Sasol shares a case study with zero investment, and a potentially massive impact.
Attendees at the 2023 CIO Day heard how Sasol is using generative AI to innovate and accelerate their business operations – all without spending any money.
The AI journey started when Sasol Group CIO, Lungile Mginqi, received a phone call while he was on holiday. At the time, ChatGPT had only just appeared on the world’s radar, and the caller urged him to investigate this new technology – or he’d face being left behind.
Lungile took the caller’s advice to heart, looked into ChatGPT, and quickly realised he wanted to drive generative AI at Sasol. “We needed to get in front of this. So we came up with a plan in terms of what we’ll do and when, and how to put in guardrails,” he said.
He added that he believes by 2024, the tech industry will be “swimming in Copilot”, and simply using AI will no longer be a differentiator – it’s how organisations use it that’ll determine the difference between success and failure.
Experimentation and identifying use cases
Phase one of the project focused on experimentation, along with a timeline that required the identification of 10 use cases. Bramley Maetsa, Sasol’s innovation projects and delivery enablement lead, elaborated by saying that at this early stage, it was important to mobilise the organisation and undertake extensive value mapping.
Bramley helped get people involved by identifying internal “change enthusiasts” – a small group of IT employees interested in generative AI, who would be keen to help with development during their spare time (at this early stage, the project had no investment costs attached).
Bramley pointed out that they received a lot of interest by simply sending an email stating that Sasol embraced the execution of AI, and if anyone was interested in joining this innovation team, they should hit reply. Thus far the team, comprising enthusiastic developers assigned at minimum capacity, has developed 10 to 30 use cases.
In October, phase two is planned to kick in, and in between Bramley and Lungile are investigating where they can drive significant value. Only once the first phase delivers, will money be put into the project. The hope is that by 2024 the generative AI tech will be integrated, and Bramley emphasises it will have a “significant impact”.
’We won’t solve AI-related problems, but we can mitigate them’
Bramley and Lungile noted that, with the development of this tech, they see potential value in several use cases particular to their industry and organisation. Examples include training an AI model to extract data during email processing (and automatically notify buyers if metadata is missing, which will save salespeople time), automating translations (that, up to now, have been outsourced), drawing up contracts (by training an AI model to implement a legal documentation checklist), product recommendations, and much more.
They admitted there have been stumbling blocks along the way, making it critical to have a clear business framework in place. For example, it was initially assumed that the project’s generative AI sandbox would “just work”, but the tenant’s instance they used didn’t have ChatGPT 3.5 Turbo. This meant they had to move to an instance located elsewhere – now they’re using GPT4 with semantic search and Chroma under the hood to automate, which is aligned to scaling the potential value this tech can offer the business.
Lungile added they’ve experienced AI hallucinations, and he’s all too aware of the potential for ethics problems and risks. “We wanted to scale quickly, but due to the demos we worked on, we realised we needed to wait for a clear and well-defined AI policy. We’re under no illusion that we’ll solve AI-related problems, but we can mitigate them,” he commented.
“The thinking behind the business framework will be key for us to drive value. I’m excited about predictive analytics combined with AI to become a catalyst in that process,” concluded Lungile.
CIO Day was made possible thanks to the support of CIO South Africa's Principal partners EOH, Makwa IT, and Liquid Intelligent Technologies. Executive partners were BCX and Workday, and Associate partners were Communication Genetics, OneConnect, iiDENTIFii, Perpetuuiti, and SAS.