Artificial intelligence has become a focal point in the realm of technology, which presents a burgeoning opportunity for procurement. Despite its novelty, AI is often hailed as a miraculous solution to various issues. However, discussions frequently revolve around potential future benefits rather than tangible business applications.
This exploration delves into the authentic AI prospects within procurement, aiming to shift the focus from speculative discourse to practical realities. It scrutinizes the genuine opportunities that AI presents, unravelling its capacity to address major challenges in the procurement landscape. By demystifying the hype and deciphering buzzwords, this guide assists in navigating the complex terrain of AI. Companies are now using AI for procurement in use cases including complex invoice processing and chatbots and digital assistants to help locate the answers to lead time questions.
Leveraging artificial intelligence (AI) in procurement empowers organizations to streamline the resolution of intricate challenges with enhanced efficiency and effectiveness, courtesy of intelligent computer algorithms. The integration of AI extends across various software applications, ranging from spend analysis and contract management to strategic sourcing.
While the academic exploration of AI dates back to the 1950s, its tangible applications within procurement functions have only recently materialized. At its core, AI represents a diverse array of cutting-edge computer technologies endowed with the capacity to learn and autonomously adjust their behaviour. In the right context, AI software can outperform the efficiency of human capabilities when addressing complex tasks. For example, our team has recently explored AI as a tool to distill complex invoices that are thousands of lines long and extract important details from email and automatically review, evaluate and post the information into SAP.
Distinguishing between strong AI and narrow AI sheds light on the varied capabilities within artificial intelligence. Narrow AI operates within predefined functions, tackling specific challenges it has been taught to address. In contrast, strong AI possesses autonomous cognitive abilities and can adeptly handle any task conceived by the AI itself.
Popular media often alludes to strong AI, also known as artificial general intelligence (AGI), as portrayed in science fiction films and TV shows (Westworld being a notable example). These representations depict machines performing intelligent tasks at or above human proficiency levels.
However, the current landscape of AI applications in procurement predominantly revolves around narrow AI, often labeled as weak AI. These solutions offer intelligent resolutions to highly specific and predefined challenges. Despite their limited scope, narrow AI presents immediate potential for enhancing operational efficiency.
Procurement AI signifies a transformative force, automating or enhancing various time-consuming tasks and providing experts with valuable insights derived from vast and intricate datasets. Conceptually, AI is a specialized software solution designed to address specific tasks. It is crucial to view AI beyond the sensationalism and recognize it as a novel form of software with the capacity to swiftly reshape work practices within large organizations.
It is essential to dispel misconceptions surrounding AI in procurement. Contrary to cinematic portrayals, Procurement AI is not a chrome-plated or plastic-coated sentient being. It should not be perceived as a replacement for human expertise or as a new team member capable of driving organizational change, strategic sourcing, or generating savings. Rather than a magical problem-solving solution, AI in procurement necessitates active expert guidance and oversight.
When it comes to procurement, any software incorporating self-learning capabilities and intelligent algorithms falls under the umbrella of Artificial Intelligence (AI). Some helpful definitions to understand when it comes to AI in procurement are:
All variants of AI rely on algorithms, which are essentially sets of rules outlining how to solve specific problems. While proficiency in mathematics enables the calculation of algorithms, they serve as the foundation for most computer software. Although the workings of algorithms in software remain imperceptible to the human eye, experts can program and reprogram them to address significant problems within software environments.
It is crucial to differentiate Robotic Process Automation (RPA) from AI in the context of Procurement. RPA offers numerous opportunities to enhance process efficiency but should not be conflated with AI. To simplify, envision RPA as a software robot emulating human behavior, while AI simulates human intelligence.
The integration of AI in Procurement offers substantial benefits, particularly in addressing intricate problems characterized by vast datasets and clearly defined success metrics. A collaborative study by Harvard Business Review and Deloitte delves into key areas where business executives anticipate significant success with AI. While challenges and opportunities vary across organizations, the following areas illustrate how AI can bring value to Procurement:
Make better decisions: Artificial intelligence facilitates the provision of timely analytics and data-driven insights, empowering decision-makers in Procurement to enhance the quality of sourcing decisions.
Identify new opportunities: By meticulously sifting through extensive datasets, AI can uncover novel avenues for savings or revenue opportunities within Procurement processes.
Improve operations: The potential of AI extends to streamlining and aligning internal business operations, even within large organizations featuring multiple business units or geographic locations.
Automate manual tasks: AI excels at automating time-consuming tasks, such as monthly processes and Procurement performance reporting, freeing up resources for more strategic endeavors.
Free up time: Through the automation of routine tasks, AI liberates Procurement resources, allowing them to redirect their efforts towards more creative and strategic tasks, including key supplier relationship management.
Capture or apply scarce knowledge: AI proves invaluable in helping Procurement organizations capture pertinent data from diverse sources, including external ones like the Internet, enriching their knowledge base.
Identify new suppliers or markets: With access to vast external data, AI emerges as a powerful tool for identifying untapped suppliers and exploring new markets, providing valuable insights for strategic decision-making.
Optimize supplier relationships: AI's data-driven capabilities hold the availability to elevate supplier relationship management, ensuring a more informed and optimized approach to handling relationships within the Procurement domain.
Embracing automation in procurement entails the systematic streamlining of processes to optimize efficiency and minimize cycle time. This transformative approach liberates employees from the constraints of repetitive, manual, and time-intensive tasks, accelerating the entirety of the procurement process.
Efficiently automate your procurement cycle in six steps:
Begin by comprehensively mapping out your existing procurement procedures. This foundational step provides a clear understanding of the current workflow and sets the stage for subsequent improvements.
Conduct a thorough audit of the current procurement process, identifying strengths, weaknesses, and potential areas for enhancement. This evaluation serves as a crucial benchmark for gauging the impact of automation.
Pinpoint labor-intensive and repetitive bottlenecks within the procurement process that are prime candidates for automation. Focusing on these areas ensures a targeted and impactful implementation of automated solutions.
Select an analytics solution that aligns with your automation objectives. A well-chosen analytics tool plays a pivotal role in facilitating and supporting the automation of procurement processes.
Develop tailored automation workflows and strategically position approval points within the process. This step involves configuring the system to handle tasks seamlessly, ensuring a cohesive and efficient automated procurement cycle.
Establish metrics to measure the success of your automation efforts. Regularly assess the performance of automated processes and implement continuous improvement strategies to refine and optimize the automated procurement cycle over time.
By following these steps, your organization can harness the power of automation to enhance efficiency, reduce cycle time, and elevate the overall effectiveness of your procurement processes.
In the early stages of integrating AI into business applications, the world of Procurement is witnessing a growing number of examples showcasing the potential of AI. Notably, machine learning algorithms are making significant strides in spend analysis, enhancing and expediting various processes, including automatic spend classification and vendor matching, with even more innovations on the horizon.
Examples of spend classification techniques exemplify the diverse applications of AI in Procurement:
This involves training algorithms to detect patterns in spend, alleviating the mundane task of repetitively classifying new spend. Humans impart knowledge to the algorithms, enhancing efficiency.
Unsupervised Learning in Vendor Matching
Algorithms autonomously detect new and intriguing patterns in vendor relationships without human intervention. For instance, machine learning algorithms can consolidate varied vendor names (e.g., DHL, DHL Freight, Deutschland DHL, and DHL Express) for increased visibility and data coherence.
Actions taken by algorithms in spend classification are reviewed by humans, who then reward or penalize the algorithm based on outcomes. This collaborative approach ensures ongoing refinement and optimization.
However, it's essential to acknowledge that achieving 100% automation may not always be realistic. Often, 80% of a process, such as spend classification, can be automated, while the last 20% may necessitate human involvement. Adhering to the 80/20 rule provides a realistic perspective on the time required for an AI-driven process and its potential to enhance existing timelines.
AI employs techniques like natural language processing to extract data on suppliers or specific markets. For instance, monitoring social media channels can provide insights into suppliers' risk positions. AI enhances predictions, such as price predictions, maintenance needs, and stock market forecasting.
AI's ability to leverage new data sources is exemplified by the utilization of external data. Market indices, company credit ratings, or publicly available information about suppliers constitute "external" data sources sifted through by AI-powered methodologies. This process identifies opportunities, offers benchmarks, and provides recommendations for performance improvement.
Consider the task of benchmarking performance. While using internal and static historical data provides a fairly accurate picture, incorporating external data, such as market reports and stock prices, introduces a whole new level of insight, enhancing the depth and accuracy of performance assessments.
When it comes to anomaly detection in procurement is on the cusp of a transformative era where automated notifications about anomalies, emerging opportunities, and recommended actions could become an everyday reality. The increasing prowess of AI in processing vast amounts of data enables it to maintain real-time awareness of the latest developments and shifts in the operating environment.
As AI becomes more adept at processing data, it gains the capability to promptly and accurately detect anomalies and changes. The system can instantly notify the procurement team when something abnormal occurs, providing immediate suggestions on potential courses of action. Furthermore, AI has the capacity to present simulations for different scenarios and highlight new opportunities based on the data at its disposal. This proactive approach ensures that human procurement practitioners are more attuned to unfolding events, allowing for swifter decision-making.
Crucially, the recommendations put forth by AI are rooted in real facts, eliminating reliance on human hypotheses or guesswork. Procurement leaders can place trust in the accuracy and objectivity of AI-generated insights, instilling confidence that their decisions are firmly grounded in authentic data. This shift from uncertainty to data-driven decision-making contributes to overall improved decision quality within the procurement domain.
Highly successful Chief Procurement Officers, as highlighted by Deloitte, are 18 times more prone to having fully implemented AI and cognitive capabilities. The following outlines seven prevalent areas where AI can be effectively applied throughout the procurement cycle.
One use case the ConvergentIS team has been particularly excited about is the use case for complex invoicing, such as hydrocarbon invoices that might have the upwards of 300-400 line items. With AI, procurement teams would have the ability to select or attach all the invoice files that you want to be processed or have the vendor send them to a predefined e-mail ID for the bot to automatically pick them up and process them from there. AI in this use case would help to accelerate processing by allowing teams to see all the key pieces of information that are being extracted from the vendor invoice file.
AI holds significant potential in contract management, with natural language processing being a tangible example. Software like Docusign Insights utilizes this capability to autonomously scan and interpret extensive legal documents, identifying potential cost-saving opportunities.
Artificial intelligence plays a crucial role in monitoring and detecting potential risk factors throughout the supply chain. For instance, RiskMethods RiskIntelligence employs big data methodologies to screen millions of diverse data sources, providing alerts within supply chain risk management software.
The automation of reviewing and approving purchase orders can be seamlessly achieved through artificial intelligence. Tradeshift's platform, featuring a chatbot named Ada, exemplifies how AI can be employed to check purchase statuses or automatically endorse virtual card payments.
Machine learning is increasingly integrated into accounts payable automation, exemplified by solutions like Stampli. This AP automation software utilizes machine learning to expedite payment workflows and detect instances of fraud.
In procurement spend analysis, machine learning algorithms are widely applied to enhance and expedite processes such as automatic spend classification and vendor matching. A more detailed exploration of this example will be covered in a separate chapter of this guide.
Leveraging big data techniques, AI provides innovative approaches to identify, manage, and utilize supplier data from both public and private databases. Platforms like Tealbook utilize machine learning to refine supplier discovery based on information sourced, cleansed, and enriched from the Internet.
Artificial intelligence is instrumental in managing and automating sourcing events. Keelvar's sourcing automation software, for instance, employs machine learning for bid sheet recognition and features specialized category-specific eSourcing bots, such as those for raw materials and maintenance/repair.
The integration of artificial intelligence in business scenarios often involves a level of human supervision, and in the context of procurement, implementations frequently leverage supervised learning. In these instances, procurement experts actively participate in training machines to execute specific tasks.
Here's a breakdown of how AI can be trained with procurement data:
The process commences with providing a set of training data to an AI algorithm, presenting it with a specific challenge. For instance, the algorithm may be tasked with observing how 100,000 invoices are classified into different spend categories.
Armed with a clear goal and initial training data, the AI algorithm is then presented with unclassified procurement data to classify based on the logic it has acquired from the training data.
In instances where the AI exhibits high confidence, data can be automatically categorized without requiring human input.
When the AI lacks high confidence, classification decisions are reviewed by procurement experts in a process known as "human annotation."
Feedback from human reviews is incorporated, enabling the procurement data to be both classified and utilized to actively teach the AI algorithm for future data classification.
As the AI algorithm receives more training and human feedback, its confidence increases, leading to the automatic classification of more data. Simultaneously, the quality of data classification improves based on human input.
When evaluating this training process from the perspective of a procurement organization, it is essential to consider tasks where there is sufficient training data, a consistent need to process unclassified data, and a tangible output that contributes to business value.
The utilization of machine learning in procurement involves deploying self-learning automated algorithms to address specific challenges or enhance operational efficiency. Machine learning (ML) empowers procurement to achieve the highest quality in terms of both volume and bottom-line impact.
As a subset of AI, machine learning emerges as the facet with the most immediate applications in the procurement landscape. It seamlessly succeeds robotic process automation (RPA) in the progression toward automated or autonomous procurement processes. Unlike RPA, which is categorized as automated statistics, machine learning possesses the crucial ability to learn and refine its capabilities over time.
Despite its significance, machine learning remains one of the most misconstrued aspects of AI within procurement organizations. Enthusiasts may categorize any application of advanced statistical methods as machine learning, while certain software vendors perpetuate misleading images of human-like machines. It is imperative to dispel these myths by delving into the core types of machine learning.
Various machine learning types play essential roles in different facets of procurement processes, each requiring a distinct level of human involvement. The four key machine learning types are as follows:
In this approach, an algorithm is trained on patterns derived from historical data, enabling it to automatically identify similar patterns in new data. Human supervision is integral, providing correct answers to guide the algorithm in recognizing patterns. Commonly applied in spend analysis, particularly in areas like spend classification.
Unsupervised learning involves programming the algorithm to discover novel and intriguing patterns in entirely new data. Without explicit human guidance, the algorithm seeks logical patterns within raw data without expecting specific correct answers. Its application is rare in critical procurement functions.
Reinforcement learning entails the algorithm determining how to act in specific situations, with behaviour being either rewarded or punished based on consequences. While largely theoretical in the procurement context, it holds potential for future applications.
Deep learning, a sophisticated category of machine learning inspired by the human brain, encompasses artificial neural networks that continuously improve their capacity to execute tasks. Representing a burgeoning prospect within procurement operations, deep learning demonstrates the potential for sophisticated applications in the procurement domain.
Employing machine learning for spend analysis is already pervasive, especially in data-intensive processes like procurement analytics. Let's delve into the intricate applications of machine learning within spend analytics, with a particular focus on addressing the complexities associated with spend classification.
Spend analysis entails the comprehensive process of identifying, collecting, cleansing, classifying, enriching, and scrutinizing an organization's spend data.
The pivotal challenge in procurement spend analytics is the classification of spend into distinct procurement categories. This task represents one of the initial applications of artificial intelligence in the procurement domain, with widespread adoption for automating the classification process.
Consider a scenario where a newly purchased computer is initially labeled as "IT equipment" in the general ledger. However, a more detailed description in the invoice line specifies it as a "desktop computer." Further complexity arises when examining the purchase order for the same item, which might feature a different description, highlighting vendor-specific or maker-specific data points. Despite all these data sources referencing the identical item, accurate classification requires a level of intelligence to discern the nuances.
While various facets of AI hold the potential to address or mitigate the challenges associated with spend classification, the prevailing trend in contemporary software solutions involves some form of supervised machine learning.
Machine learning algorithms serve as a foundation, automatically classifying new spend data into designated procurement taxonomies.
The software offers a classification tool that not only automates the process but also provides suggestions for category experts, facilitating a collaborative approach.
The machine learning-powered AI classifier assigns a confidence level ranging from 0 to 1 for each classification suggestion. Lower confidence levels are closer to zero, signifying uncertainty, while higher numbers indicate greater confidence in the classification.
Beyond new spend classification, machine learning can identify errors in prior rule-based classifications made by human category experts.
Human category experts play a pivotal role by reviewing and validating AI-classified data, offering valuable training input for future classifications. This collaborative interaction ensures continuous improvement and refinement in the classification process.
Natural Language Processing (NLP) stands as a vital branch of artificial intelligence dedicated to comprehending, interpreting, and manipulating human language. Within the realm of procurement, NLP emerges as a powerful tool capable of extracting insights from existing data and introducing innovative methods to streamline laborious processes. Let's delve into specific examples. AI, employing NLP, can effectively parse (interpret) data within legal contracts and documents, extracting pertinent information essential for procurement purposes.
Within the realm of procurement, legal contracts are rich repositories of vital information, including details on termination dates, payment terms, and renegotiation rights. Traditionally, accessing such information has posed challenges for procurement teams due to contracts being written in contractual terms and stored offline or in shared online folders, limiting data accessibility. Natural Language Processing (NLP) revolutionizes this landscape by introducing text parsing methods, allowing procurement to extract valuable data from contracts. Contract management software, leveraging parsing algorithms, efficiently scans and interprets large volumes of contracts, extracting critical information. Additionally, optical character recognition (OCR), an AI-enabled approach, automatically interprets and identifies text, even from images such as photos of previously un-digitized scanned contracts.
AI software excels at interpreting numbers compared to human language, operating on a binary system while humans think in words. Word embedding, a form of NLP, maps words and phrases in vocabulary based on their similarity and relation to others. Procurement benefits from word embedding in the analysis of text fields in purchase orders, identifying groups of purchased items within similar categories or sub-categories.
Natural Language Generation in Chat Bots
Chatbots and personal assistants represent notable applications of AI that hinge on natural language generation (NLG). These systems, extending the capabilities of NLP, interpret human input and respond through written narratives. While voice-based assistants like Siri or Alexa are prevalent in consumer applications, NLG within procurement is currently confined to pre-configured chatbots or virtual assistants automating specific tasks.
Exploring Generative AI for Procurement
Generative AI, a broad category characterized by the ability to create text, images, or other media based on learned patterns from input data, marks a significant development. Large Language Models (LLM), exemplified by OpenAI's Generative Pre-Trained Transformer (GPT), represent a noteworthy advancement. While OpenAI provides access to these models through APIs, the model weights and training code are not openly shared. The introduction of ChatGPT Enterprise in late 2023 indicates increased feasibility and security for business use cases, potentially paving the way for experimentation by procurement organizations dealing with sensitive data.
Efficiently training a model using textual records of past supplier interactions and documents can pave the way for the automated generation of "flag" protocols, discussion summaries, and more.
Enhancing Usability of External Data
The potential applications in this domain are limitless, with one example being the extraction of supplier-specific news from online articles and other internet media. This data can be summarized and delivered in a user-friendly format.
Automated Generation of Briefs and Text-Based Documents
Imagine the simplicity of generating a supplier relationship summary or creating Statements of Work (SOWs), Requests for Proposals (RFPs), and Purchase Orders (POs) with just a click of a button.
Data Classification and Mapping Optimized
Large Language Models (LLMs) are exceptionally suited to elevate procurement's efficiency in data processing, including classification, CO2 mapping, supplier normalization, and data engineering/mapping.
The capabilities of advanced chat tools are noteworthy, facilitating complex and human-like conversations. Models like GPT can generate communications that emulate human interactions, streamlining the process of requesting and receiving information from suppliers, covering aspects such as pricing, lead times, and product specifications.
LLMs exhibit the capability to analyze contracts, identify potential risks, and propose effective mitigation strategies. By comparing contract data to regulatory information, procurement teams can proactively monitor regulatory changes, receiving alerts about updates that may impact their supply chain.
Cognitive procurement stands as a prominent term buzzing within discussions on new technologies and AI opportunities, introducing fresh terminology and definitions to the forefront.
Exploring Cognitive Procurement
Cognitive procurement represents a dynamic process where self-learning AI techniques emulate human intelligence. These techniques encompass automated data mining, machine learning, pattern recognition, and Natural Language Processing (NLP). The term "cognitive procurement" finds its roots in the evolving field of advanced computer science known as "cognitive computing."
Understanding Cognitive Computing
Cognitive Computing (CC) denotes both hardware and software designed to replicate the functions of the human brain, enhancing decision-making capabilities. It mirrors how the human brain senses, reasons, and responds to stimuli to address specific tasks or challenges.
Within the realm of procurement, cognitive computing intersects with cognitive analytics (CA), offering a novel approach to extracting insights from extensive structured or unstructured data. CA mirrors the human brain's aptitude for interpreting patterns and drawing conclusions, providing valuable solutions to procurement analytics challenges. However, it's essential to note that not every AI-assisted analytics solution necessarily incorporates cognitive elements.
Cognitive computing also lends its support to procurement through cognitive sourcing, contributing to the identification of new opportunities and the automation of non-strategic sourcing activities. Sourcing assistants, exemplified by chatbots, serve as tangible instances of cognitive sourcing applications.
Given the novelty of the cognitive computing field, a cautious approach is recommended. The lack of a universal consensus on core definitions and the parameters of "cognitive" processes in a business context poses a challenge. Despite rapid technological advancements, it is advisable to validate assumptions within cognitive procurement by consulting internal or external information systems experts.
Commence with Mundane Challenges
Embarking on the AI journey involves addressing practical challenges rather than seeking miraculous solutions to revolutionize Procurement operations. Rather than viewing AI as a mystical technology, approach it from a business process standpoint. Focus on the tedious yet critical business operations that already demand time and resources. The immediate value of AI lies not in introducing entirely new applications but in integrating technology into existing processes. For instance, consider enhancing established processes like spend analysis or contract management.
Comprehensive Procurement Data Gathering
A fundamental principle is to gather an extensive range of data relevant to Procurement, even before understanding how to leverage it effectively. Instead of waiting for data quality to reach perfection, anticipate that AI technologies can contribute to interpreting and enhancing historical data quality progressively. The crucial aspect is to accumulate a wealth of data for AI interpretation. Providing AI with a substantial dataset for training enhances the likelihood of achieving better results.
Present Clear Procurement Challenges to AI
In its current iteration, AI and machine learning excel in specific, well-defined applications. Utilize machine learning for tasks such as categorizing Procurement costs based on invoice line items, while recognizing the limitations in handling intricate supplier negotiations. Assess routine tasks that consume significant time for your procurement team but yield clearly discernible outcomes in performance.
Embrace a Culture of Experimentation
Despite the potential for AI to enhance Procurement performance, uncertainties persist. Foster an open mindset toward experimentation. Consider entrusting emerging AI technology specialists with challenges and providing training samples of your data. Embrace the inevitability of mistakes and the learning curve, focusing on the anticipated business benefits. Acknowledge the rapid pace of technological evolution, where today's failed experiments may pave the way for tomorrow's successful AI methods.
Facilitate Human + Machine Collaboration
Crucially, acknowledge that all AI implementations in Procurement demand active guidance and support from procurement experts. Embrace a collaborative approach, envisioning human and machine collaboration where the expertise of the Procurement team is enhanced, not replaced, by artificial intelligence. Champion the role of change facilitator to optimize the synergies between human and machine intelligence.
Predicting the exact trajectory of procurement and AI over the next 10-20 years remains uncertain, but certain insights can be gleaned about the potential future possibilities. Analysts widely agree that existing AI applications will evolve and advance in the coming years.
McKinsey suggests that tasks like payment and invoice processing, order placement and receipt, and demand management for purchases are relatively straightforward to automate, with many of them already undergoing automation. However, more intricate tasks such as vendor selection, negotiation, and vendor management present greater challenges in terms of automation. While there will be increased automation of routine tasks, the complete automation of all procurement tasks is not an imminent reality.
The ultimate destination of Procurement AI remains unclear, but predictions have been made regarding the potential maturity levels it could attain:
While these scenarios are speculative, they represent potential outcomes based on current AI applications. The future of procurement hinges on its capacity to deliver tangible business value. The transformation of procurement aims to optimize procurement ROI (return on investment) in terms of cost savings, efficiency, collaboration, innovation, sustainability, and financial success. Measuring procurement ROI involves comparing the function's costs with the savings it generates, enabling further investments in research and development, improved customer experiences, sales enablement, sustainable offerings, and more. This is where AI has the potential to significantly enhance the impact of procurement.
Procurement professionals often have several questions about the integration of artificial intelligence. Here are some commonly asked questions and insights:
Should organizations adapt their processes for AI, or should AI align with existing processes?
AI can seamlessly integrate with existing processes, eliminating the need for extensive adjustments. The key recommendation for organizations preparing to leverage AI is to accumulate as much data as possible, even if not fully processed at the moment. A reservoir of data may unveil unforeseen benefits in the future.
Is AI in procurement exclusive to larger organizations, or is there a business case for smaller procurement entities?
Machine learning, considered a new form of software, is applicable across businesses of all sizes. While the software needs may vary for large corporations and small businesses, the emergence of suitable software providers makes AI accessible to all.
Is there a Forrester Wave or Gartner’s Magic Quadrant for AI in Procurement?
Currently, there isn't a dedicated resource like Forrester Wave or Gartner’s Magic Quadrant for AI in Procurement due to the broad applications of AI. The value of AI in procurement is primarily embedded in existing solutions, such as contract management or procurement analytics.
What does AI governance entail, especially in managing machine learning?
AI governance involves three human-machine collaboration models. "Human in the loop" requires human approval for machine decisions, suitable for critical or high-value decisions. "Human on the loop" involves human supervision of routine tasks performed by the machine. "Human out-of-loop" lets the machine run autonomously, applicable in scenarios like high-frequency trading in financial markets.
Is machine learning biased, particularly when relying on humans for training?
The effectiveness of AI hinges on the behaviour exhibited in training data. While biases can be mitigated in some applications, such as medical diagnostics, challenges arise in areas like recruitment, where human biases may be inadvertently incorporated into the AI process. To address bias, multiple data sources or rigorous bias vetting processes are recommended during training.
Where does AI fit with your business case? Download our build your own business case workbook.