South African companies can start realising significant business benefits now by utilising the high-performance computing capabilities of the locally available Microsoft Azure data centres – even if they haven’t migrated any data to the cloud yet.
By Paul Morgan, business unit lead: data, planning and analytics at Altron Karabina
There is a range of capabilities available in Azure that simply do not exist in an on-premise environment at an equivalent price point – such as enterprise scale Artificial Intelligence.
An organisation can easily get access to high-end computational services, such as Azure Databricks (an Apache Spark-based analytics platform), capable of running complex machine learning scripts at incredibly high speeds that would be cost-prohibitive using on-premise systems.
Furthermore, these services can be turned on and off whenever required. The client therefore only pays for usage, and there are no expenses associated with servers sitting idle.
However, although it’s not necessary to have the bulk of organisational data in the cloud for this type of activity, there will be costs associated with continually transferring terabytes into Azure for computation. At Altron Karabina, we therefore believe that the time is right to start moving corporate data into the local Azure cloud permanently – the platform has definitely proved itself.
The likes of AI, automation, real-time data analysis, and other sophisticated technologies available through local Azure data centres will definitely make it easier for companies to embrace innovation.
Today, decision-makers have a variety of options available to them to extract business value. While it is possible for organisations to develop solutions completely from scratch, they can also use pre-built Azure services to add value. For example, Azure Cognitive Services are able to provide pre-trained models for facial recognition, text sentiment analysis and speech-to-text services, amongst others. A facial recognition capability can be built into a standard office application by a developer who has absolutely no idea of the deep neural networks that are required to make an AI service work. This allows innovative ideas to be delivered, without needing to employ all the skills needed to create them.
This means that companies can start embracing Azure now for competitive advantage. They don’t need to procure expensive software or hardware, and they don’t need to employ huge teams of AI specialists to start delivering innovative solutions now.
Putting into practice
A local example of this type of innovation thinking was recently deployed at one of Altron Karabina’s clients in the insurance sector. The business required a way to fast track call-centre emails where the customer’s tone and language indicated frustration or annoyance.
Using Azure Cognitive Services combined with robotic process automation (RPA), Altron Karabina worked with Microsoft and the client to automate the process as emails arrive. In essence, the process identifies the probability of customer being highly upset, based on their email content, and flags it accordingly.
The business also wanted to classify claims based on attached images. Azure Cognitive Services Image Recognition was used in this case; if a claim came into the call centre with photographs attached of a damaged car, rules were applied to identify the picture and push it through the vehicle claims workflow mechanism.
Even if a company is hesitant to migrate all its data to the cloud, it is possible to start exploring a few aspects of Azure’s high-performance capabilities and what the impact can be to the organisation.
Even at an individual level, a person could use Azure Databricks to analyse their personal credit card payments and sort them into different categories – at a limited cost. The cloud can therefore work effectively for tiny things or a very complicated business process.
This year, the expectation is that more companies and people will start using AI for high-innovation projects. All of this is possible through high-performance computing, available locally now in Azure.