Autonomous Decision-Support Systems in Energy Law

1. Introduction

Autonomous Decision-Support Systems (ADSS) are computer-based systems that use data, algorithms, artificial intelligence (AI), predictive models and automated analysis to assist energy-sector decision-makers. These systems can analyse large amounts of information and provide recommendations about electricity dispatch, tariffs, grid management, renewable-energy integration, licensing, demand forecasting, infrastructure planning and compliance.

The word “autonomous” does not necessarily mean that the machine has complete legal authority. In energy law, an autonomous system may analyse information and recommend a decision, while the final legally binding decision remains with a regulator, system operator or other authorised institution.

This distinction is important because electricity decisions can affect consumers, generators, municipalities and the stability of the national grid.

2. Meaning and Functions

An autonomous decision-support system generally performs five functions:

Data collection – gathers information from meters, generators, weather systems, market platforms and grid equipment.

Data analysis – processes large amounts of technical and financial information.

Prediction – forecasts demand, generation, prices, congestion or system risks.

Recommendation – suggests a particular regulatory or operational action.

Human decision-making – an authorised person or institution reviews and adopts, modifies or rejects the recommendation.

For example, an AI system could predict electricity demand for the next 24 hours and recommend how much generation should be scheduled. A regulatory system could analyse tariff information and identify whether a proposed tariff requires further investigation.

NERSA already uses specialised regulatory methodologies and technical information when making electricity-sector decisions. Its regulatory framework includes licensing, compliance monitoring, dispute resolution and electricity-market regulation.

3. Legal Framework in South Africa

The principal legislation is the Electricity Regulation Act 4 of 2006 (ERA) together with the National Energy Regulator Act 40 of 2004 (NERA) and the Promotion of Administrative Justice Act 3 of 2000 (PAJA).

NERSA has authority to regulate electricity activities, issue licences and conditions, monitor compliance and make regulatory decisions.

The South African Grid Code also provides technical requirements for electricity-system operation. Its scheduling and dispatch rules recognise real-time balancing between electricity supply and demand and establish rules for controllability and dispatch.

An autonomous decision-support system must therefore operate within existing legal authority. It should support, rather than unlawfully replace, the decision-maker.

4. Relevant Case Laws

Pharmaceutical Manufacturers Association v President (2000)

In Pharmaceutical Manufacturers Association of South Africa v President of the Republic of South Africa, the Constitutional Court established the importance of legality and rationality in the exercise of public power.

This principle is directly relevant to AI decision-support systems. A regulator cannot simply adopt an algorithmic recommendation without considering whether the resulting decision is authorised by law and rationally connected to the statutory purpose.

An AI system cannot create legal authority for itself.

Eskom Holdings v Vaal River Development Association (2022)

In Eskom Holdings SOC Ltd v Vaal River Development Association, the Constitutional Court examined electricity regulation, Eskom's powers, NERSA's regulatory role, administrative-law principles and the consequences of decisions affecting electricity supply.

The judgment describes a comprehensive regulatory framework in which NERSA can regulate matters such as electricity production, prices, supply conditions and licence compliance. It also recognises the importance of rationality and procedural fairness.

For autonomous decision-support systems, this means that technology must operate inside the statutory electricity framework and cannot replace legally required regulatory procedures.

Earthlife Africa Johannesburg v Minister of Environmental Affairs (2017)

In Earthlife Africa Johannesburg v Minister of Environmental Affairs, the court required proper consideration of climate-change impacts in relation to a proposed coal-fired power station.

The case demonstrates an important principle for automated decision-support systems: all legally relevant factors must be considered.

An algorithm designed only to minimise electricity costs might produce a technically efficient recommendation but fail to consider environmental, climate or constitutional requirements. Such a system would therefore need to be designed to identify legally relevant considerations.

Afriforum NPC v NERSA (2025)

The recent Afriforum NPC v NERSA litigation concerned NERSA's regulatory decision-making and public participation. The court emphasised that technical expertise does not remove procedural requirements relating to participation, accountability and lawful administration.

This is significant for autonomous decision-support systems. Even where a computer model provides technically sophisticated analysis, the regulator may still need to follow consultation and procedural requirements before making a binding decision.

5. Major Legal Issues

Human Accountability

The most important issue is responsibility. If an AI recommendation causes harm, the responsible regulator or operator cannot simply state that the computer produced the recommendation.

Transparency

Affected parties may need to understand the information, methodology and reasoning behind an important decision.

Accuracy and Bias

Incorrect data can produce incorrect recommendations. Algorithms may also reproduce biases contained in historical data.

Procedural Fairness

Where the final decision constitutes administrative action, applicable procedural requirements under constitutional and administrative law remain important.

Cybersecurity

An attacker who manipulates the data entering an AI system could cause the system to produce an incorrect recommendation.

Auditability

Energy decision-support systems should maintain records of data inputs, assumptions, model outputs and human decisions so that decisions can later be reviewed.

6. Importance for Future Energy Governance

ADSS can improve energy governance by allowing regulators and system operators to process information much faster than traditional manual systems. They can assist with renewable-energy forecasting, electricity dispatch, congestion management, tariff analysis, infrastructure planning, market surveillance and environmental compliance.

However, NERSA's own regulatory materials emphasise that regulatory methodologies operate within legally authorised decision-making processes and do not themselves replace the legal framework.

Therefore, the technology should remain a decision-support mechanism rather than an independent legal authority.

7. Conclusion

Autonomous Decision-Support Systems can significantly improve modern energy governance. They can analyse complex information, identify risks and provide rapid recommendations to regulators and electricity-system operators.

However, South African law requires that important energy decisions remain connected to legality, rationality, procedural fairness, transparency and statutory authority.

The principles emerging from Pharmaceutical Manufacturers, Eskom Holdings v Vaal River Development Association, Earthlife Africa and Afriforum v NERSA demonstrate that technological sophistication cannot replace lawful decision-making.

The appropriate model is therefore “AI-assisted decision-making under human and legal accountability.” Autonomous systems may collect information, analyse risks and recommend decisions, but authorised human institutions should remain responsible for adopting legally binding decisions, explaining them, reviewing errors and protecting affected parties.