We research how agentic AI becomes a trustworthy actor in an organization’s ecosystem.
Softforge Labs is an applied AI research studio. We study how a bounded, accountable AI entity should behave toward the people and systems an organization depends on — its own teams, its clients, its suppliers, its partners, its customers, and, increasingly, other companies’ AI systems. Professional Services is how that research reaches real organizations.
Core, incoming, and associated researchers and engineers
Ecosystem surfaces every cognitive entity we build is designed against
Contextual intelligence and agentic systems as the studio’s research focus
Research structure aligned with France’s JEI / CIR innovation framework
An organization’s cognitive entity, not another chatbot
Most agentic AI is built to answer employees faster. Our research asks a harder question: what does it take for an AI actor to represent an organization responsibly — carrying memory, values, and accountability — across every relationship that organization has, including its relationships with other AI systems? We call the object of that research a cognitive entity: bounded, auditable, and designed surface by surface rather than bolted on as a feature.
Internal
Team coordination, collective memory, and structured observation inside the organization.
Clients
Service delivery, status, and co-creation with the parties who commission the work.
Suppliers
Bounded, human-supervised coordination on sourcing, monitoring, and exceptions.
Alliance & channel partners
Shared context and joint planning across organizational boundaries.
Customers
Expert-guided companion behavior for the people who use a deployed product.
Other AI systems
Machine-to-machine coordination with a counterparty’s own AI, on open protocols, under explicit trust boundaries.
Professional Services is the applied edge of our research
Every engagement we take on installs a working instance of our research into one or more of an organization’s ecosystem surfaces — never the reverse. We qualify opportunities the same disciplined way we qualify a research question: a named, funded problem; a real decision owner; evidence that the work will hold up to scrutiny.
Unisia: our flagship research platform
Unisia is a behaviorally adaptive, multi-agent human–AI coordination system. Its core research premise: interaction is itself a measurable data stream — conversation, timing, tasks, decisions, and outcomes treated as telemetry and interpreted over time, not discarded after each exchange. Expert Companion Modeling, a Behavioral Operating System, contextual memory, and multi-agent orchestration form a closed loop that decides when, how, and whether to intervene — not just how to answer.
See our projectsHow we work
Human-centered intelligence
We design AI to augment judgment and preserve human agency, not to replace either.
Research discipline
Our work is explainable, traceable, and auditable — built to hold up under scrutiny, not just under demos.
Engineering ownership
We ship what we research and own it in production — the 3 AM incident, not just the slide.
Speed with evidence
We move quickly when evidence supports it and stop when it doesn’t — no improvisation dressed as progress.