GenAI has immense potential – it can help people in a huge range of roles.
AI in Procurement
How GenAI cuts through the complexity
Generative AI powers a new generation of technology tools that will transform almost every aspect of how today’s procurement departments operate.
Generative artificial intelligence (GenAI) has quickly turned into the hottest business technology topic in the world. Less than three years after the first publicly accessible GenAI tool, ChatGPT appeared, there is an overwhelming appetite among the world’s leading companies to understand and harness the promise of GenAI. Growing interest has been reflected in the share price performance of Nvidia, the leading maker of the chips used to run AI models: its shares rose nearly 240 % during 2023 and by mid-June 2024 had gained another 190 %, making Nvidia briefly the world’s most valuable company.
As these gains suggest, GenAI has immense potential. It can help people in a huge range of roles – from call handlers to biochemists – to do more, go faster and achieve higher quality results. A 2023 joint study between Harvard Business School and Boston Consulting Group (BCG), examined the impact of using Open AI’s GPT4 across 18 knowledge-intensive tasks carried out by consultants, including idea generation, text writing and market analysis.
The study concluded that “consultants using AI were significantly more productive”. It found that they completed 12.2% more tasks on average and completed them 25.1% more quickly. The group using AI also produced results of more than 40 % higher quality than the control group. Gen AI was not suited for every task, but in areas where it was most relevant, it produced striking improvements in productivity as well as significant performance gains among less skilled individuals.
Benefits like these are especially relevant to procurement and supply chain management, where many processes currently involve laborious manual operations and resilience depends on the ability to navigate complexity and manage multiple risks.
GenAI has immense potential – it can help people in a huge range of roles.
In procurement teams as in many other departments, GenAI can provide a significant productivity gain by enabling people to complete everyday tasks around 25 % faster and more accurately. These might include drafting emails and documents, researching information, summarizing documents, producing translations or answering employees’ work-related questions. However, this layer of “everyday efficiency gains” is only the beginning. Beyond these more general benefits, GenAI has numerous potential specialist applications in procurement. According to BCG research, beyond the 44 % of procurement tasks that are already automated – a further 38 % can be enhanced with new AI solutions including GenAI.
The analysis found particularly large benefits in areas such as using predictive analytics and risk models to enhance supplier selection and generating documents to accelerate RFP and contracting processes. AI tools are particularly well-suited to produce more accurate demand and inventory forecasts based on quantitative input. By understanding the underlying context and providing know-how for decision making, GenAI is suited to propose solutions, for example optimized route planning to improve logistics efficiency, enhancing supply chain transparency and traceability through real-time tracking and managing risks through scenario modeling and analysis of supply chain resilience. In areas such as running negotiations with suppliers or calculating the effects of input cost inflation and finding ways to protect margins, GenAI can also bring major benefits, thanks to its capacity to analyze very large and complex data sets and evaluate the many options available.
BCG’s analysis also suggested that procurement teams could capture significant value in indirect spend from external service providers by using GenAI in areas including software development, creation of marketing content, professional services, recruitment and training, and customer service operations. The greatest gains from using GenAI are likely to be seen in software development spend, where between 23 % and 45 % of current spending could be saved within the next three years, and in marketing creation (22 % – 42 %). In customer services, it estimates potential savings at between 18 % and 38 %. Klarna, the buy-now-pay-later consumer finance company, provides a striking example of the potential benefits. Early 2024, it revealed that its GenAI-powered assistant was handling two-thirds of customer service inquiries in 23 markets. The assistant was doing the work of 700 full-time agents, resolving cases in less than two minutes compared to 11 minutes previously and earning satisfaction ratings on a par with human agents. Klarna estimated that this would add $40m to its profits in 2024.
AI tools are particularly well-suited to produce more accurate demand and inventory forecasts based on quantitative input.
Faster content creation of everyday marketing activities
Reduction in total consulting time within call-centers
Productivity increase for software developers using GenAI enabled coding
Efficiency gains from GenAI-enabled workflows for B2B info service provider
As this analysis suggests, GenAI represents a powerful and versatile set of tools that procurement teams can use to automate and streamline many aspects of the procurement process.
In areas such as preparing tender documents, a GenAI-powered automated assistant can help teams collaborate on the drafting and reviewing of tender documents, including checking the functional and technical specifications. Because the model is pre-trained on tender documents that exemplify best practice, it can aid users by suggesting improvements to the documents. The ability of GenAI assistants to respond to “natural language” chat-based prompts makes these tools easy and intuitive to use.
In trials, GenAI-powered assistants have halved the time taken to prepare tender documents, resulting in a 10-hour efficiency gain per tender. Users also found the assistant improved the uniformity and quality of documents, enhancing the organization’s negotiating position with suppliers and meeting employee demands for smart digital tools to improve work processes.
Similarly, a GenAI assistant can be used to compare and evaluate offers received from different suppliers. The offers are uploaded to the GenAI model, enabling users to ask the assistant to compare them to the original RFP document, summarize the various bids and answer questions on any aspect of the offers.
GenAI is also well-suited for demanding tasks that unlock the value of a company’s data assets, enabling them to generate insights from the data and identify opportunities. However, to access this value, the data held in companies´ ERP systems must first be cleaned and harmonized.
Companies frequently hold data across multiple ERP systems, each of which may structure its information differently. Using existing manual methods, it can take hundreds of hours to clean and categorize the data for detailed analysis – a task that an AI model can accomplish in a fraction of the time. GenAI’s ability to interpret natural language and understand context makes it adept at cleaning and harmonizing data, for example identifying typos and abbreviations for a single supplier such as Microsoft (Microsoft, Microsoft, MSFT,…) with one standard way of referring to Microsoft or replacing several names for the same material with a single label. It can also transform data from the disparate structures followed by different systems into a standard format for analysis.
Having cleaned and transformed the data into a standard format, GenAI can categorize the data into product and spend types, which it can analyze in order to identify potential alternative sources of supply and provide insights about the client’s spend with each supplier. It can also identify potential savings and efficiency gains and suggest possible courses of action such as consolidating the buying of a product in a smaller group of suppliers, identifying possible alternative products or suppliers, or leasing equipment rather than owning it, quantifying the savings and benefits each step would achieve.
PROCUREMENT: LET’S GET REAL!
Even among many of the companies that can see the potential of GenAI in their own business, questions remain to data security. There have been many warnings that data uploaded to public LLMs, such as the free versions of ChatGPT, is no longer confidential and will be incorporated into the training data that the model learns from. However, confidentiality concerns can be overcome.
Companies can use solutions provided by third parties, including for example Microsoft Co-Pilot or Adobe’s AI Assistant, which run on private servers in the cloud to ensure that confidential information is protected. Those who want to implement GenAI tools they have built internally or commissioned from a supplier, have several choices of how to create a secure, private LLM. They can license ChatGPT Enterprise directly from OpenAI or use one of the leading cloud providers such as Microsoft Azure, which has close ties with OpenAI, or Amazon’s AWS unit. With this in place, they can license an LLM from one of the major providers and train it using their private data without external access.
The capacity of GenAI models to invent information, events and people that are presented as if they were real is well known and numerous examples of so-called hallucinations generated by public GenAI models such as ChatGPT have been reported. However, it is important to understand the mechanism underlying these AI-generated “flights of fancy”.
The extent to which a GenAI model hallucinates can be controlled by adjusting its temperature parameter. Public versions of LLMs such as ChatGPT have a fixed temperature that users cannot alter, which means they cannot influence how much the GenAI will hallucinate. However, for companies that want to implement and train their own LLMs, there are several critical factors to consider.
First, all GenAI models hallucinate. This is desirable because one of the reasons to use GenAI is, that it can generate new content and alternative possibilities. When companies implement a private LLM trained on their own data, they can decide how much hallucination the model will be permitted by altering its “temperature” parameter rather than being forced to accept a standard temperature setting.
Other factors also have a bearing on how much the model will hallucinate. For example, some LLMs are better suited to certain tasks than others, so choosing the right model is vital to avoid unwanted hallucinations. Also, prompt design and prompt engineering require close attention since many of the problems that hallucinations create can be avoided by writing better prompts.
style="color: #000000;">The ability to fine tune the degree of hallucination they want from their LLM, will give companies greater confidence in the results their system produces. But it will never be infallible, which is why it is essential that there is always a “human in the loop” to check the outputs as a last line of defence.
Many companies have yet to explore how GenAI could benefit them, often because they are not aware of its capabilities and fail to see how useful it could be. Another common misconception is that implementing GenAI is a major task. This need not be the case, although if the company is still running legacy IT systems, this can make any implementation more challenging. But even then, implementing Gen AI with a cloud provider need not be enormously time- or cost-intensive.
However, there are key steps that companies need to follow, which start with putting responsible AI guidelines in place. These should detail the ethical, legal and technological aspects of the company’s AI governance, as well as considering the culture required to integrate GenAI successfully into the workplace. The assessment and strategy phase: With the right framework in place, companies can start to define the needs that GenAI is to meet, assess their current technology capability and skills base, and agree the outcome they are trying to reach.
Having identified the use cases that GenAI will be applied to, companies can then move to solution design. In this phase they set out how the project will be delivered including the technology infrastructure, data requirements, skills gaps and governance issues to be addressed. For example, it may be that the company will need to source additional date externally to train the models it intends to implement.
At this stage, an initial pilot should be carried out that tests a solution for one element of a use case with the aim of validating the overall approach. Concerns and reservations among employees about the impact of GenAI are natural but can be mitigated at this stage by carrying out small, controlled experiments with the technology to show how it can accelerate routine tasks.
Having achieved a successful proof of concept, attention can then shift to developing and testing the overall solution, including test-integrations with the organisation’s existing ERP system to ensure smooth functioning.
Successful implementation has technological and organizational dimensions. IT and Procurement departments must cooperate closely to ensure the system is rolled out as intended and delivers the benefits envisaged at the outset. However, the change management aspects of any roll-out are also significant. Implementation will require training provision in areas such as prompt design – which is likely to require hiring skilled personnel – as well as steps to smooth the transition to new ways of working. BCG estimates that 70% of the value potential of implementing GenAI rests on people and processes, rather than the technology itself. Some companies identify employees keen to act as “AI champions” within the organization. These champions operate as advocates, explaining the tools to colleagues and helping them learn how to get the most out of them, as well as identifying the high-value aspects of their job that require human skills.
Digital and technological proficiency
Procurement teams must develop a high level of digital literacy, ensuring they are proficient in utilizing and integrating various digital tools and platforms. This includes leveraging AI-driven software to streamline procurement processes, manage digital data efficiently, and deploy technologies like GenAI to optimize operations. Mastery in these areas is crucial for teams looking to enhance productivity and maintain a competitive edge.
Data analysis and interpretation
The ability to clean, harmonize, and analyze vast amounts of data is essential. Procurement professionals need to extract actionable insights from complex datasets, identifying trends, risks, and opportunities that can drive more informed decision-making. This skill underpins many aspects of modern procurement, from supplier selection to demand forecasting.
Strategic decision-making
In a world where AI can provide deep insights, the ability to interpret this data and make strategic decisions is critical. Procurement professionals must be adept at using AI- driven insights to enhance efficiency, foster innovation, and sustain a competitive advantage. Strategic decision-making requires not only a strong analytical mindset but also the foresight to anticipate future challenges and opportunities.
Lifelong learning
With technology and best practices in constant flux, a commitment to lifelong learning is non-negotiable. Procurement professionals must continuously update their skills and knowledge to stay abreast of emerging trends and maintain their effectiveness. This dedication to ongoing education ensures that they remain adaptable and prepared to meet the demands of the future.
Resilience
The ability to adapt and recover from disruptions is a vital skill in procurement. Whether facing supply chain interruptions, economic fluctuations, or the challenges posed by new technologies, resilience allows professionals to maintain stability andcontinue delivering results. In a dynamic environment, resilience is what enables procurement teams to navigate uncertainties and sustain their performance.
Once the implementation has been completed, it is essential to track the system’s performance and fine-tune it to ensure it continues to produce the desired outcomes. Some degree of drift is often observed in AI models over time, so monitoring is vital to identify and address this as soon as possible.
Melih Yener is a Partner and expert in Artificial Intelligence and GenAI at BCGX in Hamburg. He has a background in the TMT industry and has been helping companies develop effective digitalization strategies for several years now. In this interview, he gives an overview of where companies currently stand in relation to GenAI and how they can take the step into the future.
Everyone is talking about GenAI, everywhere you read about the benefits that the new technologies can bring for companies. Where do you see the greatest opportunities?
I see the greatest advantages above all in the wider availability of GenAI, which is no longer reserved for experts alone, but also enables less tech-savvy users to develop innovative solutions. In addition, GenAI contributes significantly to increasing process efficiency and supports decision-making processes through real-time analysis and actionable insights. Finally, GenAI enables tasks to be completed faster and to a higher quality, giving companies a decisive competitive advantage.
What is the actual implementation status in companies?Are there any industries or areas that you perceive as pioneers?
The implementation status of GenAI varies greatly between companies. Many need time to fully understand the topic. Some are starting pilot projects, but large-scale use is often still lacking. Tech companies are more open due to their affinity with technology, while there is great interest in the healthcare and automotive sectors, but there is often a lack of maturity. Automotive companies, for example, are currently investing heavily in advancing GenAI. In the healthcare sector, on the other hand, the data basis for comprehensive use is often still lacking, but there are areas, such as drug research, in which GenAI can already be used to good effect. One problem is that GenAI projects are often driven by the IT department and are also perceived as IT projects. However, the specialist departments, who know the application areas best, should be the driving force. Experience shows that projects are successful when the specialist departments take responsibility.
What are the biggest challenges?
The “last mile” – i.e. the transition from the pilot phase to day-to-day business – is crucial. The biggest challenges in the introduction of GenAI lie in scaling and integration into day-to-day business. Without this and the measurement of benefits, the use of GenAI will remain ineffective. A series of small, juxtaposed use cases will ultimately not create any significant impact. It makes more sense to focus on use cases that are scalable, i.e. that can be rolled out quickly throughout the company if successful. Companies often pay too little attention to these aspects and view them as purely technological processes. Managing expectations is also a challenge: there are two extreme poles between which companies fluctuate – either a sense of fear or the exaggerated expectation that AI can solve everything.
What advice would you give to companies that are starting from scratch?
They should start gaining experience immediately so that they don’t fall behind. It is crucial to take a comprehensive look at the topic and consider the potential of GenAI across all areas. Both the technology and what is feasible should be explored. A dual approach is advisable: take an exploratory approach and identify specific projects at the same time. Companies should first examine a broad catalog of topics and then concentrate on specific focal points. In doing so, companies should tackle two to three larger projects in order to understand what concrete benefits GenAI brings in the respective use case. The learning rate is crucial in order to be able to try things out as a team and find out what is possible.
Initially, multipliers should be involved across departments in order to involve as many employees as possible and gain their approval. Transparency within the company is crucial for the success and broad acceptance of “Colleague GenAI”.
What other factors are important for GenAI to become a real competitive advantage?
To turn GenAI into a real competitive advantage, companies should focus on larger, strategic issues. Efficiency and process optimization are important and must be done in order not to fall behind. But true game changers are innovative projects that are difficult to imitate and therefore superior to the competition. These differ depending on the industry. It is also important to find and retain experts and build up relevant skills in the company over the long term. Technology is becoming increasingly readily available, so companies need to build more than the standard offering. It is crucial to know which technologies should be developed internally and which can be sourced externally. Investment should be targeted where it will bring the most benefit.
What is your vision for the future? Where will companies be in 5 years’ time and what will be important in the future?
Given the rapid pace of technological development, it is difficult to predict exactly where companies will be in five years’ time. Nevertheless, I expect that we will reach a level where AI has human-like capabilities. AI employees could take on tasks independently and thus significantly change the labor market and regional structures.
In areas where a high level of empathy and human relationships are important, people will continue to work. However, where it is mainly about “doing”, AI co-workers will take over these tasks. This development should be positive, as it counteracts the shortage of skilled workers and supports companies: overall, the journey is moving towards “AI colleagues” who complement the workforce and relieve them of routine tasks.