LLMs free skilled staff from routine steps and make information directly accessible, cutting task-level inefficiencies rather than reengineering whole processes; but benefits depend on process fit and organizational readiness. The technology often expands capabilities—enabling work previously infeasible—so selection and readiness, not only experimentation, determine whether firms capture value.

Main Finding

LLM-based AI in administrative and knowledge work chiefly reduces two Lean wastes — skills underutilization and motion — by freeing qualified staff from routine execution and making information directly accessible. These gains occur mainly within individual process steps (not between processes). Organizational readiness acts as a boundary condition: when readiness is low, adoption yields little value; when readiness is sufficient, LLMs can both improve efficiency and expand capabilities (enable previously infeasible work). The study extends Task‑Technology Fit (TTF) logic from the individual task level to the process level and delivers a practical selection framework to guide organizational adoption.

Key Points

• Empirical basis: twelve semi‑structured interviews across twelve organizations in the Swedish energy and utilities sector (collaboration with Intric AB).
• Theoretical framing: combined Lean waste analysis, Task‑Technology Fit theory, and organizational condition assessment.
• Primary Lean wastes addressed by LLMs:
  • Skills underutilization — routine work is offloaded so qualified staff can focus on higher‑value activities.
  • Motion — faster, more direct access to information reduces unnecessary navigation and search.
• LLM impact is concentrated inside process steps (step‑level efficiency), not primarily on cross‑process coordination.
• Process‑level TTF: four process‑fit dimensions each affect distinct aspects of implementation (study synthesizes how fit maps to realized value).
• Organizational conditions: four assessed conditions explain variation in outcomes across organizations; readiness functions as a boundary condition for value realization.
• Capability expansion: a distinct outcome where LLMs enable new activities or services that were previously infeasible, not just faster execution of existing tasks.
• Practical output: a selection guidance framework that gives organizations a structured way to choose where to apply LLMs instead of relying on ad hoc experimentation.

Data & Methods

• Data: 12 semi‑structured interviews in the Swedish energy & utilities sector; interviewees from 12 different organizations (collaboration with Intric AB).
• Method: qualitative analysis synthesizing interview data through the lenses of Lean waste analysis, Task‑Technology Fit theory (extended to processes), and organizational condition assessment.
• Level of inference: exploratory, sector‑specific qualitative findings — useful for theory building and practical guidance but limited in statistical generalizability.

Implications for AI Economics

• Productivity measurement: benefits of LLMs may show up as within‑step efficiency gains and skill reallocation rather than obvious throughput or cycle‑time reductions; metrics should capture both time saved and higher‑value redeployment of staff.
• Value capture & ROI: organizational readiness materially conditions returns — investments in tooling, governance, and complementary skills may be necessary before LLM deployment yields economic value.
• Strategic adoption: firms should prioritize processes with high skill underutilization and high information‑search (motion) burdens for early LLM rollout; use the selection framework to reduce costly trial‑and‑error.
• Labor reallocation & wages: LLMs can shift worker effort from routine execution to higher‑value tasks; this may increase demand for oversight, orchestration, and domain expertise rather than replace skilled roles entirely.
• Capability creation & market effects: LLM‑enabled capability expansion can generate new product or service opportunities, altering competitive dynamics and the scope of firms’ offerings.
• Policy & training: regulators and firms should anticipate complementary investments in training and governance to realize social and firm‑level gains from LLM adoption.
• Research implications: calls for quantitative follow‑ups across sectors to measure magnitudes of within‑step gains, conditions for capability expansion, and macroeconomic impacts of process‑level TTF.