How AI helped TymeBank gain 670K customers

Read time 3min 30sec
Dietmar Bohmer, chief data scientist at TymeBank.
Dietmar Bohmer, chief data scientist at TymeBank.

Digital-only bank TymeBank has notched up around 670 000 customers since November last year, thanks to technologies such as artificial intelligence (AI), machine learning (ML) and big data.

This is according to Dietmar Bohmer, chief data scientist at TymeBank addressing the ITWeb Artificial Intelligence 2019 event, held in Johannesburg yesterday.

Bohmer discussed how innovative use of cloud-based technologies has redefined banking models, allowing SA's first digital bank to disrupt traditional banking models by delivering affordable services to under-serviced South Africans.

The bank, which soft launched in November 2018 and officially launched in February, does not have any branches and relies solely on digital means (mobile app and Web site), and kiosks.

Its heavy reliance on technology allows it to have a staff complement of only 250 employees, compared to the almost 50 000 employees in the average traditional South African bank.

Furthermore, TymeBank has a small contact centre, which means most of its customer support services and communication are delivered digitally and in real-time, explained Bohmer.

“Although TymeBank is not only about advanced technologies, they do play an important role in meeting our value proposition, which is to offer 'simple, accessible and affordable banking'. A big proportion of our transactions are free, and technology plays a big role in being able to offer a full range of banking services at significantly low prices.”

TymeBank has formed a long-term strategic partnership with retail giant Pick n Pay and Boxer stores, allowing its customers to bank at these retailers throughout the country.

It takes less than five minutes for customers to open an account on the zero-rated mobile app.

After opening an account, the user has limited services and is required to verify their details at their nearest Pick n Pay or Boxer store.

At the kiosk, TymeBank's digital system links with the home affairs database to capture biometric data, which verifies the customer's identity.

“The customer’s on-boarding process is paperless and fully digital. The ID verification is done by using ML and advanced algorithms to match their fingerprint to the home affairs system. The physical address verification is conducted by comparing our information with databases from the Deeds Office, the municipality or the customer’s home loan account. We only use official sources to regulate the know-your-customer process, as we are prohibited by law from using sources like social media, etc.”

Although the application success rate is extremely high, there are rare cases where technical issues can hinder completion of the on-boarding process.

“Through AI, we are able to detect reasons in real-time why the application could not be completed, and the client is immediately informed via various communication channels, such as SMS or an e-mail, explaining what is hindering the process,” he continued.

“The same happens during financial transactions which cannot be completed; if for instance, a client is unable to withdraw from the ATM, unknowing that they have exceeded their daily limit. We are able to detect this and immediately send them communication explaining the problem.”

TymeBank also has a separate financial education app called TymeCoach, which helps users make better decisions with their money. The app has a chatbot which uses AI to answer financial management queries and provide clients with their credit report information.

Bohmer stressed the importance of providing pro-active customer support around the clock and using self-learning technology models to consistently track banking activity. Another important factor is ensuring the right communication is sent to the right client at the right time, he added.

“Operating without a physical branch and with a small contact centre means the bank has to track and see every activity, not only to detect technical errors but also fraudulent activity. We are always better-personalising our offerings using ML to offer customers the products and services most suitable to them.”

See also