Artificial Intelligence: Your Secret Weapon for Increasing Efficiency in Procurement

Business and Technology Insights

Artificial Intelligence: Your Secret Weapon for Increasing Efficiency in Procurement

 
December 31, 2018

As artificial intelligence (AI) or machine learning based solutions are fast becoming mainstream in processes including security, predictive analytics, personalization of marketing messages, online searches, image recognition etc., businesses need to turn their attention to AI to increase automation and efficiency in procurement. Artificial Intelligence (AI) based smart procurement is characterized by its ability to read patterns, learn continually, and trigger corrective actions for optimal performance. In today’s digital world, retaining competitive edge in the marketplace depends on an organization’s ability to process business transactions at higher volumes and velocity. AI-led smart procurement solutions are essential in situations where human intervention is unviable in terms of effort and optimal decision making. The self-correcting ability of AI brings responsiveness to procurement by eliminating slower, and often redundant, manual interventions.

AI in procurement: Use cases

Efficient work allocation and exception handling

In the procurement process, an unpredicted spike in work load can make work allocation a challenging task, as it requires optimally matching complexity with the unique skills required for each task. There is a limit to which managers can scale their work allocation abilities based on tacit knowledge.

The handling of exceptions is a similar case. Manual exception handling involves three basic work flows. First, an item is allocated for processing without the knowledge of exception or error. Second, an exception, like an invalid purchase order (PO) reference, missing PO information, or even a PO line mismatch is noticed. Third, the purchasing document such as an invoice containing the exception is routed for resolution to an agent based on standard operating procedures (SOP) or the experience of the invoice processor.

How can you use AI to efficiently handle exceptions? You can train an AI-based system to read patterns from historical data to automatically allocate invoices based on the past performance of invoice processing agents. The system assesses efficiency and allocates work based on the correlation of factors such as percentage of errors, turnaround time of agents, critical nature of the invoice (utility or emergency), and even current load. AI also eliminates the need for manual inspection as it discovers exceptions using rule-based algorithms. In fact, AI can act as a learning system that is not bound by a rigid SOP. It reduces the three manual workflows to just one. The added advantage: AI-based work flow not only determines the nature of the exception but also prompts the best corrective action based on historical data.

Effective fraud detection and quality control

Can you use AI to increase efficiency in other purchasing scenarios? Certainly. Selective routing of requisitions using insights based on historical purchase records, such as nature of request (routine or outlier), pricing benchmark (lowest to highest), role of requester, seasonal variation of business, and so on is one such example. Such selective routing saves effort on redundant approvals, highlights relevant cases for approvers, and speeds up processes for superior customer experience. Similarly, you can more efficiently handle invoices that come without POs by training the AI engine to mimic human actions. For instance, it can be trained to combine invoice data elements and infer processing needs based on historical data, as well as derive the information needed for processing, such as accounting codes.

Organizations with very high volumes of invoices can use AI to detect fraudulent transactions. Standard organizational controls to prevent fraudulent transactions such as assigning greater accountability to executives for high value transactions and segregating duties might prove to be inadequate as transaction volumes increase. Process-enabled controls cannot correlate factors such as due amount, supplier, invoice date, or seasonality, especially when transaction volumes are high. AI engines can be trained to read large volumes of transactions, detect suspicious patterns, and zero in on fraudulent transactions.

You can use AI to pre-qualify items for scenarios such as quality checks, or even to synthesize responses to supplier queries. The current approach to quality checks has limitations because it is bound by a set of finite business rules, and high volumes demand quality checks based on sampling. AI can standardize quality checks by introducing a dynamic learning system that incorporates learnings from previous quality checks into future transactions. Similarly, manually responding to supplier queries becomes very tedious even with a certain level of automation. AI, on the other hand can intelligently streamline supplier query responses by proactively pushing alerts as well as synthesizing responses in real time.

The future of AI in procurement

Next generation procurement functions will be evaluated on the basis of their ability to drive higher value while delivering unique buying experiences. This is likely to drive rapid adoption of AI to streamline high volume processes and enhance efficiencies in procurement. Bottom line: AI will play a critical role in helping executives balance price with value while enforcing effective controls.

How well is your organization positioned to implement the ‘AI as the natural choice’ paradigm in procurement? 

Ravindra Lalas is associated with the Platform Solutions Unit at TCS. He has been part of TCS Accounts Payable Platform from its inception and works with stakeholders on product engineering, solution development, and transformation initiatives. In his professional experience spanning over a decade, Ravindra has worked on CRM, BI and analytics, and across industries such as Telecom, Travel, and Retail. Ravindra holds a master's degree in Industrial Engineering and graduate degree in Electrical Engineering.