How do AI novelty and scope shape employee relationships?

Makarius, Mukherjee, Fox and Fox (Journal of Business Research, 2020) propose a "Model of AI Integration" that crosses two dimensions of an AI system: its novelty ("low/moderate/high") and its scope, split into "content changing/incremental AI versus context changing/radical AI." Crossing these produces six quadrants, each entailing "the types of employee-AI relationships and human roles in working with different types of AI." They define "content changing" as change confined to "a narrower task domain, mainly based on the component knowledge of the associated system," and "context changing" as change that "may occur in broader task domains (and sometimes across the value chain creating new ecosystems) ... based on the architectural knowledge of the system." Each quadrant, they argue, produces a different amount of what they call "sociotechnical capital."

The mechanism is socio-technical systems (STS) theory, which they say "describe[s] and explain[s] the behavior of organizations and their members while providing critical insights into the relationships among people, technology, and outcomes." They extend Resnick's (2001) construct of "sociotechnical capital" — originally "productive combinations of social relations and information and communication technology" in civic and community settings — into the workplace, defining it as the organization's "ability to jointly optimize the use of AI and people." Following Barney (2001), they treat it as a form of capital that is "valuable, rare, difficult to imitate, and organization-specific," accumulated "over time with intentional investment," and they frame the whole process through an "organizational socialization approach" — integrating AI is modeled on how organizations socialize a new employee, not how they install a new tool.

The typology sits alongside two other pattern notes in this library that classify AI-organization encounters on different axes. How should humans and AI systems divide decision-making tasks? classifies decision tasks by where accuracy advantage lies and by learning configuration; Makarius and colleagues instead classify the AI system itself by novelty and breadth of effect, and derive the relationship type from the technology's characteristics rather than from comparative task performance. What problems do companies solve with generative AI? sorts generative AI deployments by the organizational problem they were observed solving; Makarius's scope axis is meant to characterize a system ex ante, before deployment, rather than classify what already happened. The "context changing/radical AI" quadrant, crossing the value chain and "creating new ecosystems," is also one concrete articulation of the system-wide case that Does AI adoption require rethinking how organizations coordinate? treats formally as costly to coordinate but synergy-rich.

The excerpt is a conceptual framework paper proposing a "future research agenda," not a study that scores any organization against the six quadrants or measures sociotechnical capital directly; the claims that managers and employees hold negative perceptions of AI integration are themselves citations to other studies (Frey & Osborne, 2017; Raisch & Krakowski, 2020; Gunning, 2017), not this paper's own data. It was published in 2020 against pre-2020 examples — recommender systems, hiring algorithms, blockchain smart contracts, hospital forecasting tools — not LLM agents, so its novelty and scope categories describe enterprise AI of that era; extending the quadrants to agentic AI would be a reader's inference, not a claim the paper makes.