How AI is impacting the payments industry in South Africa

Machine learning, for example, can be used to leverage internal and external data in assessing risk and anticipating customer needs in real time.

Kiaan Pillay, Co-founder and CEO at Stitch. Portal ERP South Africa. 

AI is quickly making inroads across every industry. The payments and fintech space is no different – here, AI can be leveraged to enable efficiencies in fraud prevention, transaction initiation, engineering productivity and much more. For fintechs, it’s less about using one tool, but rather, an ecosystem of technologies that enhance decision-making, streamline operations, and create new value for both service providers and their customers. 

Machine learning, for example, can be used to leverage internal and external data in assessing risk and anticipating customer needs in real time. For compliance, facial recognition tools are speeding up Know-Your-Customer (KYC) checks and onboarding. Natural language processing (NLP) helps virtual assistants to read and understand contracts, queries, emails, and forms. Generative AI simplifies content creation and enhances the product design process, while AI agents - autonomous assistants - can reason and achieve a desired goal, including payment initiation and support queries. 

For Stitch and payment infrastructure companies like us, AI is deeply embedded in the way we do business as a core driver of efficiency and scale. It might not be as exciting as a flashy front-end feature, but when it comes to processing payments at scale, the behind-the-scenes impact can be massive. 

“Everyone is looking for a very sexy, big uplift – some sort of magic that happens with a product. I'm not super convinced that's the thing yet. But, turn over a rock on anything less glitzy (fraud, recon, support, etc) and I bet you'll see a step change in difference over the last two years.” - Kiaan Pillay, Co-founder and CEO at Stitch

We caught up with our co-founders, Kiaan Pillay (CEO), Priyen Pillay (CTO) and Junaid Dadan (President) to discuss the ways AI is being leveraged to enable greater efficiency for the Stitch team, and how they see this technology growing over time.

Smarter fraud prevention and detection

In 2023, financial crime cost South Africa’s finance industry R 3.3 billion. More than two-thirds of the losses are attributed to fraudulent Card Not Present (CNP), or online card transactions. As a result, banks are highly vigilant about any suspicious transactions, putting intense pressure on Payment Service Providers (PSPs) to manage and respond to high-volume queries and disputes within tight Service Level Agreements (SLAs). 

AI-powered fraud prevention systems help the team at Stitch respond to potential fraud with accelerated speed and accuracy. These systems use machine learning models to detect and flag unusual transactions in real-time, while rule-based logic identifies edge cases that need a human to review and make a final call. 

A PSP that’s starting out may initially hire people to manually manage the cases, which are sent to them via email. This presents several challenges. First, the team needs to be available 24/7 to answer urgent emails – and for banks, nearly all queries are urgent. Secondly, emails are unstructured, which reduces the speed to close the case within the SLA. Third, as clients start to increase volumes, scaling at speed becomes an issue. 

To automate and scale this workflow, AI agents can be used to monitor inboxes, structure data, and trigger alerts in compliance systems. This means that teams can react a lot more quickly to non-routine incidents, helping Stitch manage high-volume queries faster. As they process more data over time, their accuracy will increase, making them even more reliable.

Stitch has been running an AI agent for over two years. It runs 24/7 and responds to all emails without human intervention. When the system launched, it had a 50% accuracy rate; within months, it climbed to 90%, and today it operates at 98% — dramatically improving response time and scalability.

Enhanced productivity and product design 

From sales, to marketing, to operations, to compliance, all functions of a fintech business can use AI to increase productivity. 

Automated versus autonomous workflows

Automating manual processes is part of every industry’s effort to increase productivity. However, you don’t need AI to achieve automation. AI goes a step further by creating autonomous workflows that can reduce time spent on recurring tasks. 

Generally speaking, AI agents can help teams uplift capacity by carrying out repetitive tasks that require a bit of thinking, such as getting approval on something. Instead of going through the process of referring to previous emails, finding time in a calendar, or double-checking figures before sending the request, the AI agent receives the objective and goes through the entire process on the user’s behalf. 

In fintech, they can also be used to enable more seamless experiences for consumers - from generating support queries based on specific issues, to accessing a fintech’s backend system to manage and initiate financial transactions, thanks to protocols like MCP which can be used to build AI agent toolkits and provide agents with easy access to existing systems. 

As maturity grows, AI agents will become a standard way of uplifting capacity, providing opportunities for teams to focus on creating value for the firm.

Empowering engineering teams

Tools like Cursor or Windsurf use simple, everyday language prompts to help engineers generate code. This opens up capacity for more complex problem-solving and additional builds. It also opens up the talent pool, gradually shifting the focus from raw technical coding skills to system-level thinking and value creation.

“The engineering skillset required is changing. First, engineers need to be very familiar with AI, and how to use AI tools to be very productive. Second, it's less about writing the code and more about thinking through how systems work, how to meet customer needs, and generally thinking at a higher level.” - Co-founder and President, Junaid Dadan

To ensure teams are focusing on projects that are adding value to the client, AI can be used to help the team build stable, self-sustaining systems faster, rather than having to spend extra time on maintenance. “This might include building redundancy and failover procedures that utilise AI to monitor performance, detecting anomalies and ideally, self-healing. For us, this means our engineers aren’t maintaining systems in perpetuity; rather, they build and automate themselves out of a completed product and move on to the next high-impact project,” said Priyen Pillay, Co-founder and CTO at Stitch. 

Accelerating sales cycles

Sales is an interesting area for AI-driven productivity at Stitch due to the tailored nature of our services. Teams at Stitch are constantly navigating unique customer challenges, tailored solutions and past decisions, but that knowledge has previously been  scattered across channels, buried in old contracts, or locked in someone’s memory.

AI tools simplify this. With a tool trained on past deals, contracts, and pricing models, reps could just ask, “How have we handled this for a similar client in the past?” and get an instant, reliable answer. This reduces friction, speeds up deal cycles, and ensures more consistent decision-making.

Enriching product development

AI is also transforming how quickly product teams can get from concept to prototype. “Our teams are encouraged to use tools like Vercel v0, GPTs or AI-enhanced dev environments to get an early signal on what’s worth pursuing. While the output is not a fully fledged product, it is enough to spark conversation, test ideas, and sometimes form a decent base to kickstart off.” said Priyen Pillay.

Product specifications and similar documents still play a significant role in the product development process. However, with the addition of these tools, the process can be enriched with artifacts that are explored before committing to a build. It's a shift from static planning to dynamic design.

AI implementation is not without risks 

Fraud is getting smarter

In the context of fraud, AI has become a double-edged sword. While the technology is helping fintechs better detect and fight fraud, it’s also giving fraudsters better tools. Today, it’s easier than ever to create fake websites, cloned voices and hyperrealistic images that make it harder to identify scams. In one case, a person lost over R6 million after being misled into trading on the “Johannesburg Stock Exchange”. 

The arms race to implement the smartest tools is pronounced for ID verification or facial verification fintechs. As deepfakes become commonplace and harder to detect, firms providing these authentication services are facing growing threats, as it becomes easier than ever to falsify KYC.

Authenticating agent-led payments 

Today, payments are centred around human-facing interfaces: users interact with the checkout page by filling in details and authenticating themselves. On the back end, financial systems connect API to API to settle the transactions.

That model is going to change drastically. In the future, AI agents will interact with these pages and make payments on behalf of the user. For example, how do we adapt 3DS authentication flows designed for humans to protocols that can authenticate a certified AI agent that is authorised to act on behalf of the user? Similarly, backend API to API integrations need to consider agent to agent activities. 

This is one of the biggest shifts in systems design and infrastructure we will see in payments, according to Junaid Dadan. 

Conclusion

Businesses that are focussed on scale, speed and a strong competitive advantage are finding more ways to embed AI tools into their day to day operations – from machine learning, to NLP, generative models and agentic AI. 

For the payments space specifically, businesses like Stitch are using these tools to streamline fraud prevention and detection, optimise operational efficiency, improve capacity and accelerate the product design and build cycle.