Workflow-integrated foundation models turn meetings and documents into durable, queryable 'artifact capital', making AI-enabled remote-first arrangements the default for much knowledge work. Offices instead become a resource for apprenticeship, conflict repair and other high-tacitness activities, while new risks—surveillance, skill atrophy and carbon costs—demand targeted ML and policy research.
Remote-Capable Knowledge Work Should Default to AI-Enabled Flexibility
The paper argues that integrated foundation-model stacks create durable, queryable artifact capital that shifts coordination economics in favor of AI-enabled remote flexibility, reserving in-person time for highly tacit, tightly coupled, or high-relational tasks.
This position paper argues that remote-capable knowledge work should default to AI-enabled flexibility because the workflow-integrated foundation-model stack changes the coordination economics that once favored daily co-presence. By foundation-model stack, we mean systems that combine natural-language interaction, multimodal capture, long context, retrieval, transcription, translation, and increasingly bounded tool use inside everyday workflows. Their organizational significance is not generic automation but the accumulation of artifact capital: durable, queryable, reusable traces such as transcripts, summaries, decisions, tickets, code comments, and retrieval layers. The argument rests primarily on capabilities that are already widely deployed---transcription, summarization, retrieval, translation, drafting, and code assistance---with bounded agents treated as an amplifying but not necessary extension. Rather than eliminating the office, this shift supports selective co-presence, reserving in-person time for tasks with high tacitness, high coupling, or high relational stakes, including apprenticeship, conflict repair, trust formation, and early-stage synthesis. Because the same systems can also intensify surveillance, skill atrophy, and compute-related emissions, we outline a machine-learning research agenda centered on team-level evaluation, privacy-preserving memory layers, scaffolded AI for learning, carbon-aware routing, and pro-agency workflow design.
Summary
Main Finding
Workflow-integrated foundation-model stacks (transcription, summarization, retrieval, translation, drafting, code assistance, and related long-context, multimodal, and bounded-tool capabilities) change the coordination economics of knowledge work by generating durable, queryable “artifact capital.” This reduces the advantage of daily co-presence for many remote-capable tasks and supports a new default of AI-enabled flexibility, while reserving in-person time for a narrower set of high-tacitness, high-coupling, or high-relational-stakes activities.
Key Points
- Foundation-model stack ≠ generic automation: its organizational value primarily comes from accumulating artifact capital (transcripts, summaries, decisions, tickets, code comments, retrieval indices) that lowers coordination costs over time.
- Already-deployed capabilities (transcription, summarization, retrieval, translation, drafting, code assistance) are sufficient to change coordination economics; bounded agents amplify effects but are not required.
- Outcome: many knowledge tasks become easier to decentralize and schedule asynchronously, favoring AI-enabled flexible/remote work as a default.
- Co-presence becomes selective: in-person interactions are most valuable for apprenticeship/mentoring, conflict repair, trust/relationship formation, and early-stage synthesis where tacit knowledge, tight coupling, or high relational stakes matter.
- Potential harms and trade-offs include intensified employee surveillance, skill atrophy (deskilling), and increased compute-related carbon emissions.
- Proposed ML research and design agenda: team-level evaluation metrics, privacy-preserving memory layers, scaffolded AI tools to support learning and apprenticeship, carbon-aware routing and compute allocation, and workflow designs that preserve worker agency.
Data & Methods
- Type of paper: position/theoretical paper (conceptual argument), grounded in observation of widely deployed capabilities rather than new large-scale empirical experiments.
- Empirical basis: references to mature, commonly used tools/functionalities (transcription, summarization, retrieval, translation, drafting, code assistance) and their observed uses in workplace workflows.
- Analytical method: economic and organizational reasoning about coordination costs and the role of durable artifacts (artifact capital) in substituting for co-presence; scenario analysis of task attributes (tacitness, coupling, relational stakes).
- Limitations: not an empirical causal study—claims are theoretical and suggestive; effectiveness and distributional impacts depend on implementation details, adoption patterns, and institutional responses.
Implications for AI Economics
- Labor supply & workplace design: lowers coordination frictions and increases feasibility of remote/async arrangements for many roles, affecting demand for office space, commuting, and urban labor markets; may shift bargaining over where and when work happens.
- Productivity measurement: artifact capital changes how productivity and knowledge production are observed and measured; firms may need new metrics that capture durable, reusable outputs and team-level coordination performance.
- Human capital & skill formation: potential substitution risks for tacit learning—necessitates design of scaffolded AI to support apprenticeship and prevent skill atrophy; workforce training and career progression models must adapt.
- Surveillance & governance: richer, queryable traces enable intensified monitoring; creates regulatory and governance needs (privacy, consent, limits on reuse) and potential effects on worker autonomy and bargaining power.
- Capital & emissions: higher compute and storage demands shift cost structures (capex/opex) and create carbon externalities; incentives for carbon-aware routing and energy-efficient model design become economically relevant.
- Firm boundaries & contracting: artifact capital may alter make-or-buy and employment vs. contracting decisions by changing asset specificity and the marginal cost of coordination across organizational boundaries.
- Research & public policy priorities: evaluate tools at the team and organizational level (not just individual task performance); fund privacy-preserving memory research; design standards for pro-agency workflows; create incentives or regulations to internalize compute-related environmental costs.
- Transitional dynamics: the office is not eliminated but repurposed; organizations will need policies to decide which activities require co-presence and how AI-enabled artifacts are governed and compensated.
Assessment
Paper Typecommentary
Evidence Strengthn/a — This is a position/theoretical paper that builds on observations of deployed capabilities and organizational reasoning rather than original empirical analysis or causal identification; no causal claims are tested with data.
Methods Rigorn/a — The paper uses conceptual argumentation and synthesis of existing technologies and organizational theory rather than formal empirical methods, experimental design, or econometric analysis.
SampleNo original dataset or sample; the paper synthesizes widely deployed AI capabilities (transcription, summarization, retrieval, translation, drafting, code assistance) and organizational examples to construct a normative argument about remote-capable knowledge work.
Themesorg_design human_ai_collab productivity adoption skills_training
GeneralizabilityApplies primarily to knowledge work with extensive digital artifacts; less relevant to manual or front-line service jobs, Effects may differ by firm size and resources (large tech firms vs small businesses), Cultural and institutional differences in workplace norms and labor regulations across countries may limit applicability, Depends on heterogeneous adoption of foundation-model technologies and digital infrastructure, Assumes responsible design and governance; outcomes change if surveillance or poor implementation dominate
Claims (10)
| Claim | Direction | Confidence | Outcome | Details |
|---|---|---|---|---|
| Remote-capable knowledge work should default to AI-enabled flexibility because the workflow-integrated foundation-model stack changes the coordination economics that once favored daily co-presence. Organizational Efficiency | positive | high | defaulting remote-capable knowledge work to AI-enabled flexible arrangements (i.e., work-location policy and coordination costs) |
0.01
|
| The foundation-model stack (NL interaction, multimodal capture, long context, retrieval, transcription, translation, bounded tool use) changes the coordination economics that previously favored daily in-person co-presence. Organizational Efficiency | positive | high | coordination economics (costs/benefits of co-presence vs. remote work) |
0.03
|
| The organizational significance of these systems is not generic automation but the accumulation of artifact capital: durable, queryable, reusable traces such as transcripts, summaries, decisions, tickets, code comments, and retrieval layers. Organizational Efficiency | positive | high | accumulation and reuse of organizational knowledge artifacts ('artifact capital') |
0.01
|
| Capabilities that are already widely deployed—transcription, summarization, retrieval, translation, drafting, and code assistance—are the basis for this shift (with bounded agents as an amplifying but not necessary extension). Adoption Rate | positive | high | deployment/adoption of specific AI capabilities (transcription, summarization, retrieval, translation, drafting, code assistance) |
0.06
|
| Bounded agents act as an amplifying but not necessary extension to the foundation-model stack for changing work coordination. Automation Exposure | mixed | high | role of bounded agents in amplifying coordination impacts |
0.01
|
| Rather than eliminating the office, this shift supports selective co-presence, reserving in-person time for tasks with high tacitness, high coupling, or high relational stakes (including apprenticeship, conflict repair, trust formation, and early-stage synthesis). Task Allocation | positive | high | allocation of in-person vs. remote time for specific task types |
0.03
|
| The same foundation-model systems can also intensify surveillance. Ai Safety And Ethics | negative | high | increase in workplace surveillance capability/use |
0.03
|
| These systems can cause skill atrophy. Skill Obsolescence | negative | high | degradation or atrophy of worker skills |
0.01
|
| Foundation-model usage can increase compute-related emissions. Ai Safety And Ethics | negative | high | compute-related (carbon) emissions associated with foundation-model usage |
0.03
|
| A machine-learning research agenda is needed centered on team-level evaluation, privacy-preserving memory layers, scaffolded AI for learning, carbon-aware routing, and pro-agency workflow design. Research Productivity | positive | high | prioritized ML research directions and interventions (team-level evaluation, privacy-preserving memory, scaffolded learning, carbon-aware routing, pro-agency design) |
0.01
|