The wave of interest in AI has highlighted the importance of Data.
Where does Automation fit in the Data World?
Data Lakes, Data Warehouses and Big Data
There are various definitions for the terms, but their usage means lots of data files, data tags and descriptions plus large numbers of data movements getting data in and out of the storage facilities.
The creation of AI models requires large quantities of data for the training activity, the testing and re-testing of results. Remember AI is essentially “Statistics” and some mathematical formulae.
The handling of large collections of data would not be practical without effective automation.
Automation of Processes
When business processes that have been performed manually are automated, there is the potential for the automation to record a lot of extra processing data which asking a person to do manually would have been too much of a burden.
Once such processing data becomes available it can be monitored for changes and analysed for trends.
For businesses that progress beyond the automation of a single process, the ability to orchestrate the automation software robots to perform the right task at the right time becomes essential.
Generally, the total capacity of automation software robots exceeds the initial processing requirements used to justify the implementation of the automation. Using the spare capacity to perform low priority automation tasks such as “Data Reconciliation” can add real value.
As the number of automated tasks grow, the complexity required for the orchestration grows. The orchestration can be designed to use the automation processing data. In a scenario where an automation is failing as an external resource is not available, it might not be appropriate to keep attempting to execution that automation so frequently but to use the automation capacity to execute more occurrences of a lower priority task until the external resource is found to be available again.
Of course, not every scenario can be anticipated and be built into the orchestration but the availability of data does offer some possibilities.
The UiPath technology that we utilise for automation solutions makes use of Snowflake to hold all of the automation processing data that is created by the software robots in order that orchestrations can be built to respond to situations as they occur.
AI with modified Data – aka Machine Learning
AI models are created by training them on large quantities of data. The generic models such as ones for extracting data from Invoices are good, but they are “Generalised” by design. It means that a good result is delivered in the extraction of data most of the time. AI is using “Statistics”. When the invoices being process by the AI model do not result in a good extraction of the data, a person will need to perform the extraction.
Manually extracting the data for the odd invoice is not usually a big issue, but if it is say 5% of 1,000 invoices then it is a chore.
Automation solutions like UiPath , provide the option to implement “Machine Learning” as an enhancement to the AI models used for such processing. This provides the capability for the invoices to be processed manually by a person, the actions taken by the person to be captured with the Machine Learning and used to enhance the model.
Processing one invoice manually, will not mean an immediate improvement in the results from the AI model.
Why? Statistics drives AI, hence multiple manually processed invoices need to be captured by the Machine Learning to have any effect. The workflow for this enhancement of the AI model can be created as part of the automated solution.
The processing analytics for such AI solutions should show a continuous improvement for regular invoices that are similar to ones which have been previously processed. Don’t expect it to be 100%, remember it is “Statistics”.
Will Semantic AI enhance Data handling?
UiPath have leveraged their extensive AI skills and technical knowledge of document understanding plus communication message understanding to create the “Ultimate Copy & Paste” functionality with “ClipBoard.AI”. It is in Beta / Early release.
Semantic AI type capabilities offer a better understanding of Data by using its Context. This will reduce the effort required to “Label” data which can be a manually intensive process and enable more automation to be created with less effort.
Automation’s Relationship With Data
Automation is contributing a lot of data into the world and Automation is essential to the handling of the vast quantities of data that are available, it is a symbiotic relationship.
#businessbeyondautomation
Article Author
David Martin
Managing Director, Ether Solutions
https://www.ether-solutions.co.uk/
