Basic machine learning technology is already used by some procurement applications in areas such as spend analytics and contract analytics. This is mostly limited to automating the processes of collecting, cleaning, classifying and analyzing expenditure data in an organization — to identify savings or paths to greater efficiency.
Today, procurement technology vendors are creating cognitive procurement advisors (CPAs) and virtual personal assistants (VPAs) that use natural-language processing (NLP) and natural-language generation to further increase automation and efficiency.
Skills are a vital piece of the puzzle and are often overlooked
“A procurement VPA can improve the end-user experience of traditional procurement tools and increase spend under management by guiding people to the correct purchasing tool,” said Magnus Bergfors, research director at Gartner. “A CPA can provide summaries, recommendations and advice in everything from supplier assessments and performance management, to risk management and compliance.”
Build a Platform for Success
The reality of artificial intelligence (AI) is that most organizations won’t create significant value unless they deploy the technology on top of the right platform, data and processes. At the moment, most organizations only manage part of their spend with procurement technology.
“The first step to automating spend management is to make sure that existing tools are both fit for purpose, and are used throughout the organization,” said Mr Bergfors. “Proper use of existing tools provides the platform for AI, as well as the data on which to train it.”
This means mandatory use of e-sourcing, and allocation of responsibility for consistent use and validation of spend analytics data to category managers. As a minimum, all contracts should be stored in a contract repository and procure-to-pay solutions should be properly configured and populated by relevant suppliers.
Another key area to address is categories of spend that are not suitable for existing solutions. Automating 100% of spend management is neither realistic nor necessary for most organizations. When a process is not automated, however, it should be because of a strategic choice rather than an oversight.
“Application leaders should prioritize investment in solutions that support automation for spending categories that are not well-served by current systems, said Mr Bergfors.
Human Learning Is Still Needed
Skills are a vital piece of the puzzle and are often overlooked. Application leaders who lack the right analytical and data science skills will struggle to understand the opportunities and limitations of AI in procurement.
“Organizations that don’t have dedicated procurement analysts will need to create this role, but it may be hard to justify a full-time position in smaller organizations,” said Mr Bergfors. “In that case, it is important to set aside time and budget to train citizen data scientists inside the procurement and sourcing teams.”
Once the right platform and skills are in place, an organization is ready to experiment with pilot programs in areas where there are existing use cases for machine learning and automation.
“We urge application leaders to explore three use cases first: spend analytics, contract analytics and risk management,” said Mr Bergfors. “The latter two use cases will require more-advanced quantitative skills to figure out what to measure and analyze, and how to score the resultant data.”