Robo-processing is changing business processes in logistics. Add AI to create a smart digital worker.

Whenever I mention process automation in logistics industry, I hear the usual complaints about software vendors overstating capabilities and papering over the gaps between what the robots could do and what they can actually do. My interlocutors are especially scathing of software tools pitched as artificial intelligence capable of easily replacing variety of administrative jobs in logistics companies.

To an outsider, logistics processes are nothing else but a long string of repetitive actions centered around sets of documents required by internal business processes, external partners, and compliance with regulations. Not surprisingly, visions of humans repetitively receiving, reviewing, approving, sending, and archiving reams of paper feature big in discussions about work automation. Possibility of human error getting introduced in the paper-based process is very high. Each error results in complaint, dispute, time-wasting search for resolution and impacts measured in real money. But can the robots really do better?

The (ro)bots are here

Employing a software bot in place of a human to prevent errors is the typical pitch heard from robotic process automation (RPA) software providers. Bots don’t get tired, bored, discouraged or malicious and they sound like an answer to COOs and CFOs worries of high administrative costs of delivering logistics services. Value of RPA has been validated in many industries of which financial services provides the most vivid examples of RPA’s usefulness.

In essence, RPA tools perform programmable scripted actions corresponding to IF, THEN and ELSE statements. Data leveraged by RPA can come in both structured (e.g. EDI message) and unstructured (e-mail, scanned PDF attachment to e-mail) formats. The scheduler component of RPA programming allows execution of scripts at prescribed time and with prescribed frequency. A non-programmer can write the script in natural language, making the RPA tools very accessible to many knowledge-workers.

The most common implementations of RPA bots in logistics companies today are those dealing with digitized documentation containing structured data and text-decoding recognition technologies like optical character recognition (OCR). Upon identifying structured and unstructured data in target documents such as RFPs, purchase orders, BoLs, or invoices RPA bots are able to correctly recognize the nature of those documents, extract key data fields and words necessary to perform document classification, determine any missing data, and launch a workflow leading to actioning the document.

There are multiple examples of projects completed by freight forwarding companies using RPA to provide data, price quoting and booking functionality via customer portals. They monitor purchase orders, perform order handover optionally adding internally-stored data/instructions, and automatically allocate the order among multiple logistics partners based upon geographic Origin and Destination information, cargo type, and cargo attributes (dimension, weight, etc.). Some 3PLs combined RPA with cargo sensor monitoring systems allowing RPA bots to generate “cargo visibility” notifications to customers and handling the much dreaded “Where is my cargo?” customer inquiries.

There are about 200 software companies selling generic RPA tools. Major players include UiPath, Blue Prism, Automation Everywhere, Appian, or RPA Labs, but distinctions between #1 and number #30 are rather insignificant. Don’t expect any out-of-the-box templates specific to logistics or supply chain management. Writing the scripts is not difficult but building integration for the bots to retrieve data from other internal or external systems will require help of IT professionals.

As logistics and supply chain professionals like to point out, the processes they deal with are unpredictable and full of exceptions or exceptional circumstances not perceived from the data contained in the documents they deal with. Insisting on humans being better with uncertainties and exceptions is the biggest barrier to deploying software bots beyond the most rudimentary arrangements of actions.

The smarter bots are here too

To be very clear, RPA is not artificial intelligence (AI). This is an ongoing source of confusion, as innovations in machine learning (ML) branch of AI allowed some RPA software companies propose solutions integrating Natural Language Processing (NLP branch of AI) and machine learning (ML branch of AI) with their RPA scripting engines. In context of these solutions, NLP or ML are not making RPA operate like human intelligence capable of multi-sensory perception (say, tone of voice in a voice mail message), rapid classification of what constitutes valid or invalid exception, accepting uncertainty, and predicting outcomes based on incomplete data.?While NLP or ML cannot provide that human-like general thinking capability, they have enabled the bots to “learn” with the help of humans or other bots by continuously expanding and enriching the dataset used to train the AI behind the RPA bots.

Due to the NLP or ML components, the implementation cost will be less predictable than that associated with deploying the more traditional RPA tools. However, logistics companies can extract greater value out of the solutions combining RPA and NLP or ML. Think of completely automating the customer service function or a bot-driven ability to assemble a quotation for service. Going a bit more sophisticated, with ML-driven price optimization behind the RPA, the system can establish similarities and determine the price band or a specific price point at which the customer is most likely to accept the quote.

Among the RPA companies which include ML with their RPA platform, there are a few which provide shortcuts for logistics and supply chain processes. Not one of them can claim automating all possible processes and administrative tasks happening between the capture of the quote request all the way to successful settlement, but having some process templates included with the product will reduce your startup cost. Look at RPA offerings from Shipamax, Smart RPA, Gleematic, or Cigen for good ideas.

Keep in mind that your transformation team will definitely want to automate processes other than those handled by these companies. If their generic RPA features inhibit ease of use (wizards and GUIs), speed of editing and configuration, or building business-specific extension, your business may be better off building solution on top of an industry-agnostic tool like Blue Prism or others already mentioned.

Even smarter bots are here?

Many tasks associated with moving freight require thinking, gaining situational awareness and analysis. Thinking and situation assessments are general, while the analysis and specific task execution is specific. That’s how to contrast discussion of RPA, ML, and AI:

• RPA is about getting the computer to perform a task that requires known predictability of the input, decision and output.
• Adding ML allows to work around unpredictable outputs and some unpredictability of inputs.
• Adding AI enables computer program to analyze information and mimic the human mind’s capacity for assessing circumstances of the situation and demonstrate critical judgment.

The deep learning branch of AI is now being tested by a handful of 3PL companies and a couple of e-commerce companies with big aspirations in logistics. Those solutions, custom built for specific purposes, combine senses (hearing, vision, touch), cognition (intelligent decision-making), and RPA technology to create robotic life-like assistants for business decision-making and business processing. Combination of RPA, ML and AI opens doors to complicated decision-making work such as auxiliary decision-making, identification of mutually beneficial resolution(s) and undertaking autonomous actions in conjunction with other systems. As the AI is being pushed all the way to the edge devices, e.g. autonomous logistics center robots, capabilities and sophistication of decisions made through RPA will increase rapidly.?With sophisticated AI behind the RPA, the bots will gain cognitive capabilities and ability to continuously correct their own behavior(s), a welcome solution in very dynamic area like logistics execution.

The implementation costs of those most advanced solutions put them beyond reach of most of the companies involved with logistics and supply chain execution. The good news is that implementing RPA in conjunction with ML today will prepare logistics business to deal with the advanced technologies of decision-by-machine already on their doorstep. Transforming your semi-automated business processes with intelligent robotic execution not only enables your business to reduce costs and reduce errors within your own processes and in interactions with your logistics partners and customers. It also adds temporary competitive edge to your service offerings and agility to fulfilling specialized requests from your customers. And that means business growth and increased profitability, something that logistics businesses have often hard time delivering for their stakeholders.

?

If you found this topic interesting, leave your comment or suggestion. Don't just “like” it, share it with your network. I will be happy to listen to your comments and respond, if needed.

On a broader note, I work on innovative applications of mathematical optimization to solve real life business planning & execution problems. All my writings draw on real life business experiences with my clients. Real experiences feature big, as nothing beats practical confirmation of theoretical solutions. If that interests you, please follow?me?to receive the latest updates and contact me for consulting project opportunities.