Intelligent Automation Week: Exploring the Ethical, Technical and Strategic Challenges of AI and RPA
Last year, as part of my role as an IT Consultant in the Version 1 Innovation Labs, I attended Intelligent Automation Week, 2019. This was a three-day event held in Twickenham Stadium, London, in which 60 industry-leading speakers discussed Robot Process Automation (RPA), Cognitive Computing and Artificial Intelligence (AI). Most speakers were C-Level executives, from companies such as Adidas, Ikea, Google, Ericsson, NHS, Ocado, Vodafone and more, discussing these topics:
- AI Tools and Technologies
- Data
- RPA
- Leadership and AI culture
In my first Medium post, I will be discussing my main takeaways from this RPA and AI-focused event.
AI Should Be Ethical
Companies are compelled to make use of big data. If standards and regulations are slow to develop (in comparison to the speed new technologies are developed), ethical principles must guide enterprise strategies to avoid the risks involved with Machine Learning (ML) and Artificial Intelligence (AI). Some examples of the risks involved are automating poverty or training ML with non-generalised data, leading to gender or racial bias.
Due to the complexity of AI algorithms, understanding the relationship between input and generated output is not straightforward. This is why some practitioners are starting to recommend a methodology called Responsible AI by Design to minimise the risk of undesired consequences of AI.
This issue is so important that the European Commission published a document in 2019, containing the Ethics Guidelines for Trustworthy Artificial Intelligence (European Commission, 2019).
The Culture Must Change
Relating to the previous point, AI requires cultural changes alongside technical ones. The technology is ready, but the organisations and the people within them are not. Sometimes, RPA is seen as the hammer and any problem within the company as a nail: this is wrong! RPA implementations can improve the business but sometimes they are just workarounds and the best solution would in fact be better integration within the business. Choosing what to automate is key; if a process is not good, it should not be automated. Instead, it should be changed.
Similarly, most improvements are tactical when they should be strategic. Innovation should be part of the existing strategy and not seen as an alternative/new one. The support from the top down is needed. IT must be fully supported by C-Level executives when the business begins its Intelligent Automation journey. On the other hand, IT within the business must educate employees so they appreciate the benefits involved with AI and understand how to use AI effectively within their roles. In Microsoft’s words (2019, p.12):
Employing a culture of AI will make staff at all levels feel empowered and create not just a business that employs AI solutions, but a truly AI-enabled organisation.
Intelligent Automation in the Public Sector
The case study shown was a recent implementation of an RPA system for the East Suffolk & North Essex NHS Foundation Trust (ESNEFT), where automation was employed to improve and verify the decisions medical professionals are making. The idea was to empower, not to replace, maximise staff skills, make staff more efficient and allowing them to spend more time with patients.
The process was challenging. As highlighted before, support from the top is vital for innovation projects at this scale, but without the OK from the board and no other examples of RPA employment in healthcare, there was nothing to show to senior management. Nevertheless, the CTO was able to set up a case study for automating invoice processing in days, using an intelligent automation platform (Thoughtonomy) hosted in Microsoft’s Cloud. In the first month, the system released 300 hours which grew to 4,500 hours a month after 1 year. Considering this time could now be used by staff for spending time with the patients and performing tasks they enjoy most, this use case is a great example of how intelligent automation can make processes more efficient and make employees’ lives easier at the same time!
IPA is not just a type of beer!
Despite its employment in the beer industry, IPA is also the acronym for Intelligent Process Automation. So, what does IPA exactly mean in this context? IPA is a general concept embracing and combining:
- Digital Process Automation (DPA): digitalising businesses and processes to become more responsive and customer-centric.
- Robotic Process Automation (RPA): handling high-volume and repetitive tasks with no human intervention.
- Artificial Intelligence (AI): in this context, Machine Learning allows systems to access data, learn for themselves and take autonomous decisions.
This is summarised by the following formula:
Today, RPA mostly works at the task level when it should be planned and designed at the strategic level. IPA is a much more general concept: an end-to-end process ideally leading to a digital ecosystem. Companies seeing IT as a cost centre are not ready for IPA — business and IT must work together to achieve a common goal.
From Data to Wisdom
Data is at the core of the new economy that Intelligent Automation is enabling. The lexical choices of new methodologies such as Data-Driven Decision Making (DDDM) highlight the centrality of data in the decisional process which must be supported by hard data, rather than based on intuition or observation.
To guide business decisions to better outcomes, raw data must be transformed and become more than what it is at the time it is extracted. The steps below outline this data journey.
Optical Character Recognition (OCR): Challenges and Future Trends
Optical Character Recognition technology faces several challenges:
- Unlimited and varied layouts: key information can be anywhere in the document, image, or message to be processed
- Poor quality of the scanned sources which are sometimes rotated on their axis
- Different formats: JPEG, PNG, PDF, Word or Excel documents
- Documents not having strict formats or following clearly defined rules: understanding, for example, the difference between debit and credit.
Traditional OCR technologies struggle with the issues above because of the way information is extracted from the documents: reading line by line, including useless information, performing post-processing checks following hand-crafted rules which are specific for each document type.
On the other hand, Cognitive OCR employs Deep Learning and Long Short-Term Memory (LSTM) to process any type of document in the same way. Cutting-edge algorithms for character detection and recognition allow the scanning of the whole document with one single pass (rather than line by line), focusing on only relevant information, improving accuracy and shortening the processing time. Needless to say, this is the future of OCR technology.
References
European Commission (2019), Ethics Guidelines for Trustworthy AI,
European Commission [online]. Available at: https://ec.europa.eu/futurium/en/ai-alliance-consultation (Accessed 19/12/19)
Microsoft (2019, p. 12), Accelerating competitive advantage with AI, Microsoft Corporation
