Ghana’s AI strategy left out the power sector – It should not stay that way

As private operators prepare to take over parts of ECG’s operations in 2027, PURC and the new Responsible AI Authority should set rules now for how data and AI are used in the power sector.

Ghana’s power sector is about to change in a way few households will notice at first.

Private sector participation in the Electricity Company of Ghana (ECG) is expected to begin by early 2027, through concession arrangements rather than a sale.

Cabinet’s April 2025 approval covered billing and revenue collection. That is where artificial intelligence is most likely to arrive first, and where it can do the most harm if no one sets the rules.

The case for better tools is not in doubt. ECG’s system losses fell only marginally, from 27.05 per cent in 2024 to 26.88 per cent in 2025, against a PURC benchmark of 21.4 per cent.

The IMF projects an energy-sector shortfall of about US$1.1 billion in 2026, driven largely by collection and distribution losses and costly generation contracts.

Every operator bidding for ECG work will promise to close that gap with data: smart meters, software that flags suspected theft, predictive maintenance on overloaded transformers, and models that decide where revenue-protection teams go.

These tools work elsewhere and can work here.

But they make decisions that land on people.

A model that flags a meter as tampered can trigger a disconnection or a penalty. An algorithm’s estimated bill can push a family into arrears.

Load-management software decides which feeders go off first when supply is short.

When a private operator makes those decisions, with data generated by Ghanaian customers, on systems the regulator cannot inspect, accountability becomes very hard to enforce.

The risk is not hypothetical. ECG’s own communications director has acknowledged that many meters are outdated or were installed without the company’s involvement, and that hundreds remain outside its digital system.

A model trained on incomplete records will be wrong more often for the customers those records miss, and its errors will not fall evenly.

The timing is fortunate. Ghana launched its National Artificial Intelligence Strategy 2025–2035 in April and committed to establishing a Responsible AI Authority within the first year.

An Emerging Technologies Bill, due before Parliament, would create an agency to set ethics and safety rules for AI.

Yet the strategy’s priority sectors are agriculture, health care, education and government services.

Energy, which the Finance Minister has called the country’s single-largest economic risk, is not among them.

That gap should close before the concession contracts are signed, not after.

First, write data and AI terms into the concession agreements.

Customer and network data should remain the property of the state utility, with full access for ECG and PURC.

Contracts should give the regulator the right to audit any model used for billing, theft detection or disconnection, and require operators to hand over data and documentation when the contract ends.

Second, guarantee human review. No household or business should be disconnected or surcharged on the strength of an automated flag alone.

A trained officer should confirm the finding, the customer should be told why, and a quick, free appeal should run through PURC.

Third, verify performance independently. If operators are paid for cutting losses, the loss figures cannot come only from their own systems. PURC or the Energy Commission should hold the reference data and confirm the gains reported.

Fourth, test for fairness. Load-shedding schedules and revenue-protection targeting should be published and checked for whether some districts or income groups consistently bear more of the burden.

Fifth, build capacity inside the regulators. PURC and the Energy Commission need in-house data scientists who can read a model’s documentation and test its outputs.

A small energy-and-AI working group under the Responsible AI Authority, drawing on PURC, the Energy Commission, ECG, NEDCo, GRIDCo and the universities, would be a practical start.

Clear rules lower risk for investors, who then know in advance what will be required of them.

They also protect the programme politically. Organised Labour already opposes private participation in ECG, and one high-profile case of wrongful disconnections could stall the whole effort.

Ghana has an unusual chance to get the sequence right: to set the rules for algorithms in its power sector before those algorithms run at scale.

The Responsible AI Authority and PURC should take it.

The writer (PhD) is an energy systems engineer and a Lead Author for Working Group III of the Intergovernmental Panel on Climate Change (IPCC) Seventh Assessment Report. He serves as a technical expert with UNESCO on the ethics of artificial intelligence. He writes in a personal capacity.

Email: This email address is being protected from spambots. You need JavaScript enabled to view it.