AI & Agentivity
Artificial intelligence is reshaping how professional services firms work. PMG's position is straightforward: AI is a tool that makes expert human work faster, sharper, and more rigorous. It does not replace the judgment, accountability, or intellectual responsibility of our consultants.
We use AI across research, analysis, drafting, and ideation. Every output that leaves PMG reflects the judgment of a human consultant, regardless of what tools contributed to its development. That is not a caveat — it is the operating standard.
Our approach is governed by four principles
Business Integrity
AI use must strictly align with our professional standards and client commitments.
Client Trust
We handle data with care that exceeds what the law requires, protecting confidentiality relentlessly.
Transparency
We maintain a clear internal understanding of how every tool we use works and what its limitations are.
Human Oversight
AI does not operate autonomously, make decisions, or carry accountability in any engagement.
What we mean by Agentivity
Agentivity is PMG's term for the application of agentic AI — systems capable of executing multi-step tasks with a degree of autonomy — in a consulting context.
The distinction matters. Most AI use involves a simple prompt and response. Agentic AI can plan, take sequential actions, use tools, and produce outcomes across a chain of steps without a human directing each individual step. This creates genuine capability, and genuine risk.
At PMG, agentic AI is used only under active human supervision. An agentic system does not have authority over client decisions, recommendations, or deliverables. It operates as a capable assistant with a human in the loop at every consequential step.
What We Don't Do
Some boundaries are worth stating plainly and non-negotiably.
PMG does not build proprietary AI models. We evaluate, select, and use tools built by vetted providers, assessed rigorously against our risk framework.
We do not use client data to train or fine-tune any AI model, internal or external. Client data is never an input to model training under any circumstances.
We do not share client data with AI platforms without explicit, documented consent from the client. This applies regardless of convenience or speed.
We do not deploy AI autonomously in client engagements. AI does not make decisions, produce final recommendations, or act on behalf of a client without human oversight.
We do not use AI tools rated above a limited risk threshold without a client-specific risk assessment and mitigation plan agreed with the client prior to deployment.
Our Published Thinking
PMG publishes research and analysis on the strategic implications of AI — not just how to use it, but what it is doing to the conditions in which organizations operate.
AI's Second-Order Risks: Infrastructure, Talent, and Truth Are Already Under Strain
Examines four consequences of AI adoption that are already manifesting but are underweighted in most strategic conversations. These are not future problems — they are present ones.
Read Full ArticleModel Risk Tier Calculator
Select your engagement scoping parameters across all 4 operational dimensions to determine the appropriate PMG AI Governance Tier and generate your real-time compliance checklist.
Deployment Mode
Operational scopeData Classification
Confidentiality levelDecision Autonomy
Human-in-the-loopArchitecture Tier
Hosting & computeTier 2 — Low to Moderate Risk
Moderate operational impact. Requires explicit client DPA alignment, prompt sanitization, and structured peer review.
Mandatory PMG Governance Checklist
0 of 5 CompletedFrameworks & Documentation
The following materials are available to clients, researchers, and auditors. Access requires a verified PMG account.
AI Platform Capacity and UAP Analysis
A reference guide to the major AI platforms PMG has assessed: what each one does, what makes it distinctive, and how it fits within our evaluation framework.
Approved AI Platforms and Risk Assessment
PMG's current approved platforms list, with risk ratings across five dimensions and plain-language guidance for staff on how each tool may be used.
AI Risk Assessment Methodology
The framework PMG uses to evaluate AI platforms, including the five risk dimensions, the four-tier rating scale, and how jurisdiction risk is assessed.
AI & Data Ethics Statement
PMG's formal statement of ethical AI principles, data handling commitments, governance structure, and bias mitigation program.