Since ChatGPT was released to the world at the end of last year, AI has become the watchword of business conversation. The question being discussed at length online and at virtually every industry conference is whether we are on the precipice of a technological wave not seen since the rise of the internet in the 1990s.
There is no doubt that ChatGPT is impressive. For most people, it represents a quantum leap from a point where they had virtually no conscious access to AI technology to one where they can ask any question and get a reasonably comprehensive answer. In some areas, it is more accurate than others, but compared to what came before, it is a transformational consumer product.
That said, the consumer product point here is key. Fundamentally, ChatGPT is a product built on the GPT-3 and GPT-4 Large Language Models. Large Language Models (LLMs) aren’t in themselves ‘new’. They are a technology that has been in use, almost exclusively at large companies, for many years.
Businesses have been training LLMs in that time on their own datasets to help them make sense of vast pools of information and find new insights. This includes the field of procurement. So, for a few pioneering large organizations and, indeed, procurement functions, ChatGPT may not have been quite the same quantum leap as it has been for consumers.
What has changed, though, is the power of LLMs and the ability to generate new content from seemingly unstructured data (generative AI). GPT-4, at the time of writing, is the most powerful LLM in the world. Whereas many of the private LLMs previously used by businesses have been trained on carefully curated databases, GPT-4 was trained on wholly unstructured data (the internet). The power of LLMs is only increasing – it is widely expected the GPT-4 will not hold the title of the world’s largest LLM for long.
Applying AI to procurement
What does this all mean for procurement and the potential applications of artificial intelligence in our industry? Hitherto, legacy ‘procuretech’, even that which is AI-enabled, has mostly enabled us to run defined processes a little faster or work with bigger volumes of structured data then conventionally possible via ‘Excel.’ Automation, ML and AI have been solutions looking for a process. A consequence of being a pioneer has often been the creation of highly bespoke models with configuration or cost barriers to change. Could these two limitations could be a thing of the past as AI-enabled supply chain technology becomes comfortable with unstructured data and democratized by design?
Today at least, the answer is yes, and no. Yes, because any user of ChatGPT or a comparable generative AI solution can create reasonable (but not exceptional) outputs faster and more comprehensively than a DIY approach. No, because getting to insight is often about specific data points which may lurk in paid or closed datasets, the answer to which is some level of integration, or a utility solution, both of which may only be available via subscription for now. This will price some users out of the ‘exceptional insights’ market, but not stop many rapidly creating reasonable outputs that may be a step change in their world.
An example? A significant procurement value add is the ability to understand markets, analysing data and searching for insights that allow us to take action – whether that be with cost, revenue, or purpose in mind or something else altogether. There are services and tools that can help with this, but it is also something that, with the right language model, generative AI can disrupt and be a perfect fit to help with. Further, a ‘democratized’ model and ‘low code, no code’ approach means that, theoretically, at least, user barriers can be broken down. Put simply, it’s easier for the end user, and it is not unrealistic to say that data analysis that would have taken days or weeks in the past will now be undertaken in minutes.
Another key role of procurement is making sense of what we find – setting up dashboards, writing plans, drafting reports, and the like. While AI may not completely replace this overnight, you may well find an AI-created report takes you 70% or 80% of the way to the finished product. The key to optimizing this is adding the right wrapper, enabling the user to define their own ‘prompts,’ questions, or instructions that will get the most out of the LLM to enable faster, more bespoke, and more insightful data interrogation and content production. We have already seen pioneers doing this, using AI to generate reports and further Intelligent Automation to take decisions based on the findings.
What the future holds
It’s easy to see how AI will make things faster, but will it create ways of doing things that are new or better? Likely yes, but we are in the early innings of AI. Perhaps the easiest way of viewing many of today’s emerging artificial intelligence capabilities is as productivity tools or as Microsoft has framed its new Office product; a ‘Copilot.’ In procurement terms, we can easily envisage our digital assistant at our beck and call.
Beyond speed and general productivity enhancement, there are three other areas where the impact of AI will be of particular interest to procurement. The first is in businesses that previously were unable to develop private LLMs because of the vast resource and costs associated with them. For those companies, AI will allow them to do things that previously they couldn’t, and that should have a very interesting impact on how those businesses operate and the competitive environment more broadly.
The second area is category disruption. We talk a lot about how procurement operates, but in some ways, the biggest change AI may have on procurement is by disrupting the status quo within the existing categories we manage. If you’re a marketing category specialist, how will AI change the way marketing is created, how it’s placed, and how it’s consumed by audiences? AI has the potential to completely upend the way things are done in the supply markets that we are experts on.
Finally, we must consider how prepared we are for a supply market where suppliers are able to use AI to better project their services and respond to our questions. Just as we will be better armed, so to the suppliers that we bring into our organizations. Some may argue that formal tender environments have long since been a creative writing competition. Still, AI’s ability to answer everything to a reasonably high standard may well narrow the playing field in a discipline that has in the past been seen as core to decision-making, particularly in the public sector.
AI Ethics will be a big topic, and it is reasonable to suggest that a vendor may be asked how AI has been used in a bid response. We may also consider its role in the products and services that we buy, the AI investments and roadmaps of our suppliers, or how a digital account manager could be deployed during a relationship, for example. In a world where development is so fast, what we value in our suppliers may evolve quickly, and the importance of partnering, relationships, and understanding the difference between buying commodities and solutions will come to the fore.
Given all this, I think there is an opportunity for procurement professionals to feel optimistic about what artificial intelligence will bring to our sector. Just as the internet changed the way we do business, AI promises to do the same. Allowing AI to take on routine work faster and more efficiently will almost certainly allow procurement teams to spend more time on work that adds value – that is something we should embrace, but also an opportunity for us to consider just how and where we will add the most value in a world redefined.
