Artificial Intelligence – Spend Analytics
From our experience, every data driven project will have several stakeholders asking “where is the AI component in this” or “does AI not do this automatically with 95%-100% accuracy”.
In order to understand what to reasonably expect from Artificial Intelligence, we must first appreciate what we mean by AI.
"AI is an umbrella term for a broad set of computer engineering techniques, ranging from Machine Learning (ML) and Rule-Based Systems to optimisation techniques and Natural Language Processing (NLP). We need to be asking which components of AI are relevant and effective for which phase of our projects."
In our experience in delivering large scale/complex spend analysis projects, successful use of AI is achieved through a combination of Rule-Based and Machine Learning approaches. The first allows for the subject matter expert (SME) / category expert to set the foundation of the analysis, and the second allows for the dynamic discovery, using both supervised and reinforced learning to associate and classify the data in ways that had not to date been fully appreciated.
The myth of a standalone ‘unsupervised’ computer program being able to intelligently analyse and classify data independently of external intelligence, is similar to expecting a French to English translation engine being able to accurately and automatically convey the emotions and meanings of a French text to an English audience without any supervision.
"In order to understand what to reasonably expect from Artificial Intelligence, we must first appreciate what we mean by AI."
Our tried and tested approach to AI has served our clients well in gaining an accurate appreciation and understanding of the genuine opportunities hidden within their global spend data. Our Spend Module has been equally successful for clients with either a single instance of data or for those with over 100 disparate sources. The flexibility of our analytical platform allows for the refresh of data using AI to occur intra-day, weekly or monthly from all sources.
We continue to work closely with teams in both academia and Microsoft Cognitive Services. As we uncover techniques and approaches that genuinely add value to our clients’ data analysis objectives, we are quick to embed these into our platform.