LLM-based multi-agent systems have ushered in a new era of agent simulation. As new tools, platforms and computational resources become available, we expect to see increased use of agent simulation in organizational decision-making.

Agent simulations model individual choices and interactions between agents.
- Suitable for dynamic systems where behaviour changes the environment
- Can be computationally expensive
- Sensitive to design: agent guidelines and inter-connections
- Custom implementation often required

Monte Carlo simulations model uncertainty by sampling from probability distributions.
- Elegant and mathematically grounded
- Computationally efficient
- Easy to implement using existing tools
- Breaks down in dynamic, adaptive environments

Aug.2025> github.com/Trading-Goos...
Sep.2025> github.com/ivebotunac/P...
Oct.2025> github.com/keithtyser/a...
Oct.2025> www.arxiv.org/pdf/2510.15949
Oct.2025> chatpaper.com/paper/199062
Nov.2025> arxiv.org/pdf/2511.12599
Nov.2025> github.com/hkirat/ai-tr...
Nov.2025> nof1.ai/leaderboard

Sources:
Dec.2023> arxiv.org/pdf/2311.13743
Jun.2024> arxiv.org/pdf/2402.18485
Jan.2025> github.com/moondevonyt/...
Feb.2025> github.com/virattt/ai-h...
Jun.2025> github.com/TauricResear...
Jun.2025> github.com/51bitquant/a...
Jul.2025> github.com/wanghsinche/...

Source: github.com/sarwarbeing-...

CivMAS Pillars >> linkedin.com/pulse/four-f...
Review paper >> arxiv.org/abs/2507.21046
AgentEvolver >> arxiv.org/abs/2511.10395