The past few decades have seen explosive growth in data, artificial intelligence (AI) and machine learning (ML) technologies, with the first AI programs actually written in the 1950s. However, despite innovative algorithms, these were unable to unlock truly scalable business value.
In effect, one could compare data science at scale to the retail industry. Slim product margins in the retail industry mean sustainability is only achievable through high volumes of sales. Similarly, the value of data science – where AI and ML algorithms are used to solve a business problem – grows exponentially if it can be executed at scale.
Hennie Fouche, MD of TrueNorth Group, emphasises that to unlock the full potential of data science, it is crucial to start the process by selecting a scalable use case. Unfortunately, this critical step is often disregarded, leading to wasted efforts on impractical or underwhelming outcomes that fail to meet expectations. Data scientists tend to operate at the intersection of mathematics, domain expertise and computer engineering. Therefore, they must be able to articulate complex problems in the context of both understandable terms and business value.
"Data is often likened to a company's 'lifeblood', drawing a compelling parallel between the necessity of food and water for humans to function optimally and the indispensability of data for AI algorithms. It is crucial, however, that this data is of high quality, for just as feeding garbage into the system yields garbage in return, the output is only as reliable as the input.
“Thus, the more varied the types of data points used, the better the model performs. It is really about the additional touch points one can access that lie outside the standard internal data that a business may analyse. The more types of data you can source to use, the more accurate and effective the model is.
“Data science has always been about leveraging the information to learn something about your business, and data is only going to become more important in the digital era. After all, digitisation has given us the ability to capture data at scale, so what is now required is to have the relevant AI tools to extract the value from this volume of information,” he says.
He points out that a great example of executing data science at scale is Google’s efforts. Research indicates that just in 2020, the business undertook 600 000 experiments for changes to its search algorithms. Roughly 4 500 improvements were made in that same year – equating to roughly 18 improvements per day, annually.
“In a similar manner to the principles of DevOps, MLOps provides a framework for continuous delivery of improvements. The implementation of continuous integration and continuous delivery(CI/CD) approach in machine learning, along with the right people, processes and technology, can lead to significant value creation. This is because faster delivery is enabled, releases are better governed and proper testing is undertaken.
“It is also worth noting that whereas, until recently, AI was very much the preserve of the large enterprise, due to the need for a big data platform to support it, SMEs can now also benefit from this technology. Thanks to the cloud, smaller entities can spin up on demand the requirements they need to implement an AI-based analysis,” he continues.
“In fact, SMEs may even find they have a huge opportunity to leapfrog the larger organisations. After all, even advanced enterprises inevitably remain stuck with a lot of legacy equipment and environments, so it is much harder to be as agile as a smaller company unencumbered by such concerns.”
Ultimately, he says, it is clear that maximising the potential of data can undoubtedly unlock greater business value, as well as unveil novel new opportunities. Research has consistently shown that companies that adopt a data-driven approach to decision-making also tend to achieve greater long-term profitability.
“Clearly, aligning with a trusted partner, one that has strong capabilities in driving scale operations in ML, knows how to set up the necessary infrastructure and processes and can help you unpack your business case, is key. In fact, it is the best way for companies of all sizes to effectively leverage AI and drive their strategic objectives forward,” concludes Fouche.
Share