In the first part of this series of articles about the contribution of AI to order automation, I presented the idea of an automation solution based on machine and deep learning — less exhaustive, but much faster and cheaper than EDI automation. In this second part, I will focus on the conceptual, even philosophical differences between the two approaches, and also discuss how they could be complimentary.
EDI and AI: different visions of automation
At the end of the twentieth century, the rise of office software and automation gave rise to an “all-computer” vision of business. As its name suggests, EDI (Electronic Data Interchange) corresponds to this vision of a structured, exhaustive and perfect world where digital data is transmitted between companies from machine to machine, replacing human exchanges based on paper documents. In order to simplify their back office and reduce costs, some companies, mainly those with unequal power relationships, have tried to apply this technical vision by forcing their customers and suppliers to send and receive their orders via EDI. Aiming at rationalisation, efficiency and above all, commercial exchanges control, several national regulations are accompanying the switch to EDI, mainly in Europe but also in other continents.
As opposed to the EDI philosophy of completely dehumanised exchanges, AI is different. With the fundamental principle being to imitate human intelligence, it is easier to imagine a collaboration between machines and humans. AI already helps humans to perform tasks as diverse as travelling, detecting diseases or entering orders in an ERP system. Of course, this is not really what is put forward in science fiction books and movies where artificial intelligence has a strong tendency to quickly become independent from its creators, or even outright alienate or exterminate the entire humanity as in the Matrix or Terminator movies.
But we are still a long way from a strong artificial intelligence, capable of learning and understanding new things by itself and therefore having emotions, or even consciousness. Current artificial intelligence are said to be weak and are ultimately only a concentrate of human cognitive capacities in extremely specific domains. Returning to the automation of commands, these specific tasks can be the identification of key information in a business document, the detection of an anomaly in this information, or predictions based on thorough data analysis. At the end, where EDI aims at a complete replacement of human interactions, AI seeks to instead help humans and make their tasks easier.
A required flexibility in the order taking process
To thrive in the 21st century, businesses must be flexible and adaptable. The loyalty relationship between a customer and its supplier is less and less based on the price and quality of the product or service and more and more based on the customer experience. Customers want suppliers that are easy to work with. They want to place orders easily, without having to comply with particular rules or formats. Customers also want to be able to make changes to their orders or get advice and recommendations on products of interest to them. Exchanges between trading partners therefore require scalability, adaptation and even interpretation.
However, EDI systems based on a technical approach fundamentally lack the agility to meet these requirements. After the cost of implementation, the significant maintenance linked to the low adaptability is the second reason why company exchanges have not been completely automated with EDI. EDI now accounts for a large share of B2B order exchanges, and although the technology has been around for a long time, it did not gain a global foothold, and the increase in the percentage of EDI orders remains limited.
AI to improve EDI exchanges?
EDI has a role to play in the automation of order flows, and AI is not intended to replace these computerised exchanges. On the other hand, it can compensate for the deficiencies of EDI and provide agility to a digital but rigid process. As with orders sent by fax or email, AI adds a layer of intelligence to the EDI transport layer. For example, it can automatically detect anomalies based on a customer’s order history and prevent an order with unusual and potentially incorrect quantities from being created automatically in the ERP system, therefore delivering it without human verification. An error in a quantity or unit of measure can have significant impacts for the supplier between the goods return (or even their destruction if they are perishable), the generation of a credit note for the customer, and potentially the degradation of a business trusting relationship. AI also solves a recurring problem in EDI related to product references. In standard EDI, the customer’s EDI system must always be up to date with the supplier’s article references and as soon as the supplier’s catalogue changes, the reference data must be updated, either by EDI or with more manual steps carried out by internal IT. An intelligent system makes it possible to view EDI orders that are in error due to incorrect product part numbers in a format that the user — not just the machine — can understand, and Customer Service Representatives (CSRs) in charge of processing orders can easily correct the part numbers, just as they would on fax or email orders. Beyond empowering business users, machine learning can transparently identify these corrections and automatically apply them to future orders from that customer, again with the goal of reducing the amount of manual and repetitive work. In an EDI-only process, the human must adapt to the computer language to handle exceptions, whereas, with AI, the computer adapts to the human language, to the benefit of the business users in charge of processing orders.
EDI + AI = efficiency + flexibility?
AI allows companies to enrich an EDI order process by adding a functional layer to speed up and facilitate order processing, while providing CSRs with total visibility on all orders, regardless of their source (email, fax, EDI, portal, etc.). Beyond this transparency, these business users also benefit from the flexibility of a platform that leaves room for their human intelligence to manage special cases and adapt to change, as opposed to a fully automated but purely IT-driven EDI process with no room for manoeuvre. By restoring visibility and agility to an opaque and inflexible technology, AI puts people back at the heart of company exchanges and even offers a competitive advantage by improving the customer experience.
If you haven’t checked out Part 1 of this series, be sure to go back and explore the history of EDI and AI.