“South African companies must embed AI into their businesses to thrive but they need the right data first,” says Hein Pretorius, CEO of technology strategy firm Onpro Consulting.
A report jointly compiled by Accenture and the Gordon Institute of Business Science, titled Artificial Intelligence: Is South Africa Ready?, predicts that in five years, more than half of consumers and enterprise clients will no longer select products and services based on brand but on the quality of the supplier’s AI. “In other words, if a consumer knows a company’s goods or services are produced by purely traditional, human-based processes, they will perceive this as inferior quality and expect a poorer customer experience,” warns Pretorius.
The report strongly advises local companies to act now in developing their AI capabilities.
AI as mark of quality
AI is a computing technology that gives human-like intelligence and predictive decision-making to the products and processes into which it is incorporated. Although it is often referred to in the singular, AI is not a single intelligent entity. Like any piece of software, it is a program that either runs on a device, like a computer, robot or smartphone, or provides information to that device, to perform specialised functions. These can be simple tasks, like understanding human speech to activate a function on your phone, or complex ones, like analyzing products and manufacturing processes to improve them dramatically. An important attribute of an AI is that it can continue to learn as it gathers new data, improving on previous attempts and even surpassing human performance.
The speed and accuracy with which an AI-enabled manufacturing process can turn out products as well as the efficiency realised when AI-enabled systems coordinate their efforts will soon amount to better quality in the eyes of consumers.
Data is critical
Before being used for complex tasks in a production environment, AI systems must first be trained. This means they are given data that is appropriate to their task and from which they can learn how it must be performed. They then train iteratively and repeatedly until they achieve an acceptable rate of accuracy. Some AI systems come pre-trained in simple tasks but for more specialised operations, they must be trained from scratch.
Without the right quantity, structure and quality of training data, an AI system will fail to learn. While programming and training an AI system can take as little as a few hours, getting the right data can take months. “Creating sufficient data that is meaningful and void of confusing information is a specialised task, requiring experience in data science,” says Pretorius. “Many companies will not immediately have this data on hand and will need this expertise to construct it over time.”
So if companies want to implement AI, they will need to start off by accumulating precise training data with the assistance of a competent data scientist.
Beyond the data
According to Pretorius, AI projects are like any other enterprise technology initiative. They should not be attempted in isolation but must be considered in relation to the business as a whole. Companies should therefore seek out a partner that combines technology strategy and data science services. “This will ensure their solution is holistic, being both technically effective and aligning with their total strategic vision,” says Pretorius.