A Cost Case for Neurotechnology in Operations

Abstract. AI offers scale and powerful assistance in operations, but its operating costs, energy use, and effects on labour deserve scrutiny. The human brain runs on about twenty watts, making human intelligence remarkably energy efficient. Neurotechnology could help make better use of that capacity. Neuroscience informed interfaces may help operators find signals, sustain attention, and make decisions with less effort, sometimes reducing the need for AI. Neurotechnology could also help people use AI more selectively and collaborate with it more effectively. A credible cost case would measure operator time, decision quality, AI usage, incident outcomes, and total operating costs, alongside environmental and labour effects.

As a Solutions Architect in operational infrastructure, I analyze how AI can improve systems and tasks. Its ability to scale brings clear benefits, but operating costs vary with the model and the volume of work. At published standard API rates, one million requests containing 2,000 input tokens and 500 output tokens each would cost about US$225 with OpenAI’s GPT-6 Luna, US$4,500 with GPT-6 Sol, or US$22,500 with GPT-6 Astra. Those are model charges; integration and human review add costs of their own. AI’s growth also raises broader questions about energy demand and changes to labour (OpenAI, 2026; International Energy Agency, 2026; International Labour Organization, 2025).

As a neurotechnology specialist, I see another opportunity: make better use of human intelligence. The human brain uses roughly 10–20 watts, comparable to a small light bulb (Yu et al., 2018). That figure does not establish that human work is cheaper than AI. It does invite us to examine whether neurotechnology can help people make better decisions with the expertise and cognitive capacity they already have.

Neurotechnology

Neurotechnology refers to devices, software, and procedures that directly measure or interact with the nervous system to understand or influence its activity (UNESCO, 2025). I also use the term in a broader, applied sense: findings from neuroscience can inform the design of interfaces, training, and collaborative systems that require no neural sensor during everyday use. This broader use extends UNESCO’s definition. Table 1 summarizes the opportunities, potential benefits, and maturity of several approaches.

Opportunity	Contribution	Potential operational benefit	Maturity
Neuroscience-informed interfaces	Use neural measures during testing to examine attention, workload, and error recognition	Better decisions with less unnecessary cognitive effort	Research methods demonstrated; operational benefit needs testing (Frey et al., 2016)
Neuroadaptive interfaces	Adjust information or assistance in response to estimated workload	Support operators during demanding tasks	Demonstrated in simulations; live deployment needs validation (Aricò et al., 2016)
Brain–computer interfaces	Use neural signals as an additional channel of input or feedback	Hands-free interaction for suitable tasks	Demonstrated in a simulator; advantage over conventional input needs testing (Dehais et al., 2022)
Neural biometrics	Use individual neural responses as an identity signal	A possible additional factor for sensitive access	Experimental; long-term stability and usability need evaluation (Zhang et al., 2025)
Human–AI error detection	Investigate neural responses when a person notices an automated error	Potentially earlier correction of mistakes	Research stage (Dimova-Edeleva et al., 2023)
Training and simulation	Measure cognitive demands during practice	Direct training toward persistent difficulties	Studied in flight simulation; transfer needs testing (Mark et al., 2022)
Digital collaboration	Study neural activity during shared tasks	Better understanding of coordination and handoffs	Exploratory; findings about team performance are mixed (Reinero et al., 2021; Réveillé et al., 2025)

These are opportunities to investigate, not demonstrated savings in infrastructure operations. They differ substantially in cost. A one-time interface study may yield a design that everyone can use without sensors. A live adaptive system requires equipment, reliable measurement, support, and rules governing neural data.

Making human computation more effective

Neuroscience-informed interface design offers a concrete starting point for the cost analysis. Neural measures can be used during design research without requiring operators to wear sensors in everyday work. The question is whether that research leads to an interface whose benefits justify the study and redesign.

An operator diagnosing an incident must notice changes, identify relevant signals, hold competing explanations in mind, and decide what to do while new information arrives. A dashboard can display every necessary metric and still make those mental tasks unnecessarily difficult.

Traditional usability testing shows whether someone completed a task, how long it took, and where errors occurred. Neurotechnology can add evidence about what happened during the interaction. In controlled research, Frey and colleagues (2016) used EEG measures related to workload, attention, and error recognition to compare interaction methods. Such measures could help explain why one operational interface places greater demands on an operator than another. They are estimates that require interpretation alongside behaviour and task outcomes, not direct readings of thought.

Consider an incident in which an operator consults an AI assistant to connect evidence scattered across a dashboard, logs, and a runbook. A neuroscience-informed redesign might put related evidence together so that the operator can make the same diagnosis without the model call. The potential saving is the number of calls avoided multiplied by their actual cost, provided operators reach an equally accurate diagnosis without taking longer. This is a testable claim, not an assumed result.

Operators could diagnose matched incidents using the current interface and the redesigned one. Researchers would compare accuracy, time, missed signals, workload, and AI use, including cases that become more complex or require sustained attention. Neural measures would be valuable if they reveal a design problem that ordinary observation, interviews, and performance measures miss—or help predict where performance will fail as demands rise. Lower measured workload alone would not prove that an interface is better: some cognitive effort is necessary for sound judgment.

Neuroadaptive interfaces extend the idea from research to live operation. In realistic air-traffic-control tasks, Aricò and colleagues (2016) used an EEG-based workload estimate to trigger assistance and observed reduced workload. This provides a reason to investigate adaptive support in infrastructure operations, but it does not establish that continuous sensing would outperform assistance triggered by incident severity, operator requests, or other information already available to the system.

Extending the cost analysis

Interface design is one starting point. The other opportunities require similarly specific comparisons. Neural biometrics would have to justify themselves against existing authentication methods on security, speed, usability, and ongoing maintenance. A brain–computer interface would have to improve a particular task enough to outweigh its setup and operating burden. Training and collaboration tools would need to produce lasting gains in performance or coordination, not merely an interesting neural measurement.

Different applications may yield different kinds of value. One might save money; another might improve reliability or the quality of work at an added cost. A third might offer no advantage over a simpler method. Comparing each with the best available non-neural alternative is how neurotechnology becomes a defensible operational choice.

Comparing the cost cases

Operational decisions draw on several forms of intelligence. They work together, and each has costs beyond the price of its immediate output. Table 2 compares the measures that would inform a cost case for each.

Form of intelligence	Potential role in operations	What a cost case would measure
Human judgment	Interpret unusual situations and decide what action to take	Operator time, decision quality, workload, errors, and incident duration
Rules and conventional models	Filter known conditions and detect recurring patterns	Compute and maintenance costs, false alerts, and missed events
Large language models	Interpret unstructured evidence and assist with unfamiliar incidents	Usage charges, accuracy, review effort, and work avoided
Neurotechnology	Improve interfaces or adapt assistance using neural evidence	Research, sensor, governance, and deployment costs; changes in performance, AI use, and total cost

A credible comparison would begin with a baseline: incident volume and duration, operator time, errors, escalations, and AI usage. It would then measure those same outcomes after a change. For the dashboard example, the key question is whether fewer model calls and less operator effort produce at least the same quality of diagnosis—and whether the benefit pays for the study and redesign.

The analysis should also report broader effects that a financial calculation may miss. Compute and sensing equipment require energy and materials. Automation can change skill requirements and the quality of work; an interface can reduce task time while increasing strain. If neural sensing continues into everyday operations, the cost case must include the burden of wearing equipment and the governance of sensitive data, including consent, access, and retention (UNESCO, 2025). Those requirements are different for a time-limited, voluntary design study.

Conclusion

The brain’s modest energy use does not prove that human work is cheaper than AI. It points to an opportunity to use human intelligence more effectively. Neuroscience-informed interfaces offer one way to test that idea; adaptive assistance, neural biometrics, training, and collaboration require their own comparisons. The cost case for each rests on measurable improvements in decisions, total cost, or the experience of the people doing the work.

References

Aricò, P., Borghini, G., Di Flumeri, G., Colosimo, A., Bonelli, S., Golfetti, A., Pozzi, S., Imbert, J.-P., Granger, G., Benhacene, R., & Babiloni, F. (2016). Adaptive automation triggered by EEG-based mental workload index: A passive brain-computer interface application in realistic air traffic control environment. Frontiers in Human Neuroscience, 10, 539. https://doi.org/10.3389/fnhum.2016.00539

Dehais, F., Ladouce, S., Darmet, L., Nong, T.-V., Ferraro, G., Torre Tresols, J., Velut, S., & Labedan, P. (2022). Dual passive reactive brain-computer interface: A novel approach to human-machine symbiosis. Frontiers in Neuroergonomics, 3, 824780. https://doi.org/10.3389/fnrgo.2022.824780

Dimova-Edeleva, V., Soto Rivera, O., Laha, R., Figueredo, L., Zavaglia, M., & Haddadin, S. (2023). Error-related potentials in a virtual pick-and-place experiment: Toward real-world shared-control. Proceedings of the 45th Annual International Conference of the IEEE Engineering in Medicine and Biology Society. https://doi.org/10.1109/EMBC40787.2023.10340244

Frey, J., Daniel, M., Castet, J., Hachet, M., & Lotte, F. (2016). Framework for electroencephalography-based evaluation of user experience. Proceedings of the 2016 CHI Conference on Human Factors in Computing Systems. https://doi.org/10.1145/2858036.2858525

International Energy Agency. (2026). Key questions on energy and AI. https://www.iea.org/reports/key-questions-on-energy-and-ai

International Labour Organization. (2025). Generative AI and jobs: A refined global index of occupational exposure. https://www.ilo.org/publications/generative-ai-and-jobs-refined-global-index-occupational-exposure

Mark, J. A., Kraft, A. E., Ziegler, M. D., & Ayaz, H. (2022). Neuroadaptive training via fNIRS in flight simulators. Frontiers in Neuroergonomics, 3, 820523. https://doi.org/10.3389/fnrgo.2022.820523

OpenAI. (2026). API pricing. https://developers.openai.com/api/docs/pricing

Reinero, D. A., Dikker, S., & Van Bavel, J. J. (2021). Inter-brain synchrony in teams predicts collective performance. Social Cognitive and Affective Neuroscience, 16(1–2), 43–57. https://doi.org/10.1093/scan/nsaa135

Réveillé, C., Vergotte, G., Dray, G., Jean, P.-A., Perrey, S., & Bosselut, G. (2025). Trajectories of interbrain synchrony during teamwork: Links with team composition and performance. Social Cognitive and Affective Neuroscience, 20(1), nsaf081. https://doi.org/10.1093/scan/nsaf081

UNESCO. (2025). Recommendation on the ethics of neurotechnology. https://www.unesco.org/en/legal-affairs/recommendation-ethics-neurotechnology

Yu, Y., Herman, P., Rothman, D. L., Agarwal, D., & Hyder, F. (2018). Evaluating the gray and white matter energy budgets of human brain function. Journal of Cerebral Blood Flow & Metabolism, 38(8), 1339–1353. https://doi.org/10.1177/0271678X17708691

Zhang, Y., Zhang, H., Li, Y., Wang, Y., Gao, X., & Yang, C. (2025). A longitudinal EEG dataset of event-related potential. Scientific Data, 12, 1069. https://doi.org/10.1038/s41597-025-05378-x