PlanVector AI has announced the launch of PWM-1F, its first project-domain foundation model, designed to serve as a core intelligence layer for enterprise project agents, analytics platforms, and project-centric software environments.
PWM-1F is a Project World Model (PWM) and Temporal Causal Inference (TCI) engine, purpose-built to analyse how projects behave over time. The model is available immediately for integration by project management platforms, ERP and PSA vendors, and other enterprise systems that manage complex project portfolios.
Unlike conventional “AI for project management” approaches—which typically connect large language models to project documents and reports—PWM-1F is designed to answer control-level questions. These include where risk is accumulating, why margins are drifting, and how a project is likely to evolve based on its current trajectory.
A project-native intelligence layer
PWM-1F runs on top of leading general-purpose foundation models, but its behaviour is shaped through project-specific post-training. Rather than learning from live customer data, the model is trained on a broad range of synthetic project scenarios, expressed as time-series metric histories. This allows it to recognise patterns in project telemetry and translate those patterns into structured assessments of project health, risk drivers, and performance trends.
This capability underpins PlanVector’s Temporal Causal Inference (TCI) approach, which focuses on understanding how changes in project metrics over time influence outcomes, rather than simply summarising static information.
Because the model is trained on project metrics rather than a single platform’s data schema, PWM-1F can adapt across environments. When a host system provides recent project metrics, the engine interprets the trajectory, infers current health, and identifies the most significant positive and negative drivers. Platforms can then use these outputs to support early warnings, explanations, and “what-if” scenario discussions.
Where deeper analysis is required, PWM-1F can integrate with host platform APIs to pull in additional contextual data such as tasks, risks, or project phases.
First production deployment with PlanAutomate
PlanAutomate, an enterprise project management solution for project-based organisations using Microsoft Dynamics 365 Finance, is the first platform to deploy PWM-1F in production.
Within PlanAutomate’s system, the engine ingests metric snapshots from Dynamics 365 Finance and PlanAutomate’s project data model. It identifies projects whose metric trajectories resemble emerging risk patterns, explains variances in terms of underlying drivers, and feeds its assessments into PlanAutomate’s analytics and conversational interfaces.
Built for enterprise platforms
PWM-1F is built on Google Gemini and functions as a domain-specific, multi-modal project world model layered on a general foundation backbone. It is intended for use by software vendors and enterprise platforms that require project-aware AI capable of reasoning about project behaviour, not just project content.
PlanVector AI’s PWM-1F is available now for project management platforms, ERP and PSA vendors, and other enterprise systems that manage projects.
