AI has quickly become universal, but genuinely “intelligent” software is still some way off. Data scientists and digital ethicists talk about the “AI singularity”, where software innovates without human intervention. And that is still years away.
But systems based around complex algorithms and machine learning—the building blocks of AI—are everywhere today. They’re in the smart assistant on your phone, online chatbots, facial recognition, and self-driving cars. And as practical applications for advanced AI—deep learning and neural networks, for example—emerge, the technology will become even more deeply embedded in decision-making and operations.
These AIs all share one thing in common: they need to work hand in hand with data. And there are two ways in which organisations can approach that partnership. Either they can feed data—pre-existing and newly generated—to AI systems as an additional process, or they can intrinsically link processes to AI systems, designing data flows around them.
Both approaches have consequences for finance and supply chain leaders’ roles and responsibilities. And in both cases, measuring and optimizing how data is gathered, what the “supply chain” for that data looks like, and its impact on the organisation is crucial.
So why does AI demand a renewed focus on data? After all, big data and analytics have been revolutionising decisions for years. The reason is the big AI efficiency gain: autonomy. To let AI systems make decisions, we have to trust that they understand what’s going on.
Take something as simple as an online chatbot for consumer queries. One study estimated that a system would need 2 million answers paired with 200,000 questions to be able to satisfy users. Researchers presenting to the Association for the Advancement of Artificial Intelligence (AAAI) used more than 600,000 tweets just to compile a machine-generated analysis of sentiment in 30 Twitter hashtags.
The more training data AI systems are given, the better they perform. In many cases that means feeding them huge pre-existing datasets and then watching how the algorithms perform. But equally important is how they evolve based on what they learn from the data.
AI systems for mission-critical processes in a corporate environment—much more complex than Twitter hashtags—will require vast amounts of information to be trained. But to apply their learning, they need a consistent flow of real-time data gathered, categorised, processed, and reassembled to produce reliable outputs.
This is the digital supply chain.
Both disciplines must also ensure the flow of data through the organisation is seamless for the AI approaches being used. If the digital supply chain is fragmented or fractured, even the most modest AI is going to struggle to meet the expectations set for it.
Cloud makes all the difference. When organisational data is hyper-available on a consistent platform to the right systems—when there’s a genuine real-time digital supply chain for AI systems to use—the result should be new efficiency and effectiveness across many core business processes.
Better yet, cloud ERP vendors can offer easy-to-use, AI-infused ERP applications off the shelf, trained on industry standard data or tuned to best practice. Combining a modest internal digital supply chain with powerful models pre-trained on external data offers a relatively simple way into AI.
SCM and finance leaders’ roles will change in different ways. With AI handling the routine processes, the job becomes exception remediation, problem-solving, and decision-making instead of ensuring data is up to date; rather than regular maintenance, it’s about failure maintenance.
Getting to the start line requires a shift in mindset. When Birmingham University realised that digital native students were coming into the organisation expecting seamless services, the team knew they had to redesign their digital supply chain.
They discovered they could plan for services that would really differentiate their student experience both now and in the future by adapting their digital supply chain to templates offered by cloud ERP vendors. It meant asking operational parts of the university about desired outcomes—not existing process maps or data flows. Finance and SCM could look across those outcomes to build a coordinated, integrated digital supply chain fit for the next generation of AI.
As the University realised, cloud means that essential AI processes – that might drain the computing power of on-premise systems—can be applied to your own data. Cloud can deliver “data enrichment as a service”. It enables the automation of data pipelines to train AI and apply algorithms to ongoing data streams. It means being able to deploy self-training models as required. Security and management is baked in.
Fully exploiting your digital supply chain means building something unique—and cloud means doing so with a big head start, not reinventing the wheel.
Within five years, out-of-the-box, ERP-based AI will be married to proprietary systems built with cloud-based tools to deliver huge efficiencies, new products, and better customer experience. And many of us will care more about the digital supply chains that feed our AIs than the physical ones that feed our warehouses. Is your SCM function ready?