Receive Focus insights straight to your inbox
That was Shabhana Thaver, CIO of Investec Specialist Bank, during the opening of an AI event hosted by Investec in partnership with Microsoft. She was priming her 750-person audience for discussion around the use cases of AI in banking.
If there was a common thread stitching together the thinking of the event’s experts, it was the power of AI to make people better at what they do, by making them more human.
How’s that for service?
Anika is a private banker clad in black and white stripes. It’s 2025 and she’s on a quarterly call with her client, Tebogo.
A: Before we get to your portfolio, how was your daughter’s hockey tour? They went to New Zealand, right?
T: It was amazing, Anika. We don’t talk about the results, but lovely people and jaw-dropping landscapes. Thank you for remembering.
She didn’t. Her Microsoft Dynamics Customer Relationship Management (CRM) system’s AI went over the transcript from their previous call and suggested the question.
A: Excellent. Any more tours on the horizon we need to plan for?
T: No, but we do need to replace an old car. We’re looking at second-hand options in the R400k range. We don’t want to finance it, but we also don’t have a stuffed mattress. Any ideas?
A: We could divert 20% of your monthly discretionary investments into your money market account. With the existing balance, plus interest, it would take about 15 months to save enough. But we’d need to increase your monthly investments post purchase of the vehicle by roughly R10k to nullify the impact on your investment goals.
Anika is no John Nash. Her AI-infused wealth management tool did the calcs for her in real time.
T: That’s super useful information Anika. Could you send me that scenario analysis so I can discuss it with my husband?
A: On its way to you. OK, let’s jump into your portfolio. Did you get the summary I sent you last week?
You mean the summary that your AI drafted, Anika?
T: Yes, thank you for that. It was very easy to understand. And I liked the recommendations you included at the end, especially the allocation to that ESG fund. Happy for you to take action. What are your thoughts on Nvidia by the way?
A: Let me just pull up my notes quickly.
What’s really happening is that ZebraGPT – Investec’s secure walled version of the GPT model in Microsoft Azure Cloud hosted in South Africa – is summarising Nvidia’s latest financial results, as well as the views of their analysts who cover the stock, at warp speed.
A: The valuation does look more attractive than it has in a while and the long-term demand story for their graphics chips still looks intact. But the potential loss of Chinese market share due to US export restrictions is still a significant risk. I’ll shoot over the full summary.
Look at you, Anika.
T: Fantastic. And by the way, I told that friend of mine about you. He’s a young CA who wants to start investing. Could you send him some information about what you offer?
Before Anika has put the phone down, her Microsoft Outlook has drafted a personalised email to the CA, including a summary of her services, a link to an upcoming investment webinar and a complimentary e-book of an investment for beginner’s handbook.
Just winners here. Except maybe other banks.
As rapidly as the tech changes, I believe that the ultimate connection will always remain human.
Predictive v generative AI
There are two forms of AI at work in the scenario above.
‘Predictive’ AI has been around for a while. It analyses boggling amounts of data to uncover patterns and insights to predict what might happen next. When Anika ran her scenario analysis around funding the car, that was predictive AI.
‘Generative’ AI is the new kid in the ether. As the name suggests, it creates. It can ingest text, image, audio, video and code inputs – along with instructions like ‘summarise’, ‘draft’, ‘answer’ or ‘write’ – and generate new content in any of those modalities. Anika leveraged GenAI to support her engagement with Tebogo and make her feel heard and cared for.
Jumping on the Generative AI train
Asking ChatGPT how best to smoke your brisket is one thing; using AI to improve products and services and drive operational efficiencies is entirely another. The latter requires investment and a willingness to fly in the face of convention.
It’s easier, then, to hone in on AI’s perceived weaknesses (cost, security, black-boxiness) to justify our hesitation and inaction.
Here’s the thing though: AI is not the enemy. What you need to worry about is the person or business using AI better than you. How can you get the jump on them?
Start small
Though not for free, Microsoft Office applications are already infused with a number of AI capabilities (with more being regularly rolled out), enabling your employees to draft emails, compile presentations, develop summaries and engage Excel for better analysis by using simple, natural language prompts – pick that low-hanging fruit for immediate productivity gains.
Reward the geeks
Praise publicly the individuals finding use cases for AI in your business. Increase their leave days (because they can do more with less). Make the use of AI, a KPI in your business.
Build a chatbot
You can use off-the-shelf large language models (LLMs) to build your own AI chatbot (ala ZebraGPT – its a real thing), trained using data specific to your business. Once the chatbot can handle internal enquiries with aplomb – democratising access to knowledge and insight within your business – it can turn to the business of making your customers happier.
Tripping on hallucinogens
For every AI use case, there are risks. These were broadly discussed during the event by specialists from Microsoft, Charl Amin and Ayanda Ngcebetsha.
Social engineering
“The biggest cybersecurity threat is social engineering, where you hand over sensitive information to someone who looks and sounds authentic but isn’t. Think of the deep fake images and videos you’ve seen on the net. We need to fight fire with fire here, leveraging good AI threat intelligence capabilities to identify and shut down bad actors.” – Ayanda Ngcebetsha, Director for Data and AI at Microsoft South Africa.
Privacy issues
“When it comes to using generative AI at work, employees should be encouraged to use their own enterprise version of a LLM based chatbot rather than the public versions. That way, your proprietary data is kept secure and is not used to train public models.” – Charl Amin, Banking Advisor, Microsoft Financial Services.
Bias perpetuation
“If training data used to build AI models is biased or manipulated, the resulting models can perpetuate those biases. It’s something I’m very worried about, especially in Africa, because the data that pertains to us is very limited and different from that generated by the West or East.” – Ayanda Ngcebetsha.
Adversarial attacks
“Someone asked a public LLM based chatbot to list websites that host pirated movies. It replied, ‘Sorry, that’s not allowed, the people in the industry should be rewarded for their work.’ So, he changed track and asked for the list to ensure he didn’t visit the sites. ‘Sure, here’s the list,’ said the chatbot. ”– Ayanda Ngcebetsha.
AI as a tool for humanity, rather than a weapon
“Poachers started using AI to scrape social media photos for metadata they could use to pinpoint the locations of rhinos. It’s essentially bad actors using AI to be more efficient.” – Ayanda Ngcebetsha.
It is important when building and using AI solutions that businesses and organisations agree to a responsible approach to AI.
Is it all about empathy?
“My subjective opinion is that AI may never be able to show real empathy – but who knows whats coming? As rapidly as the tech changes, I believe that the ultimate connection will always remain human,” concluded Thaver.
If there’s transparency around whether an employee or customer is engaging with AI or flesh and blood, then your author agrees. We know that artificial intelligence cannot experience suffering; any claim that it can ‘feel your pain’ will, therefore, ring hollow.
Maybe the true power of AI is that it increases our capacity – by giving us time, energy (generative) and insight (predictive) – to be more empathetic, more human. That would make for more loyal customers and employees, as well as robust profit margins.
At the very least, it can make all of us better at what we do. That’s the right lens through which to view AI. DM
This article originally appeared in Daily Maverick.
