Co-authored by Andreas Freakley (primary author) and Pawel Krzysztofik
Click here for Part 1
Most organizations have data available that will allow them to streamline their processes or generate new insights using artificial intelligence and automation. Automation-focused modules can be facilitated for streamlined, mature processes with little exceptions and strong data structure, for example, where a company is already manually unlocking data and employees perform a large number of repetitive tasks with little need for decision making.
On the other hand, artificial intelligence or machine learning algorithms can provide more cognitive approaches and help to unlock data with high levels of complexity and unstructured data sources. AI can work independently but can also collaborate with human experts to improve quality or ensure legal requirements.
The Infosys Consulting Building Blocks Framework
Our artificial intelligence and automation building blocks framework is a strong tool to help a company structure its unlocking approach and identify the right technology for specific tasks along the process.
The framework consists of 5 main areas, which group algorithm types based on the nature of their expertise:
- Conversational | Interact with humans and facilitate information exchange
- Process Oriented | Streamline and scale processes and drive efficiency
- Decision Oriented | Analyze data and act based on the outcome
- Expert Oriented | Assist subject matter experts and provide input for decision making
- Explorative | Future oriented, identifying solutions with unclear answers
Phases of the data unlocking process
The process for unlocking data follows four main phases with a clear outcome. The phases can be associated with the artificial intelligence and automation areas from our building blocks framework (above) and have individual KPIs.
- Extracting Data
The extraction phase focuses on extracting data from the data source and making the data machine readable. The important factors for this phase are extracting the information from the source documents with accuracy and completeness, while preserving the original structure of the document. Depending on the data source, specific algorithms often deliver agreeable results for only one of the mentioned factors. To achieve optimal results, multiple approaches can be applied and the results combined.
- Analyzing and Mapping Data
The goal of the mapping phase is to identify relevant information within the extracted data and map it to the respective target data fields. Depending on the complexity and the number of anomalies within the source data, a fully accurate mapping may not always be achievable. In this case, the machine learning algorithm can provide mapping suggestions to the subject matter expert and, based on human decisions, improve its confidence levels over time. Through establishing such a feedback loop, AI can increase its accuracy over time to a point where human quality checks may not be needed anymore.
- Checking the Data
The outcome of the mapping phase can be assessed by a human in the data check stage. This can be done to achieve multiple objectives:
- Ensuring data quality
- Fulfilling regulatory requirements or
- Taking final decisions on mapping suggestions and potentially improving the algorithm
As this is the stage with the largest part of human interaction and could be a bottleneck in terms of scalability, it is important to design the interface in a manner that the highest data quality is maintained while optimizing handling time.
Often, manual data quality checks can be reduced after the initial configuration of the AI system once it is certain that the algorithms produce output with a certain level of accuracy and confidence. At this point, only random data sets can be checked or those identified as containing algorithms.
- Leveraging the Data
While this stage per se is not part of the unlocking process, having a clear idea how the data will be leveraged once available is probably the most important part of the initial assessment.
The first step is to run an initial design workshop and have a closer look at the business value and define the problem statement. Both business and user assumptions are used to detail out the target vision and a first set of key performance indicators are defined for future measurement. Using the Infosys AI&A building block framework, the initial draft is then broken down and fitting algorithms are matched to the initial architecture.
As an added benefit, needed roles and resources can be easily identified through this high-level view. After a dependency check regarding existing infrastructure and technology, the concept can be adjusted and then acts as a roadmap to unlock an organization’s data sources.
Proofs of concepts to assess the feasibility can usually be provided by a small team of data scientists in a few weeks and can be then quickly scaled across an organization using the results of the early assessment phase.
Given the rapid developments in the markets right now, this is the right moment for organizations to explore the possibilities and leverage the value behind AI and automation.
Want to learn more? Get in touch with our AI experts today.
Co-authored by Andreas Freakley (primary author) and Pawel Krzysztofik
Associate Partner, Infosys Consulting
Dr. Harald Gunia is an enterprise architecture and artificial intelligence expert. He has more than 29 years of experience in all major AI technologies, including, machine learning (ML) and robotic process automation (RPA). He was worked in more than 10 industries and has deep expertise in digital capabilities and large-scale business transformations. Harald holds an M.Sc in computer science and a PhD in artificial intelligence.
Principal, Infosys Consulting
Pawel has more than 11 years of experience in artificial intelligence, consulting and data science. He has worked in consulting, financial institutions and the chemical industry, where he grew his skillset in data science and artificial intelligence. Pawel heads up the Infosys AI Labs department in Europe where he leads a team of data scientists, RPA developers and machine learning experts. He delivers value to his clients starting from exploring AI potential through prototypes to implementing scalable solutions. He holds a BSc. in computer engineering from Wroclaw University of Science and Technology and a BSc. in financial management from the Manchester Metropolitan University.
Senior Consultant, Infosys Consulting
Andreas has more than 14 years of experience in disruptive technologies, consulting and account management. He’s passionate about driving value by leveraging artificial intelligence and automation, and exploring new business opportunities or creating environments where both internal and external process experience is improved by working hand in hand with a virtual AI-powered colleague. Being part of our global AI&A practice allows him to focus on proving the value of machine learning and connected technologies to organizations taking their first steps in this domain. He leads small and agile teams that can quickly build prototypes and deliver tangible results using a company’s own data. He holds a BSc. from Frankfurt School of Finance and Management and is a certified application scientist.