By Jasmine Badiee Konrad, PhD and Jake Wiggins, PE
Most engineering organizations are experimenting with AI, but few have a clear framework for deciding where it should actually be used.
At the same time, power infrastructure demand is increasing, project timelines are tightening, and the engineering workforce is under pressure to keep pace. These forces are driving interest in tools that promise to accelerate engineering work. AI is quickly becoming part of that conversation.
The challenge is not whether AI can assist engineering work. It’s where it should.
The Problem with Ad Hoc AI Adoption
Across the industry, AI adoption is happening unevenly.
Some teams are experimenting aggressively, applying AI tools across a wide range of tasks. Others are hesitant, concerned about reliability, risk, and accountability. In many cases, decisions about where to use AI are being made without a consistent framework.
This creates two problems:
- Overuse, where AI is applied to tasks that require engineering judgment or carry significant risk
- Underuse, where AI could meaningfully improve efficiency but is avoided due to uncertainty
In engineering, both outcomes matter. The consequences of error, and the responsibility for those decisions, remain with the engineer.
Not All Engineering Tasks Are the Same
Engineering tasks vary significantly in how well they lend themselves to AI support.
Some tasks:
- have well-defined inputs
- follow established procedures
- produce consistent, verifiable outputs
Others:
- require interpretation
- depend on tacit knowledge
- produce multiple defensible solutions
These differences influence not just whether AI can be used, but how it should be used.
In practice, AI is often more effective as a tool for accelerating analysis and exploring options than for fully automating engineering work. In many cases, the effort required to verify AI-generated outputs, especially for high-risk tasks, can approach or exceed that of performing the work manually.
Risk Matters as Much as Feasibility
One of the most important considerations in applying AI to engineering workflows is risk.
Tasks that are technically well-structured may still carry high consequences if performed incorrectly. In these cases, even if AI can produce a plausible output, the level of verification required can significantly limit its practical value.
Conversely, lower-risk tasks may offer opportunities to use AI more freely to improve efficiency, provided outputs are appropriately reviewed.
Technical feasibility alone is not enough to determine whether AI should be used.
A More Deliberate Approach
To address this, NEI engineers developed a framework for evaluating where AI can effectively support engineering knowledge work.
Fig 1. Framework for evaluating where AI fits in engineering workflows.
The framework considers:
- how structured a task is
- how much output converges across engineers
- the consequence of error
Together, these factors help determine whether AI is best used to automate, accelerate, clarify, or support engineering work.
The goal is not to replace engineering judgment, but to apply AI in ways that enhance engineering workflows while maintaining technical accountability.
What This Means for Engineering Organizations
As AI tools continue to evolve, engineering organizations will need more deliberate approaches to integrating them into their workflows.
That includes:
- identifying where AI can provide real efficiency gains
- recognizing where engineering judgment must remain central
- aligning tool use with the level of risk associated with each task
Organizations that approach AI adoption in this way will be better positioned to improve productivity without compromising quality.
About the Research
NEI engineers Jasmine Badiee Konrad, PhD, and Jake Wiggins, PE, presented this work at the 2026 IEEE Rural Electric Power Conference (REPC), where it received a 2nd Place Paper Award.
To explore the full framework and methodology:
Download the paper:
Leveraging AI in Engineering Knowledge Work: A Framework for Evaluating Effective Use Cases
© 2026 IEEE. Personal use of this material is permitted. Permission from IEEE must be obtained for all other uses, including reprinting/republishing this material for advertising or promotional purposes, collecting new collected works for resale or redistribution to servers or lists, or reuse of any copyrighted component of this work in other works.