Artificial Agency
Generative behavior engine for lifelike NPCs
About
Artificial Agency, founded in Edmonton by former Google DeepMind researchers, builds a behavior engine that replaces fixed dialogue trees and decision trees with runtime decision-making: developers give each NPC motivations, rules, and goals, and characters improvise their behavior in response to the player. The engine plugs into existing games or underpins new ones, and the company says it can also drive companion agents, overseer agents, and adaptive tutorials. It came out of stealth with $16M in 2024 and launched the behavior engine with a Cubico tech demo in 2025. Investor and funding claims are company-reported, and real-world studio deployments are still early.
What it can do
- Generative NPC behavior
- Motivation- and goal-driven agents
- Runtime improvisation
- Companion and overseer agents
- Adaptive tutorial generation
- Integration with existing games
Setup
- Contact Artificial Agency for engine access
- Define NPC motivations, rules, and goals in the authoring environment
- Integrate the engine with your game build
- Tune behaviors in playtests and ship
Prerequisites
Game studio account; supported engine integration
Use cases
Works with
Trust score — 8/10
Scored 8/10 (verified tier, metrics-v1). Evidence: TechCrunch July 2024: $16M out of stealth; GlobeNewswire July 2025: behavior engine launch with Cubico demo; Funding and customer traction claims are company-reported; Real studio deployments still early. License: Proprietary engine; enterprise licensing.
How the trust score is built
Full methodology →- Adoption · stars/downloads, log-scaled vs peers
- 7.0/10
- Maintenance · recency, releases, issue load
- 8.0/10
- Openness · license + docs
- 6.0/10
- Manual review · sanity read + red-flag check
- 8.0/10
- · TechCrunch July 2024: $16M out of stealth; GlobeNewswire July 2025: behavior engine launch with Cubico demo
- · Funding and customer traction claims are company-reported
- · Real studio deployments still early
Sources
Every fact above was checked against these sources. Where a claim comes from the provider and couldn't be independently verified, the text says so.