

With HTC, a major automobile manufacturer leverages AI to streamline its sourcing and procurement.
THE CLIENT
The client is an Indian subsidiary of a multinational automobile manufacturer.
THE PROBLEM
The client was facing recurrent interruptions in their manufacturing process, leading to reduced revenue accretion. The problem was recognized when engineers were confronted with supply issues pertaining to the purchase of automobile parts. Non-ideal, wrong selection of suppliers from the global suppliers’ database (GSDB) triggered deferrals in receiving the sought-after parts within the required timelines. It led to an irredeemable loss of time in the procurement stage. Build costs overshot the budget, as the parts did not arrive at the manufacturing facility on time.
Further, the loss of time had a direct impact on the production cycles. To sum up, they had to deal with:
• Delays of one to more than four weeks
• Inability to proactively recognize and exclude erroneous codes
• Failure to map GSDB codes correctly.
We embarked on an exhaustive examination of procedural obstacles and developed a solution using artificial intelligence (AI) and machine learning (ML) techniques. In the process, we:
• Utilized historical data of successful POs to train the system
• Deployed a self-learning AI algorithm to send prompt notifications to sourcing teams when incorrect GSDB codes were fed into the system
• Leveraged an AI-based ranking engine to provide an estimate of the site code accuracy
• Collated and codified past data of spare part dealers and fed them to the tool along with a detailed breakup of codes used for the different parts
• Conducted a feasibility analysis to develop a proof-of-concept (POC)
• Derived detailed percentages of successes or probabilities for different suppliers of parts and inputted them into the self-learning AI algorithm to inform sourcing teams about possible incorrect GSDB codes
• Used the AI ranking engine to prepare a list of accurate site codes based on all the probability percentages achieved, in a descending order to arrive at the most accurate supplier for the part (trials were made to train other algorithms but none were as accurate as Naive Bayes).
With the ML tool doing the job of selecting the most ideal supplier based on the given parameters, the purchase department is now channelizing greater time and attention towards resolving supplier-end issues. The flow of the parts
became smoother. On the other hand, with the proper utilization of time, labor hours were utilized more effectively and the whole process could be completed within the desired time limit. Also, finishing the job within the planned schedule led to an expected rise in profits. However, the most obvious benefits include: