Ritesh Kumar, Director of Procurement and Supply Chain Intelligence at The Smart Cube, examines the hesitancy towards adopting large language AI models and the benefits that they can bring to the workplace.
Over the past six months, the deployment of large language models (LLMs) has become more prevalent across a range of industries. From financial services and telecommunications to healthcare and media, LLMs are being used to develop and improve new Artificial Intelligence (AI) tools that are capable of handling complex queries in business operations.
As more companies realise the benefits of using LLMs for work process automation, demand for generative AI solutions is likely to increase.
Nevertheless, procurement teams across industries have been hesitant to adopt large language AI models in their mainstream processes. This has been due to concerns ranging from the absence of established procurement use cases, high operational and maintenance costs, security and privacy worries over data dissemination and access, or a lack of LLMs’ contextual understanding, which can make AI tools unreliable for strategic decision making.
However, organisations are increasingly adopting generative AI. In fact, an April 2023 Writer’s survey reported that nearly one in five companies use five or more generative AI tools – and their procurement teams are also beginning to invest in advanced digital solutions.
A recent US survey by Arize AI (a Machine Learning observability platform provider) also noted that over half of its respondents were looking to utilise LLM applications in the next 12 months – so, the significant opportunity is clear.
Use of large language AI models by procurement teams
While the use of large language tools in the procurement function is not as common as in other areas, such as content creation and marketing, firms have started to use natural language processing (NLP) tools, such as chatbots and virtual assistants. These are powered by LLMs to assist with procurement tasks in order to streamline processes, reduce costs, and make better-informed decisions.
One procurement function which companies are using LLMs to assist with is vendor selection. These models help procurement teams to analyse vendor information and determine which vendor would be the best fit for their organisation. This is based on factors including price, quality, and delivery times.
As well as this, procurement professionals can use LLM tools to analyse spend data in order to identify trends and patterns. For example, GlaxoSmithKline used the IBM Watson platform to analyse procurement data from across its global operations.
This helped the firm to identify opportunities to negotiate better deals with suppliers and optimise procurement processes, resulting in significant cost savings.
Further, large language tools can help procurement professionals manage supplier relationships more effectively – analysing data and providing insights on supplier performance, such as quality of goods, delivery times and responsiveness to customer needs.
For instance, Amazon uses an NLP model to identify patterns in customer complaints, such as delivery delays or product defects, which alerts Amazon’s procurement team. The team is then able to take action with the supplier.
There are multiple additional use cases which are increasingly being explored by procurement teams. For instance, these tools are used for streamlining supply chains by leveraging scenario modelling, generating accurate demand forecasts by leveraging historic training data, and optimising inventory levels by feeding training data rich in demand patterns or seasonality.
Further use cases include performing a holistic risk assessment to aid improved procurement decision making, as well as utilising these tools for a more accurate spend classification and analysis to drive strategic decision making.
Drivers behind organisations’ use of LLM tools
One of the key factors behind organisations’ use of LLM tools in procurement is that it improves efficiency. AI tools can help automate manual tasks such as data entry and analysis. This can save procurement professionals a significant amount of time and allow them to focus on more strategic tasks.
Elsewhere, LLM tools can support decision making. By using AI to analyse large volumes of data, procurement professionals can gain insights into market trends, supplier performance and other key factors that can help inform procurement decisions.
Additionally, AI tools can help identify potential risks such as supplier fraud, contract non-compliance, and supply chain disruptions, allowing procurement professionals to take proactive measures to mitigate these risks, in addition to enhancing supplier relationships and identifying cost-saving opportunities.
Further to this, LLM’s can help identify cost-saving opportunities, such as identifying suppliers with lower prices, negotiating better contracts and reducing waste.
Concerns amongst procurement professionals
Currently, there are several barriers to the large-scale adoption of large language AI models. This is because AI is still in its development phase.
One of the major current concerns procurement teams have regarding the adoption of large language AI models is that accuracy is not always guaranteed. They can generate misleading and biased information or make mistakes in contracts, negotiations or other important documents.
The output of any generative AI model depends on the prompt provided by a user. A misleading prompt could produce inaccurate results. Therefore, extensive testing and training is needed to obtain high-quality outputs.
Furthermore, procurement professionals also have security concerns regarding large language AI models. AI systems can be vulnerable to cyberattacks, which could put an organisation’s sensitive procurement data at risk of breach.
Tools like ChatGPT are built without any real corporate privacy governance frameworks, making it challenging for companies to leverage these models in their chatbots.
Additionally, some procurement professionals are hesitant to adopt large language tools because LLM tools lack the contextual understanding and decision-making abilities that humans possess.
Procurement professionals need to consider factors such as existing supplier relationships, pricing negotiations, contract terms or any other external forces affecting supply at the time of the negotiation of contracts. These factors require human judgment and expertise, which cannot be fully replicated by an AI system.
For example, an AI system may reject a supplier that does not meet its stringent criteria in one area, while overlooking the supplier’s reliability overall.
However, this should not discourage the use of large language AI models within procurement. Instead, procurement teams should consider using AI and human intelligence (HI) in conjunction with one another.
For example, a buyer can establish a team of experts to review generative AI outputs or intervene in the generative AI process whenever necessary, adding valuable contextual insights and judgement.
Combining AI and HI
When it comes to transactional procurement (P2P) functions like invoicing, contract management, and accounts payable processes, there is significant scope for these activities to be undertaken by AI, as they are repetitive.
However, in terms of more strategic tasks – such as category strategy development, business requirement gathering, and supplier management – AI should be utilised to assist and accelerate human decision making, as opposed to replacing it.
If AI is harnessed correctly, procurement specialists will be able to spend a greater amount of time focusing on tasks which cannot be automated, and which require some form of HI and emotional intelligence. Individuals will be able to shift their attention towards fostering relationships with stakeholders and suppliers, developing these vital connections.
Technology isn’t there to take the place of humans – it is meant to enable more time for us to do our jobs better. As such, the optimum approach towards AI will be about discovering the best use cases for it in tandem with HI – in the specific contexts of an organisation – then striking the right balance between the pair so as to best meet the business objectives.
