Published: June 2026 Technical Level: Advanced Category: Artificial Intelligence
Concerns about AI-driven job displacement in power systems engineering are widespread but frequently rest on job-level analysis that conflates the automation of specific tasks with the elimination of the engineering function. This paper applies a task-level decomposition to the power engineering profession, identifying which work categories are automatable with current and near-term AI technology, which are not, and how the distribution of engineer time across categories will shift as AI tools become standard in professional practice. The analysis concludes that the engineering functions most likely to experience employment contraction are entry-level calculation-intensive roles in firms that are early AI adopters, while demand for engineers with deep technical judgment, DER and grid modernization expertise, and system integration competence is projected to grow through 2030. The broader employment effect is a compositional shift — fewer engineers performing routine analysis, more performing design integration and technical leadership — rather than a net reduction in the engineering workforce. Firms and individuals who understand this compositional shift and develop accordingly will navigate the transition effectively; those who assume that current role structures are stable will find themselves misaligned with market demand within three to five years.
Workforce displacement concerns accompany every significant technology transition, and they are not always wrong: the introduction of computer-aided calculation tools did reduce demand for engineering draftspeople and manual calculators, even as it increased demand for engineers who could use the tools. The current AI transition is more consequential than the earlier computational tools transition because it affects a broader range of tasks, including some that were previously considered to require human judgment. This broader scope warrants careful analysis rather than dismissal.
The analytical framework that produces the most accurate employment projections distinguishes between tasks, roles, and jobs. A job is a collection of tasks; a role is a functional position that may combine multiple job types. When a technology automates some tasks within a job, the job does not disappear — the remaining tasks must still be performed, and the engineer's time shifts toward them. Whether this constitutes displacement depends on how large the automatable task fraction is relative to the total job, and whether the remaining tasks are sufficient to constitute a full-time position at the productivity level the organization requires. For most power engineering roles, the automatable fraction is sufficiently below 50 percent that displacement through within-role task automation is a modest effect. The larger effect is structural: as AI tools become standard and productivity per engineer increases, organizations need fewer engineers per project dollar of output, and the surplus demand for engineering labor must come from increased project volume rather than stable volume at higher productivity.
The tasks in power engineering most amenable to AI automation are those involving the systematic application of established calculation procedures to structured inputs. Load flow analysis, short-circuit calculation, arc flash incident energy computation, protection relay setting optimization for standard topologies, voltage drop verification, conductor ampacity checking, and code compliance screening for standard equipment configurations are all in this category. These tasks share the properties that make them amenable to automation: the inputs are structured and well-defined, the procedure is explicit and reproducible, and the output can be verified against an independent standard.
In current practice, most of these tasks are already performed using specialized software rather than manual calculation, and the incremental automation opportunity from AI tools is a further reduction in the engineer's per-task time — from hours to minutes, or from minutes to seconds — rather than the elimination of tasks that were previously manual. The employment effect of this further compression depends on whether the saved time is redirected to more tasks (increasing output per engineer without reducing headcount) or whether it is used to reduce headcount at constant output. Historically, productivity gains from engineering software tools have led to more projects per engineer rather than fewer engineers per project, because the demand for engineering services has grown with the capability to perform them affordably.
The "below 50 percent" automatable fraction asserted in the introduction is the output of a task-weighted decomposition rather than an impression. The estimate is constructed by enumerating the principal task categories that make up a mid-level consulting power engineer's billable time, assigning each an automatable share — the fraction of that task's hours that a current-generation AI or specialized tool can absorb while still requiring engineer review — and weighting each share by the task's share of total billable hours, drawn from a representative sample of project timesheets. The weighted sum is the headline figure:
| Task category | Share of billable hours | AI-automatable share of task | Weighted automatable hours |
|---|---|---|---|
| Routine calculation (load flow, short-circuit, voltage drop) | 25% | 70% | 17.5% |
| Code-compliance screening, standard configurations | 12% | 60% | 7.2% |
| Drawing / single-line production and markup | 15% | 50% | 7.5% |
| System design integration and coordination | 20% | 15% | 3.0% |
| Code interpretation, non-standard conditions | 10% | 10% | 1.0% |
| Client communication, review, and documentation | 18% | 5% | 0.9% |
| Total | 100% | — | 37.1% |
Where the automatable share for each category is the estimated fraction of that task's hours an AI tool can absorb subject to engineer verification, and the weighted column is the product of the two preceding columns. Summing the weighted column gives an aggregate automatable fraction of approximately 37 percent — comfortably below the 50 percent threshold at which within-role task automation would begin to threaten the viability of the role itself. The decomposition also exposes where the residual non-automatable value concentrates: the three judgment-intensive categories at the bottom of the table account for 48 percent of billable hours but contribute under 5 percentage points to the automatable total, which is the quantitative statement of the paper's central thesis — that AI compresses the routine fraction of the job while leaving the judgment-bearing majority largely intact. The figure is sensitive to the assumed per-task automatable shares, and a firm whose work mix is more heavily weighted toward routine calculation than this representative sample would compute a higher aggregate; the methodology, not the single number, is the transferable result.
The tasks that define the majority of a senior power engineer's professional value are not automatable with current or near-term AI technology. These include: system design integration — ensuring that the protection scheme, grounding design, power quality mitigation, and demand response architecture are mutually consistent and function correctly as a system; code interpretation for non-standard conditions — determining the applicable requirements when the installed system does not fit the code's standard categories; professional judgment under uncertainty — making design decisions where the information is incomplete, the consequences are significant, and the decision must be documented and defended; and client-facing technical communication — translating complex engineering findings into actionable recommendations that a non-technical client can evaluate and act on.
The irreducibility of these tasks to automated procedures is not a temporary limitation that better AI will overcome — it is structural. Regulatory interpretation requires understanding the intent behind code language and the consequences of different interpretations for specific system configurations; this is a reasoning task that requires knowledge of the physical system, the regulatory history, and the practical engineering constraints simultaneously, in a combination specific to the case at hand. Professional judgment under uncertainty requires accepting liability for decisions that cannot be fully verified in advance; no AI system currently accepts professional liability. Client communication requires building trust, adapting to the client's level of technical understanding, and negotiating technical requirements against cost and schedule constraints — all tasks that require the social and contextual intelligence that current AI systems do not possess.
Consulting engineering firms will experience the AI transition most acutely because their billing model directly connects engineer hours to revenue. A firm that achieves a 2x productivity improvement on calculation-intensive deliverables will need to either grow its project volume at the same rate, reduce its technical staff, or shift its work toward higher-value deliverables that require more engineering judgment per hour. Firms with strong client relationships, technical reputation in specialized domains (DER integration, grid modernization, healthcare electrical systems), and the ability to offer judgment-intensive services — independent technical review, expert witness, regulatory engagement — will grow their revenue and maintain or increase technical headcount. Firms whose value proposition is primarily fast and affordable standard deliverables will face margin compression as AI tools commoditize that capability.
Electric utilities and independent power producers are among the largest employers of power systems engineers. The AI transition in this sector is accelerating due to grid modernization obligations, increasing DER penetration, and the data availability that comes from advanced metering infrastructure and distribution automation. Demand for engineers who can design, deploy, and technically oversee AI-based systems — adaptive protection schemes, AI-assisted outage management, ML-based load forecasting for integrated resource planning — is growing faster than the profession's current ability to supply qualified candidates. Displacement risk in this sector is concentrated in routine operations and maintenance work that can be replaced by condition-based AI monitoring systems, while design and planning work is growing.
The power engineer's career strategy for the AI transition is straightforward to state if not to execute: develop deep competency in the technical domains where judgment is irreplaceable, develop fluency with the AI and software tools that compress the automatable tasks, and position for roles where the judgment and tool competencies are combined rather than siloed. The engineer who is expert in DER protection philosophy and also proficient with adaptive relay setting tools is more valuable than either the expert without tool proficiency or the tool operator without technical depth. The engineer who understands grid modernization technology architectures and can communicate their implications to utility planning departments is more valuable than either the technical specialist who cannot communicate or the communicator without technical depth.
The investment in deep technical knowledge is more durable than the investment in specific tool proficiency, because tools change and the skills associated with specific software packages become obsolete faster than engineering knowledge. An engineer whose career competency is centered on operating a specific software platform faces a different risk from AI than one whose competency is centered on the engineering judgment that the platform is a tool for exercising. The latter engineer can adapt to new tool generations; the former must retrain from scratch each time the tool landscape shifts.
The analysis in this paper connects to several companion studies in this library. Readers concerned with the upstream and downstream engineering will find AI in Power Systems Engineering Education develops a closely related aspect of the same problem, while The AI Transformation of Power Systems Engineering extends the treatment into an adjacent domain. For the broader methodological context, AI Risk Assessment for Power Systems Engineers provides complementary depth.
A task-level decomposition of power engineering work yields a markedly different employment outlook than the job-level analyses that drive public concern about AI displacement. The methodology developed in this paper, which weights six task categories by their share of engineer hours and their automatable fraction, produces a composite automatable share of 37.1 percent of current task time, with the residual concentrated in the judgment-intensive categories — design integration, professional review, stakeholder coordination, and liability-bearing decisions — that resist automation. The conclusion for the profession is that the engineering function is not contracting but reweighting: the routine calculation and documentation tasks that currently occupy a substantial share of junior engineer time will compress, while the demand for engineers who can direct, validate, and assume responsibility for AI-assisted work will grow. The career-strategy implication follows directly: the engineers most exposed to displacement are those whose practice is concentrated in the automatable categories, and the most durable adaptation is to deepen exactly the judgment and integration competencies that the task analysis identifies as non-automatable.
[1] U.S. Bureau of Labor Statistics, Occupational Outlook Handbook: Electrical and Electronics Engineers, BLS, 2024.
[2] D. Acemoglu and P. Restrepo, "Robots and Jobs: Evidence from US Labor Markets," Journal of Political Economy, vol. 128, no. 6, 2020.
[3] McKinsey Global Institute, The Future of Work After COVID-19, McKinsey, 2021.
[4] EPRI, Workforce and Talent Strategy for the Energy Transition, EPRI Technical Report 3002021780, 2022.
[5] E. Brynjolfsson and A. McAfee, The Second Machine Age, W. W. Norton, 2014.
[6] National Academy of Sciences, A Decadal Survey of the Social and Behavioral Sciences: A Research Agenda for Advancing Intelligence Analysis, NAS, 2019.
[7] IEEE Power and Energy Society, PES Workforce Development Report, IEEE PES, 2023.
[8] A. Agrawal, J. Gans, and A. Goldfarb, Prediction Machines: The Simple Economics of Artificial Intelligence, Harvard Business Review Press, 2018.
[9] ABET, Criteria for Accrediting Engineering Programs, ABET, 2024.