State-owned power producer Eskom is looking to cut the number of AI pilots across the organisation in a bid to avoid wasteful expenditure, according to its CIO, Len de Villiers.
Speaking at the industrial digital and intelligent transformation forum hosted by Huawei at Mobile World Congress in Barcelona, Spain on Monday, De Villiers said organisations must balance the quest for innovation with their duty to ensure a return on investment from AI spend.
“You will not stop AI affecting your organisation – if you ignore it, you will be in trouble. We’re going to make sure that AI is not inhibited but carefully directed. I am sitting with 220 AI pilots in our company and the issue with all those pilots is they have to have a business case. That business case must make sense, and there has to be a return on investment,” said De Villiers.
AI at Eskom has been deployed across a range of business units. Some of the most promising use cases relate to developing predictive maintenance capabilities, embedding intelligence into the national grid and the digital twinning of power stations.
One of the most important drivers for these projects is the ongoing liberalisation of South Africa’s electricity sector, which aims to stabilise electricity supply and reduce costs by introducing a large number of independent power producers into the sector.
Pricing models
The envisioned system aims to make extensive use of dynamic pricing models, feed-in tariffs for residential and commercial customers, and the wheeling of power across the grid. Eskom aims to use AI to manage the complexity of the system and maximise its efficiency, too.
De Villiers cited Johnson & Johnson, which ran 900 AI pilots and eventually cut 700 of them to reduce costs. “Can you imagine all the money spent on the 700 that was wasted?” he asked.
Part of Eskom’s rationalisation exercise involves asking department leads to prioritise their AI pilots and select those with the most promise.
Read: How liberalisation is rewiring South Africa’s power sector
“You can’t do everything your company wants to do on AI; it’s just impossible. Give each business unit – and there are very big business units in our company – a challenge to prioritise at least three or five, work with them per business case and let them manifest. But don’t boil the ocean: you will lose money and you will lose focus,” said De Villers. — (c) 2026 NewsCentral Media
Get breaking news from TechCentral on WhatsApp. Sign up here.