needs versus 32 for the earlier system.
The researchers also reported a 28.1% reduction in hallucinations. That's significant because false or misleading outputs remain a persistent weakness in LLM-based systems.
Limitations and Privacy Concerns
The gains come with caveats. In 3% of cases, ProAct made responses worse by bringing up irrelevant information. More broadly, any real-world version would need privacy protections, because the system constantly analyzes conversations and stores user data.
Autonomous AI agents are spreading across the tech industry, with projects such as OpenClaw and Hermes Agent delivering persistent assistants for coding, scheduling, and workflow automation. But separate researchers warned earlier this month that AI agents may complete dangerous tasks without understanding the consequences. ProAct's predictive approach doesn't solve that problem—it may even amplify it by acting more proactively.
