In this article:
- Blue Yonder
- Freight Tracking Evolution
- AI Example Use Case
- Removing Unnecessary Complexity
Blue Yonder permalink
Last week there was a major announcement that JDA is re-branded to Blue Yonder.
In 2018, JDA purchased BlueYonder, a data science based company that was utilising AI and ML algorithms to help organisations make better supply chain decisions, with the promise if improved efficiencies.
With the re-branding, we will now see the focus shift completely to be marketing the powers of AI and ML. It says a lot about the impact this tech is having on the logistics industry, even if it is just marketing impact for now.
We are yet to see if/how this technology will be used to enhance JDA TMS, however, with the re-branding of JDA to Blue Yonder, we can be assured that there will be some interesting developments in the years to come.
But what will they be, and how can you get ready for them today? There is much that can be done today, to get ready for the coming wave of AI and ML. As the old saying goes, failing to prepare, is preparing to fail, and in the AI space, it seems Gartner think failure is even more likely, with the bold prediction that 85% of all AI projects will "not deliver" for CIO's.
For many businesses, realising the benefits of AI technology is years off. But it will be critical for companies to ensure they have the correct foundations in place, to take advantage of AI and ML in the years to come. It is said that data scientists spend only 20% of their time on algorithms, and the rest is spent preparing and cleaning up data.
Digitisation is a word that gets used a lot in the IT industry. Companies who have higher levels of maturity in the digitising their processes will be best placed to take advantage of AI. But what exactly does digitisation mean? Basically, it just means, turning information into an electronic format. Wiki says:
Digitisation means converting information into a format that can be processed by a computer.
Freight Tracking Evolution permalink
Digitisation though is not a binary step. it takes place in stages, and evolves over years, and decades. Lets take freight tracking as a example, and look through the some of the stages of digitisation we have seen over the decades.
- Bar-codes are added to manual consignments, to allow for scanning within freight facilities. Bar-codes only started to be used in supermarkets in around 1974. They are now on every product we purchase, and every item, package, box and pallet that moves through the supply chain.
- In the 90's, with the advancements in mobile communication, drivers are provided with MDT's to receive pick-up instructions, and also to enable consignment scanning at pick-up and delivery points. This also enabled generation of electronic POD.
- As the internet finds its ways into the majority of businesses and warehouses, consignments labels are generated for each item at the despatch location , removing the need for manual data entry for billing purposes, allow for partial POD's and item level freight tracking.
AI Example Use Case permalink
Lets look at an example use case where an organisation is wanting to take advantage of AI to enable forecasting for the number of drivers required to execute deliveries in the afternoon, and provide accurate ETA's to customers. Tee inputs into the algorithm could potentially be the following(although more would likely be needed):
Historical consignment data
Consignment data (destination info)
Consignment item details, including current location
Driver capabilities - vehicle capacity
Potential Issues permalink
Lets take a quick look at some of the potential issues a project of this complexity might face.
1. Manual procedures
Even though your organisation has digitised much of its operations, it still has long standing customers who are not able to provide consignment data electronically in time for the AI algorithms to run, and decisions to be made.
Possible resolution: Allow for this "unknown" quantity in the algorithm, and utilise historical data to forecast consignment data
2. External dependencies
The algorithms accuracy is now dependent on the availability of information from many sources. Any delay in receiving information will effect the accuracy of the forecast.
Possible resolution: Build in fall-back options for every data source and enable user input for external figures
3. Missing parameters
If the algorithm does not include critical parameters that have an impact on the number of drivers needed, the accuracy of the system will not be trusted by operational personnel, and the system will not be used.
Resolution: AI algorithms must be designed with operational people. This is a huge change management challenge, as these systems are seen as a threat. People will hold onto critical information if they are not part of the solution.
For potential benefits to turn into real benefits, appropriate scale must exist to ensure that the costs of implementation are lower than the savings to be realised by the organisation.
Resolution: Engage with experts with a proven track record in the logistics industry, as well as your geography.
5. Unnecessary Complexity
Complexity come in many forms. If you are utilising multiple internal systems to capture consignments for example, or customer forecasts are in different formats, this added complexity will mean data scientists will first have to interpret and standardise data before the AI algorithms can run.
Resolution Remove complexity where possible. Not doing so will only add to the complexity of your AI projects.
Removing Unnecessary Complexity permalink
Organisations who have removed as much complexity as possible will be in a far better position to take advantage of the latest AI technology, faster and cheaper. They will also be able to evaluate the benefits of technology faster, meaning they can fail faster too. If an organisation must spend $X millions of dollars getting data standardised before they can run a POC or pilot site, they will always be playing catch-up. Some examples of unnecessary complexity:
- Workarounds on-top workarounds
- Non-standard processes/lack of documented processes
- Multiple systems supporting the same/similar processes
Don't expose operations to greater complexity permalink
The example we are looking at relies heavily multiple external factors. In a real world example these would likely increase, to also include sub-contractor availability, warehouse and sort facility integration and would continue to increase in complexity before it was able to produce valuable input to operations. If you are going to be impacting operations, ensure AI is complimentary to existing processes, and not replacing them.
Start with Value added Services permalink
Lets go back to our original example of freight tracking. One of the areas the logistics industry struggles with across the board is tracking freight through depots and cross docks where predominantly manual processes exist. Organisations could serve customers better if they had a vision based system which could track freight in real time from the moment it is unloaded from a truck, to when it leaves the facility. This type of computer vision system has been used by Amazon Go supermarkets to remove the need for checkouts! Such systems could provide valuable input to operations and customers, and the solutions could grow in complexity over time.
- AI is extremely reliant on a healthy digital diet. Make sure you are feeding it well.
- Remove unnecessary complexity where possible before embarking on bespoke AI solutions.
- Utilise AI to add value and limit direct impact on operations.
It is going to be interesting to see how AI branded companies and projects start to delivery real value for customers. in the mean time, get your digital foundation in place and reinforced.