Case studies

Built a graph machine learning fraud pipeline (using Memgraph) that scores more than 3.5 million records a day in about two hours across seven live graphs, spotting scam payments, mule accounts and fraudulent recipients in real time. It has also rolled out agentic and generative AI to nearly 5,000 staff.

The payoffAI fraud tools saved clients R673 million in the 2026 financial year, stopping over 394,000 scam payments, blocking 131,000 fraudulent recipients and shutting more than 80,000 mule accounts, at a monthly false positive rate of about 2.1 percent.

Uses in-house AI-driven credit risk models and alternative data (helped by a data analytics and credit-risk partnership with Brazil's Nubank) to lend to customers who lack a traditional credit history, and to score risk at low cost for its digital-only, kiosk-plus-app model.

The payoffTymeBank/Tyme Group uses AI and alternative data (bolstered by a data-analytics and credit-risk partnership with Brazil's Nubank, which took ~10% in its Series D) to score and lend to thin-file customers at low cost via its digital-only kiosk-plus-app model, with SME lending (merchant cash advances) up around 30% year-on-year in 2024. TymeBank South Africa reached profitability in December 2023, and Tyme Group achieved unicorn status at a $1.5bn valuation in late 2024 (not 2025). By June 2025 the group served about 17.5 million customers across South Africa and the Philippines, adding roughly 450,000 a month.

Embedded AI into its 2026 to 2028 strategy: an AI recommendation engine called SmartNudge for business cross-selling, conversational AI across its mobile apps, AI-driven marketing, and an internal Microsoft Copilot Studio service-desk bot named Karabo. Runs on cloud and Microsoft Power Platform.

The payoffAI-enabled cross-sell tools lifted campaign outcomes by about 20 percent, SmartNudge reaches a 66 percent acceptance rate, conversational AI now handles about 65 percent of all digital queries, and the Karabo bot resolves about 99 percent of employee IT queries. Standard Bank ranked second in the Evident AI Index for Banks, Middle East and Africa.

Deployed roughly 30 advanced AI use cases in 2025, including Microsoft Copilot for customer engagement across 13 South African languages and Azure-based compliance and data-privacy tooling, to automate internal work and speed up payments processing.

The payoffThe 2025 AI programme saved about 312,000 staff hours, cut global payment processing times by around 30 percent, and delivered roughly R186 million in realised benefits. Nedbank placed fourth in the Evident AI Index for Banks, Middle East and Africa.

Partnered with Google Cloud to build Vitality AI on Vertex AI and Gemini, mining a de-identified dataset of more than 2,800 dimensions of lifestyle, clinical and behavioural data to personalise health screening and nudges for members.

The payoffPersonalised, data-backed screening drove a 5.5 times increase in screening rates in fully funded healthcare settings and a 19 percent improvement in early cancer detection, with actuarial value for insurers through lower healthcare costs and improved loss ratios. Rollout is planned across Vitality's global base of about 13 million customers from 2026.

A fully digital insurer that uses AI to assess risk and generate binding car, home and single-item quotes in about 90 seconds, and handle claims and cover changes without a call centre. In 2026 it became the first insurer in the world to give a final, binding car insurance quote inside a native ChatGPT app.

The payoffExact profit figures are not public, but the AI-first model let Naked run 100 percent of its policies online and raise a 38 million US dollar Series B extension in early 2025 (from IFC, DEG, BlueOrchard, Yellowwoods and Hollard), reported as the largest insurtech investment in Africa, to scale its AI-powered growth.

Through its Shoprite X innovation hub, feeds Xtra Savings loyalty data (about 2,500 card swipes a minute, said to be Africa's largest consumer dataset) into AI for demand forecasting, dynamic pricing and stock optimisation. It uses a machine learning algorithm to define optimal Sixty60 delivery regions in real time, and launched Pixie, a swipe-based AI shopping assistant that predicts and suggests grocery orders.

The payoffThrough its ShopriteX innovation hub, Shoprite feeds Xtra Savings loyalty data (about 2,500 card swipes a minute, since risen to ~2,700, from tens of millions of members, one of the largest consumer datasets in Africa) into AI for demand forecasting, pricing and stock optimisation. It uses a machine-learning algorithm to define optimal Sixty60 delivery regions, and launched Pixie, a swipe-based AI shopping assistant that predicts and suggests grocery orders. Rand savings are not published, but data-driven forecasting cuts costs and stockouts, ML delivery-zone routing helps meet the 60-minute window, and Pixie became one of Sixty60's fastest-adopted features, used by about 98% of Xtra Savings Plus members, deepening basket size on a service that grew into roughly US$1 billion in annual sales (FY2025 sales ~R18.9bn), not a business "worth" $1 billion. Sources: - https://techcentral.co.za/how-shoprite-is-using-ai-more-groceries/253336/ - https://www.dailymaverick.co.za/article/2022-11-13-artificial-intelligence-is-playing-a-bigger-role-in-retail-and-shoprite-is-leading-the-race/ - https://www.shopriteholdings.co.za/newsroom/2022/the-shoprite-group-innovates-using-ai.html - https://www.news24.com/business/companies/embargo-for-0830-monday-morning-shoprite-says-tinder-style-ai-shopping-assistant-on-sixty60-is-a-massive-hit-20260711-0631 - https://www.itweb.co.za/article/shoprite-adds-ai-shopping-assistant-to-sixty60-app/rxP3jqBELozMA2ye - https://africannewsagency.com/shoprites-ai-pixie-reaches-98-adoption-among-sixty60-members/ - https://launchbaseafrica.com/2025/10/16/sixty60-the-1-7m-seed-investment-that-built-a-1b-delivery-empire-for-shoprite/

Signed a multi-year collaboration with Google Cloud (announced November 2025) to unify fragmented network, billing, support and mobile-money data into a single BigQuery backbone and use Vertex AI and Gemini to optimise network performance, personalise customer care and fight fraud with machine learning.

The payoffReal and largely accurate, with one figure to drop. On 25 November 2025 Vodacom Group announced a multi-year strategic collaboration with Google Cloud to migrate and unify its critical data platforms onto Google Cloud's data cloud (including BigQuery) as a single secure source of truth, and to apply Vertex AI and Gemini (plus Veo and Imagen) to optimise network performance, personalise customer care, and detect fraud with machine learning, across a group serving more than 223.2 million customers (93.7 million financial services users). Remove the "roughly 1 billion US dollar" figure: no official Vodacom or Google Cloud disclosure states a deal value, and the $1B number originates only from a single third-party analysis headline that presents it speculatively as a question with no sourcing. Financial terms were not disclosed. Sources: Vodacom (https://www.vodacom.com/news-article.php?articleID=16507), Google Cloud press corner (https://www.googlecloudpresscorner.com/2025-11-25-Vodacom-Announces-Multi-Year-Strategic-Collaboration-with-Google-Cloud-to-Boost-Africas-AI-Advancement), PR Newswire (https://www.prnewswire.com/news-releases/vodacom-announces-multi-year-strategic-collaboration-with-google-cloud-to-boost-africas-ai-advancement-302624944.html), Connecting Africa (https://www.connectingafrica.com/ai/vodacom-google-cloud-sign-multi-year-deal-to-advance-ai-services). The unsourced $1B claim: Global Data Center Hub (https://www.globaldatacenterhub.com/p/will-vodacoms-1b-google-cloud-bet).

Showmax uses a machine learning recommendation engine (built with personalisation vendor Recombee) that analyses viewing behaviour, genre and other signals to auto-curate personalised content rows and surface what each subscriber is most likely to watch.

The payoffExact figures are not public, but the recommendation system is credited with higher engagement and conversion and lower subscriber churn by helping viewers find value faster, a key retention lever for MultiChoice's streaming push against global players.

Uses AI, machine learning and robotics to personalise product offers, and built a machine learning application called Manila to automate regulatory decisions faster and more accurately. Rolled out Navi, an AI virtual assistant in the banking app, and later embedded AI into advisor workflows for real-time, human-backed digital support.

The payoffSpecific rand impact is not disclosed, but the AI tooling targets faster, more accurate regulatory decisions and personalised cross-sell to a South African base where digitally active banking customers now exceed 23 million, defending market share against digital challengers.