Building the Foundations for Trusted AI Agency
Our research studies and develops the formal, technical and organisational foundations required to delegate consequential work to AI without surrendering control or accountability.
Governance Inside the Agent Loop
Exact Nonblocking Control under Partial Observation and Dynamic Authority
agent governance · runtime assurance · supervisory control · formal methods · partial observation · dynamic authority
Design, Potential and Realization in Agentic Systems
A framework for agentic systems that connects configuration to possible and realized behaviors.
agentic systems · system design · potential · realization · authority · reliability
Enterprise AI Adoption as an Organisational State Transition
A state-transition theory of organisational AI adoption
enterprise AI adoption · organisational state · delegation · hysteresis · complementarities · state transition
The Authorisation Bottleneck
Maintaining delegated authority under shared review capacity
authorisation · delegation · policy renewal · queueing control · human review · enterprise AI
Building the Decision-Native Enterprise
An operating model for sustained, authorised action
enterprise AI · decision systems · standing authority · policy renewal · operating model · human oversight
The Supervisory Evidence Gap
A shared evidence record linking source information, access permissions, unresolved questions and supervisory decisions.
continuous supervision · supervisory technology · regulatory reporting · evidence integration · data governance · cross-organisational information sharing
Agent governance and control
Permitting only the actions that keep an acceptable outcome achievable, based on available information and the agent’s authority at its current runtime state.
Proof-carrying authority
Certificates that prove a governing decision was authorised and carried out exactly as intended, with a governance memory that preserves that certified decision over time.
Design and reliability
Separating what an agentic system may do from what it actually does, distinguishing potential from realisation and applying minimum necessary permissions based on cost, latency, exposure and risk.
Organisational adoption
Adoption as a state rather than a score: sustained delegation of consequential work, the evidence loop that holds it in place, and the thresholds that govern entry and exit.