In chain drug today, there are few terms that are as buzz-worthy as artificial Intelligence (AI) and blockchain. Often used as solution terms like “the store of the future will be enabled by AI” or “blockchain will provide traceability,” both of which may eventually be true, there are few specifics into how they actually work and may be applicable. In order to better illustrate how these technologies function, consider the following example, where a buyer and a vendor reach a contract on a product purchase.
The contract details go into the normal areas of timing, conditions of acceptance and payment methods, in addition to terms and fees. All related changes that influence the execution of that contract are also captured. The product is then palletized, placed on a trailer and managed by a third party. From there, the product is delivered, accepted and placed in inventory. Now, expand that normal supply chain story to include the truck arrival versus reception, the total of the driver’s route history, the changing condition of the tractor and the trailer, the manner it was loaded, variable traffic along driving routes, road conditions, weather changes, and legal requirements for distances traveled, including mandatory driver rest periods. It can also include the days the driver has been on the road, ongoing third-party disputes, and witnessing and avoiding accidents along the route. Each of these variables have captured data that has a statistical impact on the desired outcome: an on-time and safe delivery of acceptable product. This data is also highly dynamic, changing over time and circumstances.
The example above demonstrates AI, as it expanded the shipment process into new data sources (competitive bids, for example) with constant reevaluation of existing sources and statistical outcomes in an automated and high-intensity statistical analysis. On the other hand, if all the details along with the agreements, conditions and responsible parties were stored securely online as an integrated object with all-party access, this would be a blockchain.
AI is not a substitute for the matrixes of existing processes that make up modern retail, rather it is a means of finding new ways of making automation responsive to changing needs. It is also not a single set of methods, but directly related to the context and method used.
• Machine Learning: This is the most common form used within retail and is based on leveraging very large amounts of data and outcomes against a set of existing processes. It uses statistics and probability to review new data sources and the likelihood that there is a correlation to take use. Machine Learning is often directly linked to an automated process such as replenishment, call systems or labor scheduling.
• Neural Learning: Sometimes referred to as mimicking “ah-hah” thinking, this is a means of taking data that may have no common intersection points, but have an aggregation of multiple new patterns of information, to suggest alternative processes and even outcomes. Often it focuses on metadata (data about data or the timing, size and dimension of a transaction) to create a new landscape of information. Geo-Location is one such solution where the relevance of metadata occurrences is overlaid into a complex solution, therefore optimizing a new route to a map with changing variables. Possible outcomes include the placement of merchandising within a store or timing of labor.
• Deep Learning is said to be the multi-layers of Neural Learning and Machine Learning as it breaks down data and observation into new definitions and alternative outcomes. It was originally intended to mimic human thinking processes of evaluation and intuitive leaps with hyper-intensive data crunching to expected outcomes and then those outcomes review to an optimal scenario build.
In sum, each of these processes has shown a need for a business review. For example, Machine Learning tends to select more narrowly defined data sets and criteria, which can rapidly lead to bias in the results. Neural Learning can suffer from initial algorithms aging poorly as data leads solutions elsewhere, while Deep Learning can lead to responses that are internally innovative but fail in the real world, such as diagnostic health tools. Despite these challenges, AI-assisted evaluations of health still have high potential in the chain drug industry, just not as a single automated system.
Blockchain is an area of interest to all of the intersecting parts of chain drug including health, insurance, retail operations, finance, patient control of information and procurement agreements. There is clear value in chain drug to enabling a very secure set of transactions that are viewable by levels of permission and stored as one easily accessible object. However, while interest is high, understanding is limited and usage beyond experimentation is very low. With little over a decade in use, it is still being redefined into solutions.
Blockchain originally emerged in 2008 as a concept where a large online public group would participate in, support and proof “blocks” of transactions so they could be added to a relational “chain.” Additionally, it was initially established to hold a transparent set of transactions that would have a value (bitcoin) defined by the group and unrelated to governments. Bitcoin is one outcome of this concept, but it is the only original concept that has gained industry attention as it provided a high degree of independent security and traceability to a set of changes and binding agreements. As it currently stands, blockchain can be roughly grouped into public, private and hybrid categories, which vary on the degree of open access and third party validation.
The original form of blockchain is public and has shown to be secure, but is expensive and time consuming to reach independent validation. Private blockchain is no different than current shared ledgers, which function quite well without the overhead. However, hybrid is emerging as a solution among major logistic groups that strive for the security of third party validations with private group rules.
Overall, AI and blockchain are complex concepts that will inevitably be integrated into the internal systems of retail and chain drug companies throughout the next decade. As such, it will be crucial for the industry to go beyond buzzwords to fully understand how these technologies can be used in generating successful business outcomes.
Dave Marcotte is a retail analyst with Kantar.