Artificial Intelligence (AI) is revolutionizing numerous industries, and procurement is no exception. This article provides a comprehensive overview of AI in procurement, covering everything from basic definitions and types to examples of AI in procurement and current trends.
Our goal is to not only highlight the benefits but also the challenges and prerequisites of implementing AI in procurement.
Artificial Intelligence is the ability of a machine to imitate human skills such as logical reasoning, learning, planning, and creativity to assist with tasks traditionally performed by humans. Due to the data-intensive and complex, yet repetitive tasks in procurement, the use of AI in procurement proves to be particularly useful and advantageous.
Artificial Intelligence can be categorized into different types depending on the level of specialization:
Machine Learning: Machine Learning is a field of artificial intelligence that enables computers to learn from data and improve over time without being explicitly programmed to do so.
Deep Learning: Deep Learning (DL) is a subset of machine learning that uses neural networks to analyze large and complex datasets.
Generative AI: Generative Artificial Intelligence (often referred to as Gen AI) includes algorithms designed to generate new content such as audio, code, images, and text.
LLMs (Large Language Models): A Large Language Model is a specialized form of artificial intelligence (AI), specifically deep learning, trained on large amounts of text. It can understand existing content andcreate new, original content using Generative AI (Gen AI).
Supervised Learning:
In supervised learning, the training data must be carefully prepared and labeled. The model learns from this data by recognizing the relationships between the input features and the outcome.
Example: You want the AI to categorize suppliers into "high risk," "medium risk," and "low risk." Various data such as delivery punctuality, company location, debt level, number and severity of quality defects, and compliance with legal regulations are used for this purpose.
The training data includes this information for each supplier as well as the respective risk category. The AI learns the relationships between these factors and the risk categories. The AI's performance is then measured using a test set, for example, by the accuracy of the model.
Unsupervised Learning:
Unsupervised learning is referred to as "unsupervised" because the algorithm does not require labeled data. Unlike supervised learning, which works with predefined examples, unsupervised learning analyzes unstructured data to independently identify patterns and relationships. The algorithm clusters the data or detects anomalies without prior instructions.
An example is cluster analysis, where suppliers are grouped based on characteristics such as quality, delivery reliability, and cost management. Unlike supervised learning, the AI identifies these groups independently without prior instructions. These automatically identified groups can then be used for specific development programs, such as training or quality initiatives.
Reinforced Learning (RL)
Reinforcement Learning (RL) works with unlabeled datasets, similar to unsupervised learning. In RL, an agent learns to make optimal decisions through interaction with its environment. The agent receives rewards for good actions and penalties for bad actions to develop the best strategy. An example is an agent learning to play a video game by receiving rewards for positive actions and penalties for negative actions. However, there are fewer use cases for RL in procurement.
A "strong" AI possesses intellectual abilities equal to or superior to those of a human. This means it can autonomously and intuitively perform a wide range of complex tasks that require both analytical and emotional intelligence.
"Weak" AI, on the other hand, cannot learn independently in a universal sense. Its learning capabilities are limited to recognizing complex patterns in large datasets. All current applications, both general and in procurement, are still based on weak AI.
Artificial Intelligence is revolutionizing purchasing and procurement by automating complex processes, performing big data analyses, and more. These technologies increase transparency, improve decision-making, and automate manual tasks, leading to greater efficiency and productivity. AI also helps reduce errors and manage risks, contributing to significant cost savings.
Investments in AI and its relevance have grown exponentially in recent years. Currently, only 36% of companies use AI in procurement, but 76% of the remaining firms plan to implement it in the near future, with many concerned about falling behind technological advancements.
Studies show that the use of AI significantly enhances user performance, and the continuous improvement of AI models and the availability of large datasets promise that the benefits of AI in procurement will continue to grow in the future
Investments in AI have surged, reaching $25.2 billion in 2023, which is nine times higher than in 2022 and 30 times higher than in 2019. AI is becoming increasingly popular not only to investors but also to the general public, especially since the introduction of ChatGPT. This is visible in the graph demonstrating a rapid rise in search volume for AI, reflecting the growing interest in this technology.
A survey by the Kloepfel Group shows that the majority of companies do not yet use AI in procurement, but many plan to do so in the future. Currently, 36% of the surveyed companies use AI in procurement, while 64% have not yet implemented AI technologies. However, among the companies without AI, 76% plan to implement it in the near future.
At the same time, only 2% of German companies see themselves as leaders in AI adoption compared to their competitors, 13% consider themselves pioneers, while 43% view themselves as laggards. Alarmingly, 38% fear they have already fallen behind technological advancements (source: Bitkom). This indicates that while many companies want to adopt AI, they are concerned about already lagging behind.
Currently, all known use cases of AI in procurement, as previously mentioned, are based on "weak" AI, which has limited learning capabilities. However, within this weak AI, there are different levels of autonomy and automation with which the AI can operate in procurement.
Low Automation:
On the left end of the scale are tasks such as supplier development and complex negotiation strategies. These tasks require a high degree of human intuition, experience, and strategic thinking, which is why they can currently only be supported by AI to a limited extent.
Supportive:
In the middle of the scale are tasks such as risk management, contract management, and supplier evaluation. These activities benefit from supportive AI, which performs complex data analyses and provides decision-making assistance. Here, AI can efficiently process large amounts of data and recognize patterns to enable informed decisions, while the final decision still rests with humans.
Fully Automated:
At the right end of the scale are tasks such as order processing, supplier qualification, and simple negotiations. These activities are well-suited for complete automation by AI, as they are easily learnable by the AI. Here, AI can take over repetitive processes, increasing efficiency, minimizing errors, and significantly reducing the workload of employees.
Mercanis is a holistic procurement suite focused on S2C (Source-to-Contract) that stands out with its use of AI, intuitive user experience, and fast implementation for customers.
Mercanis offers the following modules:
Through the use of artificial intelligence in sourcing and procurement, digitalization, and automation, the platform achieves significant cost savings in pricing and processes. As a single source of truth, it brings transparency to the supplier base, enabling data-driven decisions.
Enrichment of Supplier Profiles: Mercanis automatically consolidates information on suppliers from your system components, such as the ERP system, into a central supplier database to increase transparency in the supplier landscape. Our AI then further enriches these supplier profiles from various online sources (supplier articles, reviews, contact details, and more) to identify the core competencies of the suppliers. This enables a detailed assessment of the suppliers' capabilities and strengths.
Recommendation of Alternatives: Another example of AI in procurement is a supplier recommendation system. Mercanis analyzes similarities in quality, performance, pricing structure, and other relevant characteristics to find suitable alternatives or additions to existing suppliers. This promotes the diversification of the supplier network.
Artificial Intelligence in sourcing enables the recognition and analysis of project data from sourcing events to understand specific requirements and goals. Simultaneously, it evaluates the capabilities of all suppliers to automatically identify and recommend the best suppliers for your project.
Bid Comparison: With AI-powered bid comparison, you receive a quick and clear presentation and summary of the submitted bids. This enables informed decisions in a shorter amount of time.
Overview and Negotiation Recommendations: The AI provides negotiation recommendations based on key criteria such as price, availability of essential documents, completeness of the supplier questionnaire, etc.
Automation through AI: With a simple upload of the contract, predefined fields for deadlines and terms are automatically filled in by the AI. Additionally, the AI performs active risk assessments based on existing contractual relationships, leading to better decision-making.
Mercanis GPT Chatbot: The Mercanis GPT-Chatbot uses procurement-specific generative AI. The chatbot reads the contents of the contracts and allows the purchaser to query the contents similarly to ChatGPT. The data is securely stored in Germany and not used for development purposes – the customer retains exclusive ownership of the data.
Artificial intelligence in procurement automates manual tasks and relieves buyers, allowing them to focus on more strategic activities. This leads to improved speed and productivity, and a reduced administrative burden.
For example, a study found that business professionals using AI were able to write 59% more business documents per hour. For procurement, which often involves a lot of text, such as free-text orders, generative AI in procurement is particularly relevant.
Today's business leaders and managers are confronted with an unprecedented amount of data from various sources, increasing the pressure to make critical decisions accurately.
AI can assist in decision-making by quickly, efficiently, and objectively recognizing patterns and analyzing large, multidimensional data sets that would be difficult to handle manually. An example of this is identifying the right supplier from a multitude of vendors with the required capabilities.
Well-trained artificial intelligence makes partially fewer mistakes than humans and often makes fewer erroneous decisions due to data misinterpretation.
Thanks to its ability to quickly analyze large, multidimensional data sets, AI is also excellent at identifying various risk indicators, which can be used to implement preventive measures.
Through increased productivity, improved decision-making, error reduction, and risk minimization, costs can be optimized.
According to McKinsey, 33% of respondents report cost savings from using artificial intelligence in supply chain management. Of these, 9% report cost savings of 10-19%, while 24% report savings of less than 10%.
Due to the continuous improvements in AI, the benefits of AI will become even more significant in the future. In certain fundamental tasks, such as image classification, AI has already surpassed human performance.
One of the reasons for this is the trend towards increasingly larger and better-trained models, as well as the rise in global data volume. Forecasts predict that the global data volume will reach 284.3 zettabytes in 2027, compared to 2 zettabytes in 2010.
In procurement as well, digital procurement platforms like Mercanis can collect and digitize more and more data, making it effective for AI-supported procurement tools.
According to companies, the top 3 reasons they do not use AI are lack of knowledge (72%), incompatibility with existing systems (54%), and data availability (53%).
Not all concerns about the use of AI are therefore justified, and many of these hurdles can already be easily overcome with the current state of technology.
Robotic Process Automation (RPA) is a technology that uses software robots to automate repetitive, rule-based tasks. An example of an RPA application is when Mercanis sends you notifications to remind you about contract expirations. The rule here would be: If the contract term is only a few days from the current date, then send a notification.
The main difference from Artificial Intelligence (AI) is that AI is data-driven and uses machine learning to recognize patterns in data and learn over time, rather than just executing fixed rules.
Therefore, RPA is ideal for simple, repetitive tasks, while AI is used for more complex tasks.
Many people fear that an AI procurement-software could jeopardize their job in procurement. Indeed, tasks such as filling out Excel files, writing standard texts, and conducting research can be done more efficiently by AI.
However, this does not mean that the affected employees will become unemployed. Instead, they will be relieved of tedious routine tasks and will have more time for tasks where humans are indispensable.
There are many such tasks in procurement, such as personal communication with suppliers, strategic supplier development, and strategic decision-making. Therefore, procurement professionals do not need to worry about AI taking their jobs.
Rather, they should actively support the implementation of AI to spare themselves repetitive tasks and focus on value-adding activities.
"AI is clearly better (than professionals) at processing large amounts of data. Its limitations lie in specialized and individual knowledge."
- Digitalization expert Fabian Kittel in an interview with BeschaffungAktuell
There are currently significant discrepancies in the use of AI depending on the size of the company. According to the German Federal Statistical Office, only 10% of small companies use AI compared to 35% of large companies. However, AI is not only effective for large companies.
Advances in technology, especially with large language models, have made the use of AI simpler, more cost-effective, and more accessible. Huge amounts of data are not required to train AI and integrating it into the business no longer requires deep technical knowledge.
According to the German Federal Statistical Office, 41% of respondents state that they do not use AI due to high costs. Historically, AIs were indeed very expensive, as they required large amounts of data and significant computing power. Training models like ChatGPT, for example, cost around 5 million US dollars.
However, for most companies today, AI is neither expensive nor difficult to implement. The reason is that most companies do not need to train their own AI but can rely on pre-trained models. These pre-trained models are available to individual companies at relatively affordable prices, as exemplified by ChatGPT. This allows a broader range of companies to leverage the benefits of AI.
As previously mentioned, most companies do not need to develop their own AI. Therefore, they generally do not require vast amounts of data. The data requirements depend on how specific the business needs are and, consequently, how extensively the AI needs to be trained.
If the need is relatively general, such as summarizing supplier offers, a fully pre-trained AI can be used. However, if you need a specialized AI that needs to be trained specifically on your data, a less pre-trained AI will be required, which demands more data.
Since most AIs have already been trained on extensive datasets, they can handle many AI use cases in procurement without additional effort.
AI in procurement refers to the application of artificial intelligence to support and optimize procurement processes. This includes tasks such as supplier evaluation, contract management, and data analysis to improve efficiency and decision-making.
Implementing AI software in procurement leads to increased productivity, optimized decision-making, error reduction, more effective risk management, and cost optimization.
Generative AI, such as Large Language Models (LLMs), can create new content and is particularly useful for understanding and generating text. Other AI tools in procurement focus more on analyzing and optimizing existing data, such as machine learning for sourcing and supplier evaluation.
Artificial Intelligence supports purchasing and procurement by analyzing large amounts of data, recognizing patterns, and making predictions that lead to better decisions in supplier selection and price negotiations. It can also identify and manage risks in the supply chain.
To successfully implement AI in procurement, companies need to identify the need, define clear AI use cases in procurement, clearly specify the expected use and input of the AI, collect sufficient data, and set clear KPIs to measure success.
Machine learning in procurement improves decision-making by analyzing large datasets, recognizing patterns, and making accurate predictions. Machine Learning is capable of analyzing larger datasets and avoids typical biases that occur in human data interpretation. This enables informed, data-driven decisions that lead to better outcomes.
AI can promote sustainability in procurement by analyzing ESG (Environmental, Social, Governance) criteria, evaluating supplier performance, and supporting companies in making decisions and selecting sustainable suppliers.
The costs vary depending on the scope and complexity of the AI applications. While specific, self-trained AI solutions can be expensive, pre-trained models are often more cost-effective and easier to implement.
With the use of AI, purchasers will be relieved of routine tasks and can focus more on strategic and value-adding activities, such as developing supplier relationships and making strategic decisions.
Fabian Heinrich is the CEO and co-founder of Mercanis. Previously he co-founded and grew the procurement company Scoutbee to become a global market leader in scouting with offices in Europe and the USA and serving clients like Siemens, Audi, Unilever. With a Bachelor's degree and a Master's in Accounting and Finance from the University of St. Gallen, his career spans roles at Deloitte and Rocket Internet SE.
.webp)
.webp)
.webp)
