Artificial intelligence is an increasing part of our everyday lives, powering our smartphones and the Internet of Things. But few people really understand what it is, how it works and more importantly why it is so important to procurement.
This document attempts to answer these questions and specifically address its benefits to the procurement process by addressing the following checklist:
- What is Artificial Intelligence or AI?
- What kind of AI is used – supervised or unsupervised?
- Which AI model drives the application?
- How accurate is the model?
- What is the accuracy of the results?
- How much time will it take to train?
- How much SME resources are needed?
- What are the expected benefits?
What is artificial intelligence or machine learning
The Oxford English Dictionary defines artificial intelligence (AI) as the theory and development of computer systems capable of performing tasks that normally require human intelligence, such as visual perception, speech recognition, decision making, and translation between languages.
For many procurement professionals, the language used in data science can be confusing. It’s much easier to explain by simply saying: powered by AI.
However, when investing in large-scale technology transformation projects, it is important to understand which element of the process is being powered by AI and how the AI itself works. What powers does artificial intelligence have?
Artificial intelligence, or machine learning, typically falls into two broad categories:
- The computer is presented with example inputs and their desired outputs. “Teacher” or “training set” data is used to establish a general rule that maps inputs to outputs.
- No labels are given to the learning algorithm, allowing it to find structure in its input on its own. Unsupervised learning can be an effective method to discover hidden patterns in data.
Both unsupervised and supervised learning can be used to create basic behavioral profiles for different entities, which are then used to find meaningful insights and anomalies. In the field of procurement data analysis, machine learning is a method for developing complex models and algorithms suitable for the following tasks or processes:
- inventory management
- Bill Payment – Bill Fraud
- Supplier Relationship Management
- sales pipeline
- Marketing Analysis
- customer segmentation
- On time delivery
- Operational KPIs
- Supplier Onboarding
- Spend Analysis
“Powered by AI” is a common attribute or term used to sell the benefits and merits of digital transformation solutions. The goal of these analytical models is to provide procurement teams with reliable, repeatable decisions and to learn from historical relationships and trends in the data. So it’s important to have a general understanding of how it works and what drives the AI.
What is the real power behind AI?
The power behind AI is a set of structured learning algorithms or codes used to analyze input data. Open source coding software such as R or Python is often used to develop the AI power these software models use. Within these applications there are several models or libraries that can be used to “power” the AI.
Below is a list of the most popular decision libraries that Power AI uses
- learn decision tree
- Learning of association rules
- Artificial Neural Networks
- deep learning
- Inductive logic programming
- Support vector machines
- Bayesian networks
One criticism leveled at AI models is the term “black box,” which has been compared to an airplane flight recorder. It is true that AI models are not always transparent. It can be difficult to understand how they derived the results and how accurately the calculation or prediction was made.
Construction and accuracy of AI models
A model can be more than 95% accurate and can be optimized using multiple iterations, called epoch(s), to achieve stable functioning. This does not mean that the output of the model is accuratesimply that the model itself is stable and works as it should.
Data training and validation time
The model can take several hours to complete. However, it can take days or even months to get accurate results and there is a risk of overfitting the model. In addition, it may be necessary to reserve a specialist resource in advance to validate the results.
Accuracy is critical and in many cases depends on the training dataset itself. The training set must be checked carefully to ensure that there is no data bias. Bias occurs when one or more groups appear more frequently than others and skew the results, creating bias.
Repetitive tasks that are predefined and use a flow process called Robotic Process Automation that replicates manual tasks make uploads faster, the process smoother, and produce very accurate results.
What does artificial intelligence do in summary
For most buyers, these models appear complex. Models that initially require highly skilled data engineering and data science experts to write the code are costly to produce. For many, AI mirrors the same sentiment that powered big data a decade ago, with its multiple success use cases and sourcing benefits.
With the proliferation of marketing hype and obfuscated technology “black box” solutions, it is more important than ever to fact-check and get clear answers about how AI is being operated.
The goal of procurement digital transformation is to use technology to perform tasks and processes more efficiently to seamlessly deliver benefits without disrupting ongoing operations. As with many sourcing decisions, it is important to understand the benefits of AI, the rationale for sourcing, and the return on investment.
About the author
Edward McGeachie has over 25 years of procurement and supply chain experience at Schlumberger, IBM & Lenovo PC Division. He led global procurement, operations and logistics teams before founding a center of excellence for Lenovo GSC supply chain analytics.
Co-founder of Seaforth Analytical Services, UK based supply chain analytics company. He holds an MBA and B.Eng.(hons) in Industrial Engineering and is also a certified Black Belt in Lean Six Sigma.
He has developed an innovative spend analytics platform using a hybrid of the processes described above called Accelerated Insight®.
Microsoft Azure and Google TensorFlow are also great places to explore how to program AI.