TCS - The TechCentral Show

Pick n Pay has switched on an AI shopping companion called Penny inside its asap! app, and in this episode of the TechCentral Show, retail executive for omnichannel Enrico Ferigolli takes editor Duncan McLeod through what it does and why it matters.

Built on Google's Gemini models, Penny lets customers build a grocery basket by asking for what they want in their own words – by voice, text or photo – instead of searching and scrolling. Tap the "Ask" button, request a carbonara recipe, and Penny returns the method alongside a carousel of options for each ingredient to drop into the basket. It handles re-orders, meal planning to a budget and ingredient substitutions, and it reads photographs, too: Ferigolli describes snapping a handwritten shopping list and having Penny build the basket from it.

In the conversation, Ferigolli is candid about the limits. Penny does not place orders yet – it assembles the basket and hands the customer back to checkout – and its language support, while broad, is stronger by voice than by text and still maturing for South Africa's African languages. He explains why Pick n Pay tested several large language models before settling on Gemini, how the system draws on the app's own search, order history and Smart Shopper data rather than plugging Gemini straight into its databases, and where a retail-media layer fits in.

For more detail on the launch, see TechCentral’s full report on Penny and how it works.

Ferigolli discusses:

• How conversational shopping changes the asap! experience

• Why Gemini won on accuracy, speed and cost

• What Penny can and can't do at launch

• The multilingual and multimodal ambitions behind it

• How the 2025 asap! rebuild set this up, and what comes next

Pick n Pay is "a little bit behind" in online grocery, Ferigolli concedes in the podcast – but with Penny and the features to follow, he reckons it will "get ahead really, really fast".

Don't miss the discussion.

South Africa has fewer than 400 public electric vehicle charging stations – up from zero just 15 years ago – and EV adoption remains stubbornly slow. Yet Charge, formerly known as Zero Carbon Charge, is betting big that a coast-to-coast network of off-grid, renewable-powered charging stations is exactly what’s needed to fire up the local market.

In this episode of the TechCentral Show, Joubert Roux, co-founder and director of Charge, joins TechCentral’s Nkosinathi Ndlovu to make the case for the company’s ambitious, R1.8-billion plan to roll out a charging station every 150km along South Africa’s national highways – and to explain why he believes the company is taking on “a timing risk, but not a business risk”.

Roux walks TechCentral through the December 2024 launch of Charge’s first site near Wolmaransstad and the unit economics underpinning the roll-out: just seven vehicles a day at each station are needed to reach Ebitda break-even. He also explains why every facility is designed to operate entirely off-grid, citing data showing that EVs charged on Eskom’s coal-heavy network emit 5.8 tonnes of carbon-dioxide a year, more than a comparable petrol car at 4.4 tonnes.

The conversation also tackles Charge’s unconventional fundraising strategy: a tokenised public offering on Mesh rather than a JSE listing, planned for June 2026. Roux argues that South Africa’s institutional capital is “extremely conservative” and that tokenisation will finally let ordinary investors into an infrastructure deal that has historically demanded R1-million minimums. The Development Bank of Southern Africa has already committed R100-million.

Roux and Ndlovu also discuss:

• How landowners hosting Charge stations receive 5% of charging revenue, and the rural economic development case that sits behind that model;

• The offtake agreement with transport aggregator Zimi covering 50% of capacity at upcoming N3 corridor sites;

• Charge’s formal objection to Sanral’s proposed policy giving it powers over businesses within 60m of national roads or 500m of interchanges, and the broader regulatory headwinds facing EV infrastructure;

• How BYD’s planned 1MW supercharger network and incumbent operators like GridCars – which already records 5 000 charge sessions a month – are reshaping the competitive landscape;

• Plans for 35MW truck-charging facilities and a long-term target of 120 stations across the national route network; and

• Roux’s prediction on when South Africa will hit its EV tipping point – and the two ingredients he says the market still needs: sub-R500 000 EVs and a genuinely reliable national charging network.

Don’t miss the discussion!

AI Diagnostics, a Cape Town-based med-tech company that has built an AI-powered stethoscope designed to detect tuberculosis, recently raised R85-million in a pre-series-A funding round.

In this episode of the TechCentral Show (TCS), Nkosinathi Ndlovu speaks to the company’s CEO, Braden van Breda, about the funding round and its mission to transform TB screening across Africa and Asia.

The round, led by the Steele Foundation for Hope, will fund the deployment of the Ostium – an AI-powered digital stethoscope that works in tandem with the company’s proprietary AI.TB model.

The device is designed to detect signals associated with tuberculosis in lung sounds in real time, putting point-of-care diagnostic capabilities into the hands of the nurses, pharmacists and community health workers who often serve as the first contact in resource-constrained health systems.

In a wide-ranging conversation, Van Breda explains why AI Diagnostics chose to tackle one of the world’s most underfunded diseases, and how an AI model can pick up TB signals that escape even trained clinicians. He also unpacks the painstaking work of building a TB-positive lung-sound dataset from scratch.

Van Breda also delves into:

• What the Ostium compares to a traditional stethoscope;

• How the company handles real-world recording conditions, from noisy clinics to paediatric patients and HIV co-infection;

• How AI Diagnostics stops the model from silently degrading as the device moves from a Khayelitsha clinic to rural Zambia or Vietnam;

• The company’s goal of securing World Health Organisation certification and the path towards that;

• How the Ostium is positioned against – or alongside – AI-assisted chest X-ray tools such as CAD4TB and Qure.ai's qXR; and

• What it would realistically take to put an Ostium in every primary healthcare clinic on the continent by 2030.

Don't miss the discussion!

The pace at which artificial intelligence is reshaping the threat landscape is outstripping the ability of most organisations to defend themselves, with shadow AI, synthetic identity attacks and a looming quantum computing disruption all converging at once.

That’s the view of DataGroupIT CEO Werner Lindemann, who joined Duncan McLeod on the TechCentral Show to unpack what business leaders should be doing about AI and information security.

Lindemann, who spent more than 30 years in senior roles at BCX and Clickatell before joining the security solutions distributor, says the African threat environment is no longer a watered-down version of what is happening elsewhere. Attackers are deploying the same AI-powered tools globally, and AI-enabled phishing campaigns now achieve click-through rates that traditional defences were never designed to withstand.

A bigger blind spot, he argues, is shadow AI – employees pasting sensitive data into unapproved AI tools without oversight. Lindemann says this is fast eclipsing the shadow IT problem of the past decade because the tools are free, frictionless and often invisible to security teams.

The conversation also tackles the credibility crisis facing identity verification. With AI now able to clone a CEO’s voice in real time or generate synthetic profiles that pass biometric checks, Lindemann believes traditional verification methods are fundamentally flawed. A big challenge is helping boards understand the issue in business rather than technical terms.

Lindemann also weighs in on the rise of the chief AI officer role, following Sanlam’s recent appointment, and on whether African organisations are equipped to adopt AI at the pace global peers are setting given the continent’s acute skills shortage.

The discussion closes on quantum computing. Lindemann challenges the conventional view that the quantum threat is a decade away, and outlines what business leaders should be doing now to prepare for the post-quantum cryptography world – even if the risk still feels distant.

Award-winning South African film director Donovan Marsh has pivoted to artificial intelligence filmmaking and believes generative AI tools could fundamentally reshape how movies are made – and who gets to make them.

Marsh, whose 30-year career includes directing the Hollywood submarine thriller Hunter Killer starring Gerard Butler and Gary Oldman, the Spud films and iNumber Number, is the latest guest on the TechCentral Show.

The economics, he tells TechCentral editor Duncan McLeod, are extraordinary: a single complex scene in a traditional production requires crews, equipment, locations and days of scheduling, while AI tools collapse much of that overhead into work that can be done at a desk.

But Marsh is clear that the creative work has not disappeared. He still directs shot by shot, much as he would on a conventional set, and uses a patchwork of different AI tools – no single product yet does everything. He has found that simpler prompts produce better results, saying over-prescription tends to degrade output quality.

Marsh acknowledges the disruption this implies for camera operators, lighting crews, set designers and extras. But he argues that AI filmmaking could prove liberating for smaller filmmaking markets like South Africa, where the budgets to make ambitious local movies have dwindled.

He has co-founded Dragon Studios AI with Ronnie Apteker and Stephen Cholerton, and is developing what he believes will be among the first AI-generated feature films. The tools are not quite there yet for a full 90-minute production, he says, but the gap is closing fast.

Marsh also weighs in on where the so-called “uncanny valley” still trips up generative video, the future of the acting profession and what AI filmmaking could look like by 2029.