How AI & Deep Learning can help in Supply Chain Decision Making By Krishna Khandelwal Chief Business Officer
2 2 WHAT IS ARTIFICIAL INTELLIGENCE Set of algorithms and techniques which can help machines in mimicking human behaviour. Recommendations
3 3 IT IS NOT NEW 7-8 Boom and Busts cycle Language translations, Intuition etc Failed in generalising solutions at scale Summer of 1956 at Dartmouth College in US where the foundation of Artificial Intelligence as a Discipline of computer science was first laid
4 4 IS THE BUZZ REAL THIS TIME? Commercially Deployed @ SCALE now
5 5 WHAT IS SO DIFFERENT THIS TIME? Structured Data Algorithms Compute Power Machine Learning Deep Learning
6 6 SUPPLY CHAIN AT THE CUSP OF DISRUPTION On an average around the world 180 Million Packages are delivered in a day. Certain problems are designed to be solved by Human Intuition Certain problems are designed to be solved by Algorithms
7 7 COMPONENTS OF ARTIFICIAL INTELLIGENCE Reasoning Knowledge Presentation Planning Natural Language Processing Perception Generalised Intelligence
8 THE JOURNEY A B 8 C
9 9 CHALLENGES AT EACH POINT A Customer placing an Order Free text address - High contextual component Unpredictable delivery timelines Customer and location properties
10 10 B Order Packed and Loaded CHALLENGES AT EACH POINT Multi model complex package movement Dynamism against static planning Resource planning Ever increasing storage cost Large Stock/Inventory/Warehouse management
11 11 CHALLENGES AT EACH POINT C Customer receives the Order Timely delivery of the package Lack of delivery proofs To deliver in first attempt Visibility to customer Cost vs Serviceability Finding the right location to deliver
12 WHAT DOES AI OFFER? A E C F D 12 B Magic at every point
13 13 WHAT DOES AI OFFER? Reasoning, Planning, Generalised Intelligence, Perception - Planning of orders done using AI Without AI With AI
14 14 WHAT DOES AI OFFER? Generalised Intelligence - Route planning can consider fuzzy human decisions
15 15 WHAT DOES AI OFFER? Natural Language Processing - Contextual understanding of address/text No universal format for writing Address PIN/ZIP codes have a very high error rate. Localities can be wrong too Delivered lat logs can be extremely wrong Landmarks are sometimes very local in nature Raw Address Abuse of free text address field Different language - Deciphering context is a challenge Different script Decipher the Address Information Content Local database may be unavailable Formatted Descriptive Co-ordinates Precision Type
16 16 WHAT DOES AI OFFER? Natural Language Processing - Contextual understanding of address/text Taufik Hidayat won his 1st gold medal in olympics on Aug 21st, 2004 Person Count Date Premise Locality Sub-locality [Shangri-La Hotel], [BNI, Kota], [Jl. Jend. Sudirman No.Kav 1], [Karet Tengsin], [Jakarta] [10220] Locality City Pin
17 16 WHAT DOES AI OFFER? Reasoning : Overlap of route clusters, routes choice (traffic - time vs distance) Knowledge Presentation : Road / Route constraints, # of docking bays, vehicle specific constraints etc Planning : Dynamic planning of load, inventory, fleet, resources etc Natural Language Processing : Contextual understanding of customer s comments, feedback, address etc Perception : Driver behaviour, location preference etc Generalised Intelligence : Intuitive load planning (Over-loading, SLA delays)
18 17 POTENTIAL IMPACT ARTIFICIAL INTELLIGENCE McKinsey estimates that firms will derive between $1.3trn and $2trn a year in economic value from using AI in supply chains and manufacturing
19 18 ARTIFICIAL INTELLIGENCE - AN AID Is it here to replace people? AI systems will interact with human beings in the real world only to make humans more efficient Google AlphaGo AI defeated Lee Sedol but Go players across the world are using this to improve their skills already
20 THANK YOU KRISHNA KHANDELWAL krishna@locus.sh www.locus.sh THANK YOU!