Current Work
Building the Control, Memory, and Operating Systems for AI-Native Organizations
Michael’s current work focuses on moving AI from isolated tools and experiments into governed systems that can act, coordinate, retain context, and operate safely in consequential environments.
This work is carried out through Michael’s current leadership at Plurilock, through R2 Advisory, which he founded, and through teaching and research at Virginia Commonwealth University — with the published Ideas on this site documenting the thinking as it develops. Not all of it is owned or commercialized through a single entity.
Central Thesis
The AI Challenge Is No Longer Just Intelligence
Model capability continues to advance quickly, and that progress matters. But model capability is now advancing faster than most organizations’ ability to assign authority, coordinate agents, manage memory and context, govern delegation, verify action, maintain accountability, and integrate AI into how the organization actually operates.
Michael’s current work focuses on that coordination and control layer — the systems that let an organization delegate real work to AI safely, repeatedly, and accountably, rather than the models themselves.
Workstreams
Four Areas of Current Work
Workstream 01
Governed Agentic Systems
AI systems increasingly receive access to tools, data, workflows, and decision authority. A prompt describes intent, but it does not define what an agent may do, on whose behalf, or what happens when it reaches the edge of its authority. Governing that requires bounded authority, delegation, approval, pause-and-resume control, traceability, fail-closed operation, and human oversight designed into the system rather than assumed.
The Agent Delegation Contract
The Agent Delegation Contract is an evolving specification Michael is developing for expressing authority, delegation, constraints, evidence, and accountability in agentic systems. It is a working draft, not an adopted industry standard — a way of stating explicitly what an agent is permitted to do so that permission can be checked and enforced rather than assumed.
View the Agent Delegation Contract specification →From prompts to contracts
The practical shift this work argues for: enterprises putting agents into production need a record of what an agent was authorized to do that survives the moment the agent acted — not just a log of what it attempted.
Workstream 02
AI-Native Enterprise and Knowledge Work
Files, folders, projects, applications, and chat sessions are inherited from an earlier computing model built around a single human user working alone. Knowledge work increasingly runs through persistent project or mission agents that need durable context, semantic and episodic memory, orchestration across models, and work products both humans and machines can read — coordinated rather than scattered across disconnected tools.
iDragonFly and applied prototypes
R2 Advisory is using iDragonFly and related prototypes as an applied environment for exploring AI-native knowledge work, coordination, memory, and governed execution. These are applied research prototypes, not a mature commercial product — a working environment for testing how coordination and memory should actually function before those patterns get generalized.
Coordination as the bottleneck
As model capability increases, the constraint on expert work shifts from generating output to coordinating who — human or machine — is doing what, with what authority, informed by what memory.
Workstream 03
Secure Mission Systems
Secure AI does not stand apart from cybersecurity, identity, and data architecture — it depends on them. In defense, intelligence, critical-infrastructure, and healthcare environments, AI-enabled systems have to operate inside existing disciplines for identity, cross-domain operations, DevSecOps and MLOps, human authorization, and operational resilience, at a level of rigor high-consequence environments already require.
Identity and evidence for non-human actors
As agents act with delegated authority, identity, provenance, and audit evidence can no longer be human-only or after-the-fact. The record that supports a machine decision has to be produced at the moment the decision is made.
Built on two decades of security and mission delivery
This work connects directly to Michael's experience leading OT cybersecurity at Honeywell, national cyber-analytics programs at Raytheon, and cyber/analytics portfolios at Booz Allen Hamilton — sectors where the cost of an ungoverned system is measured in real consequence, not inconvenience.
Workstream 04
AI-Native Operating Models and the Future of Consulting
AI changes the economics of expertise. The consulting pyramid — bundled labor, priced by the hour, scaled through headcount — was built around a production layer that is no longer strictly labor-dependent. As production compresses, expert judgment becomes a control layer and coordination and delegation become the core skill, and firms need new operating models, not just productivity tools bolted onto the old one.
The Diamond Model of Consulting
The Diamond Model is Michael's structural thesis on how AI restructures professional services — one application of the broader thesis behind this page, not the whole of it. It describes what replaces the pyramid once production is no longer the bottleneck.
How the Work Connects
A Layered Model, Not Four Separate Projects
The four workstreams above are not independent efforts. Each addresses one layer of a single stack, read from the underlying technology up to the organization that has to run it. A model that cannot be governed is a liability; an operating model with nothing underneath it is aspirational. The layers below need each other in order.
Models and Tools
The underlying AI capability — language models and the tools they can invoke.
Agents and Workflows
Goal-directed processes that use models and tools to accomplish multi-step work.
Memory and Knowledge
The durable context, structured knowledge, and semantic/episodic memory agents draw on across time.
Authority and Delegation
What an agent is permitted to do, on whose behalf, and within what boundary — the domain of the Agent Delegation Contract.
Coordination and Control
How multiple human and machine actors are orchestrated so delegated work stays coherent and accountable.
Human Judgment and Accountability
Where human review, approval, and accountability are mandatory rather than optional.
Enterprise and Mission Operating Model
The organizational structure — roles, incentives, and governance — that makes the layers below sustainable at scale.
Governed Agentic Systems sits primarily at layers 4–6 (authority, coordination, and accountability). AI-Native Enterprise and Knowledge Work sits primarily at layers 3–5 (memory, delegation, and coordination). Secure Mission Systems runs through every layer as a cross-cutting discipline. AI-Native Operating Models sits primarily at layer 7 — the organizational structure that makes the rest sustainable.
Current Artifacts
What Exists Today
The Agent Delegation Contract
Specification — In DevelopmentAn evolving specification for expressing authority, delegation, constraints, evidence, and accountability in agentic systems. A working draft, not an adopted standard.
View the Agent Delegation Contract specification →iDragonFly / IDF-Genesis
Applied Research PrototypeAn applied environment R2 Advisory uses to explore AI-native knowledge work, coordination, memory, and governed execution — not a commercial product.
The AI-Native Enterprise
PublishedA fourteen-essay season on trusted human-machine coordination at scale, spanning agentic systems, enterprise memory, trust architecture, and operating models.
Explore →The Diamond Model of Consulting
PublishedA structural thesis on how AI restructures professional services, from the bundled pyramid to a delegation-based operating model.
Explore →Research Program
Applied ResearchFive structured research themes — secure AI transformation, symbiotic intelligence, AI-native operating models, governance and trust architecture, and human-AI collaboration.
Explore →University Teaching
Teaching & Academic WorkOngoing teaching at Virginia Commonwealth University, connecting applied executive experience to the next generation of technology and business leaders.
Ways to Engage
Four Distinct Pathways
Executive and Enterprise Leadership
For organizations addressing AI, cyber, data, product, platform, or operating-model challenges at the executive level.
Discuss a Strategic Challenge →Strategic Advisory
For secure AI architecture, mission-system strategy, agentic governance, and technology strategy through R2 Advisory.
Visit R2 Advisory →Research and Product Collaboration
For the Agent Delegation Contract, AI-native knowledge work, coordination, memory, and control-plane architecture.
Start a Conversation →Speaking and Executive Briefings
For boards, leadership teams, conferences, and university audiences on secure AI, governance, and mission systems.
Explore Speaking →Connect
Discuss a Strategic Challenge
For governed agentic systems, secure AI architecture, AI-native operating models, and mission-system strategy — grounded in two decades of leadership across cybersecurity, national security, and enterprise technology.