Safety has always been easy to understand in a Volvo, making its case in plain sight with steel and seatbelts. Safety in AI calls for a more nuanced understanding, because the guardrails are invisible and the consequences of poor decisions are not always obvious.
It was fascinating, then, to see the safety imperatives of the two worlds come together last week, as the unstated theme of a key session at NTT Data’s Horizons Innovation Summit in Milan. While NTT is a Japanese organisation, the NTT Data division has its roots in South Africa, with a genesis in Dimension Data in Johannesburg. Today it is one of the top IT services groups in the world, and one of the few providers of a “full stack” of enterprise-scale AI integration.
Bill Wilson, global head of AI for government at NTT Data, was joined on stage at the Summit by Ian Thomas, Volvo Cars’ global head of strategic initiatives for data, AI and intelligent automation, to explore the role of trust at the intersection of the physical and the digital worlds.
Thomas did not have to invoke the Volvo legacy, as Wilson had a very personal history with it: “Growing up, our family car was a Volvo T40,” he said. “that Volvo went on to save my father’s life.”
Thomas responded that it was the kind of story he heard frequently: “There’s an amazing emotional connection between Volvo Cars, brand and safety. But Volvo has a very strong affiliation with performance as well. That performance shows up differently today. It shows up still under the hood, of course, but it shows up also in the software, the data and the intelligence in the vehicle itself.”
This meant, he said, that. Volvo was not treating AI as a separate layer pasted onto the business. It was part of the same design philosophy that made safety central to the brand in the first place.
“It’s safer, it’s smarter and it’s more personalised than ever before. It’s safer because we’re using AI technology to allow the car to understand its environment. This gives it context. It’s smarter because we’re then using AI technology to take decisions based on that context, and that decision will sometimes support the driver and sometimes intercede in a critical safety situation, And it’s more personalised through the intelligent assistants that are supporting the driver in car, but also as part of a broader connected car experience.”
This also define where the risk lies. Once a machine is reading its surroundings, forming a judgement and acting on it, safety becomes a software issue as much as a physical one.
“I genuinely don’t think about technology when I think about AI. I think about AI as a step change in how we can leverage knowledge at massive scale. Never before have we been able to process so much data, identify patterns, discern meaning from context, inform decisions and even take actions in workflows. What that really means to me is that we are now compressing the time from an idea to a real world impact.
“It also raises a really important question, and that is, how do we balance the speed of innovation that’s now available to us, with the responsibility and the trust that people place in our brand.”
That question lies at the heart of every corporate AI strategy, even if executives do not realise it. Volvo’s case, it carries extra force because the brand has spent decades persuading people to trust its judgement in life-and-death situations.
Thomas was refreshingly frank about where value is coming from.
“We’re yet to discover a single application of AI that delivers huge value. Rather, we are seeing the compound effect, where across a process or across a value stream lots of improvements through AI add up to give us true value.”
No hunt for a killer app, then? Well, maybe.
“One of the most interesting areas that we’re looking at is in engineering. Our software engineering teams are leveraging AI assisted tools to write code, perform testing and to improve documentation. Our hardware engineering teams are looking at how best to optimise designs by looking at various different features, such as materials, weight, cost, sustainability and, of course, mechanical engineering performance. The power of that is that we’re able to assess options and variations on design much earlier in the design life cycle, so we actually produce much better products and better outputs.
“The advent of AI assisted tools allows designers to sketch concepts and play with variations on design. We’re seeing the compression of timelines from months to days in the early, conceptual part of the design phase. Of course, the real power is going to come when we plug these AI design tools into engineering data, when we’re able to reference materials libraries, and reference design standards on the fly within the workflow. Then we start to see the agentic patterns emerge.”
The same logic applies in supply chain, where the stakes are not conceptual at all.
“The modern consumer vehicle is comprised of tens of thousands of different technological components and software components. Those are supplied by hundreds of companies or suppliers around the world. It’s very complex, trying to understand the supplier resiliency. The geopolitical climate means that we’re seeing huge disruption to supply chain and logistics.
“We are undoubtedly currently reactive. There is an opportunity to start looking at our planning data, to start looking at our supplier data, and to start looking at geopolitical indicators and supply chain signals, so that it becomes an adaptive process.”
That discussion is especially relevant beyond the car industry. Any business operating across volatile markets, uncertain logistics and fragile supplier relationships will recognise the appeal of moving from reactive to adaptive.
The challenge is one NTT Data is almost uniquely positioned to solve, thanks to its full stack approach.
“Data is the single greatest limiting factor to the application of AI,” said Thomas. “We’re asking agents to act within context, and what that means is we really have to understand the context of an organisation. So we’re after enterprise level understanding now, and so much of that is tacit. It’s still in people’s minds, as opposed to any kind of structured documentation.”
That may be the most practical warning of all. The limiting factor is the messy, partial, unstructured state of knowledge inside the business. AI can only act responsibly when context has been captured well enough to guide it.
Thomas also put responsible scaling in plain terms: “We need to remain compliant, so we build guardrails that allow us to move at pace without breaking things as we go. We’re also balancing people and technology. Remember that AI is about augmenting what people do today. It’s about making them more effective.
“Safety of people in the physical world has always been something Volvo has designed its products around. As we now design the agentic systems of the future, we’re trying to take that thinking around safety and trust into the digital world and the AI age.”
* Arthur Goldstuck is CEO of World Wide Worx, editor-in-chief of Gadget.co.za, and author of “The Hitchhiker’s Guide to AI – The African Edge”.