Last October, I published a blog proposing a different mental model for AI agents: treat them as cognitive skills and products, not as digital employees. That framing has since resonated strongly with Forrester clients, particularly technology leaders building agentic capabilities inside the enterprise. But the concept of a cognitive skill in that blog was deliberately loose. It was intended as an intuitive corrective to anthropomorphic thinking about AI agents, not as a rigorous architectural construct.
Over the past several months, Betsy Summers, Sam Higgins and I have developed that idea much further. In our new Forrester report titled ‘Make Skills The Foundation Of Your Agentic AI Journey’, we define the cognitive skill as the atomic unit of capability, show how skills compose into roles and workflows, and make explicit a claim that the earlier blog only hinted at: human skills and agent skills ultimately describe the same cognitive work.
That has a surprisingly profound implication. A cognitive skill is executor-agnostic: it can be defined independently of whether it is performed by a person, an AI agent, or a human-agent team. Once work is decomposed into workflows, roles and skills, the executor becomes an implementation choice rather than the defining characteristic of the work itself. We call this approach the Skills-Oriented Agentic Architecture (SOAA).
This has massive implications for enterprises on an AI Voyage.
HR curates skills taxonomies, competency frameworks, and job architectures to describe human capability. IT builds agent catalogs, capability roadmaps, and governance models to describe agentic capability. Both functions model the same cognitive work using different vocabularies, different systems, and different governance mechanisms. That separation made perfect sense when software executed deterministic logic and humans supplied all the reasoning, judgment, and decision making. It makes far less sense when agentic systems can reason, plan, adapt, and pursue goals. Once leaders accept that human and agentic capabilities increasingly share the same unit of design, several important consequences follow:
- Workforce planning and AI investment converge onto a common capability surface. Leaders can decide whether to build a capability internally, hire for it, automate it, or compose a human-agent team around it, all against the same inventory of cognitive skills.
- Work becomes dynamically composable. An orchestration layer can assign skills to humans, agents, or hybrid teams at runtime based on proficiency, availability, cost, governance, or policy.
- Governance itself begins to converge. Rather than maintaining separate governance mechanisms for people and AI, organizations can increasingly manage both through a common skills registry, shared behavioral contracts, and consistent policy surfaces.
This convergence forms one of the foundational ideas behind a broader research theme that we are currently developing at Forrester, called the Cognitive Operating Model.
Our objective is to define a unified operating framework for designing, allocating, governing, and evolving cognitive capability across the full human-agent-physical estate. In doing so, we hope to connect skills-based portfolio design with enterprise architecture, workforce strategy, organizational design, and governance into a coherent operating discipline.
We’re only at the beginning of what promises to be a fascinating evolution for how enterprises will organize work in the cognitive era. Join us at Forrester’s AI Forum in Singapore (August 20th) and Sydney (August 25th), where we’ll explore these ideas in depth – and later this year, we’ll present the Cognitive Operating Model at the Forrester Technology & Innovation Forums in Austin (September 14–15), London (September 30–October 1), and New York (November 4–5). At Forrester, we have a lot of thoughts on how to do agentic AI right, and we love to discuss these ideas with you. Connect with Betsy, Sam or me by scheduling time for inquiry or guidance.
