part0010

Archaeologists excavating a cave on the coast of South Africa identified the remnants of pigments, pointing to a prehistoric time when our early ancestors engaged in cave art. This pivotal period in our evolution was when we demonstrated some of the earliest traits of modern human cognition. Evolutionary biologists have consistently linked the development of early art forms to symbolic thinking and the capability to develop language. With language comes the ability to exchange ideas, create stories, and invent technology. [41]

These origins of symbolic thinking can be traced to more than 70,000 years ago, a period when Homo sapiens started doing some incredible things, such as leaving our African ancestral home to create dwellings in Europe and East Asia. In the process, we drove other species of early man to extinction. Only as recently as 30,000 years ago did we start to see evidence of harpoons, clothing, and lamps.

The beginnings of behavioral modernity fundamentally transformed the role of the human species on planet earth, and these unprecedented achievements were due to a revolution in Homo sapiens’ cognitive abilities. In Yuval Noah Harari’s Sapiens he goes on to highlight that “The appearance of new ways of thinking and communicating, between 70,000 and 30,000 years ago, constitutes the Cognitive Revolution.” [12] .

Just like our early Homo sapiens ancestors prior to some 70,000 years ago, automations using RPA are effectively “dumb” robots, perfectly adapted to the low-complexity work they have been required to do, but automation is developing a rapidly growing set of new cognitive capabilities. Intelligent automation, sometimes referred to as cognitive automation or hyperautomation, is a nascent capability emerging through the convergence of the relatively primitive RPA automation tools with a number of key technologies defining the current digital age, such as artificial intelligence and smart analytics.

Intelligent automation is the combined application of a number of solutions that collectively create a cohesive technological fabric, patchworking previously siloed capabilities to create something much greater than the sum of its parts. It provides businesses from across the spectrum with a portfolio of complementary technologies fused with cognitive capabilities that can help advance organizations up the digital mobility ladder, providing them with a set of tools to reimagine and implement the future of work now.

Intelligent automation can widen the scope of automation across your enterprise by going beyond the simple rules-based logic of old to infer complex reasoning and support ongoing decision-making. It goes beyond the traditional lexicon of IT by transforming data into understanding, and understanding into action. A variety of tools in the intelligent automation arsenal can provide your enterprise with opportunities to create unique configurations of technological capability that can impact the speed, complexity, and innovativeness of your business transformation. This will provide the opportunity for your organization to deliver beyond incremental operational efficiencies to create new and enhanced products and service revenue streams. Intelligent automation upskills your technology, your processes, and your people. This is our second cognitive revolution.

Extending our earlier example, a digital worker for a telecommunications provider could be used to support call center advisers in changing the customer’s address details across a number of IT systems, such as the CRM database and billing system. With intelligent automation, a chatbot on the company’s website could be offered as an alternative self-service touchpoint to customers to reduce the waiting times typically experienced by contact centers.

The customer could request an address change via the online service chatbot, which, following a customer security verification process, could then initiate the same digital worker to update the internal systems with the changes. Given the new address details provided by the customer, the automation is then also able to confirm that there is, in fact, a faster broadband service available at the new address.

Following a check of the customer’s existing credit history and available deals, it is able to offer the customer a faster broadband option and a new TV package for a 6-month period via the chatbot. On acceptance, the digital worker progresses through a sequence of steps to order new products and services, such as approving the dispatch of the latest broadband router to the customer’s premises while updating the customer via text message at key moments during the provisioning of the new service.

By utilizing a mix of tools from the intelligent automation technology stack, a differentiated service is provided without changing any of the organization’s key IT systems, such as billing, CRM, and provisioning. This considerably de-risks the digitalization of this process by negating the need for complex integrations and time-intensive and costly bespoke development. Each layer of automation intelligence can be added iteratively, further reducing high-risk “big bang” approaches in favor of rapid market experimentation of new services. These can be tested for customer satisfaction with specific target groups and finely tuned to maximize value.

With the proliferation of data, organizations have the unique opportunity to glean valuable insights on how their products and services are being used by consumer groups, identifying valuable usage behaviors or highlighting the features they value the most. Intelligent automation can provide the ability to not only support analytics in identifying these hidden data patterns but significantly improve the time taken to launch innovations to the market. In his series of Reith Lectures recorded for the BBC in 1964 titled The Age of Automation , Sir Leon Bagrit, one of the leading British industrialists of his time and an automation pioneer, said, “Our main problem in the successful application of automation is one of imagination—and, I suppose I might say, of courage as well.”. [13]

Any claim that a specific technology can support an organization in achieving its digital transformation goals is bold and understandably to be met with skepticism. Intelligent automation brings only a potentially useful set of tools. Just as buying a treadmill to make you fitter requires you to schedule a regime of increasingly challenging combinations of distance, speed, and inclination, the application of intelligent automation requires considered planning and alignment to your existing organizational capabilities. This needs to be merged with talent, organizational change, and leadership to deliver results.

Intelligent automation provides the technological capability to go beyond the cost reduction-only initiatives of RPA and start the transition across the automation benefits trajectory as shown in figure 3.2 . It provides businesses across the digital mobility spectrum with a rich set of tools that can help drive value-based change, delivering demonstrable incremental value along the journey with little need for blind faith.

Figure 3.2 shows that as the organization continues to invest in automation and intelligent automation capabilities, the enterprise has the potential to incrementally transition to new waves of value. Movement up the benefits trajectory takes the organization from using basic RPA automation to reduce costs, such as reducing employees’ time or third-party spend, to improved process efficiencies, for example by increasing the capacity to process higher volumes of work. Benefits associated with higher quality allow the organization to significantly improve the quality and consistency of the products and services they deliver; for example, by reducing error rates through automation.

To say that data is important to modern organizations is an enormous understatement. Various metaphors have tried to articulate its level of importance; from data being the “currency” of digital, the oxygen with the power to both suffocate and drive significant change [31] to being “the new oil,” data is the bedrock of our new economies. Intelligent automation is critically dependent on the availability of scaled, interpretable data. Focusing on hoarding voluminous amounts of data over the last decade in the hope that it would, someday, lead to a “eureka product” has created enterprise scale “data swamp” repositories with little structure and management, significantly impeding the prospect of valuable insights, and fragmented silos of data locked across a plethora of different IT systems.

As well as accessibility, the value and usefulness of this data is in large part determined by whether or not it is structured (think of finding a book in a collection that is ordered alphabetically and by genre) or unstructured (books that are in no particular order). In reading an email on a computer, the information is already digital. There is no digitization process to follow. The email’s “To” and “From” fields are metadata (useful information about the information), whereas the subject line contains content. Metadata is structured data; it’s stored in a highly prescriptive format and, as such, can be easily searched in a database system.

Unfortunately, however, even with structured data, most organizations suffer from “data blindness” using less than half of it, [42] with less than 1% of the structured data ever actually being analyzed—an array of missed opportunities for insight.

Like a river polluting a lake, the challenge is twofold, requiring a strategy that warrants both cleansing and structuring the river of data as it arrives at the enterprise and applying the same to the data lakes, swamps and ponds that exist internally. Here intelligent automation provides the capability to support data engineering initiatives that can be applied to cleanse and structure data at source and from opaque enterprise content.

Many organizations and institutions have provided varying definitions of artificial intelligence. For Amazon, AI has been a key bedrock of their business model and they define it as “the field of computer science dedicated to solving cognitive problems commonly associated with human intelligence, such as learning, problem-solving, and pattern recognition.” [43]

The advance in learning models has been the most critical factor in recent AI breakthroughs. It was based on the realization, during the 1980s, that given the complexities inherent in trying to replicate human cognitive activities such as speech, traditional computational approaches of attempting to program all of the possible sequences and permutations required would never capture the subtle nuances of human systems such as language. Machine learning emerged as an alternative to explicit programming. By providing the AI system with large data sets and a specific goal, the system itself could “work out” the underlying rules or algorithm that could, in turn, better predict the outcome of that goal.

For example, attempting to program a computer system to recognize an image of a dog would be complicated given the many types of dogs, the many different animals that share similar features, and the differing angles a photo could be taken from. A machine-learning approach would provide a large data set containing some images labeled as “dogs” and some images labeled as “not dogs”. The AI system would utilize machine learning to review the data and identify its own objective patterns of what a “dog” typically looks like, in the process creating its own set of unique rules, or algorithm.

This algorithm could then be continually updated and refined by being given more data sets (or more pictures of animals that have dog-like features, such as a wolf). In turn, this would provide the AI system with an increasing level of statistical probability of accurately identifying a dog from a picture. Machine learning can be unsupervised, as per the example above where no human or expert input is required following the provision of data sets, or supervised.

In supervised machine learning, a human supports the system to identify the correct outcome. For example, the AI system may identify a Chihuahua as having a 50% probability of being a dog, a threshold that initiates a clarification request and response from the human. The human can notify the AI system that it is, in fact, a dog (as opposed to a frog in a fur coat), and the algorithm is subsequently updated by the AI system, so future Chihuahuas or equivalent borderline cases can be more accurately predicted.

The salient point is that the output of machine learning is prediction. These algorithms or prediction machines may predict the likelihood of the animal being a dog or the impact of the failure of a driver’s side brake pad, and therefore what actions the car needs to take. Prediction is about taking data you do have to generate data you do not have, in turn informing the potential for making better decisions. AI systems can feed inputs to automations to take the required actions while simultaneously feeding the AI system with more data. This further refines the outputs, creating an intelligent, automated closed-loop system.

These automated IT systems communicate through data flows and supporting software programs to interconnect many different systems together in a closed loop. As human beings, we communicate with each other using language, both verbal and written. Although we take it for granted, this system is complex, requiring a reciprocal understanding of syntax, semantics, and pragmatics that comes with a lifetime of training. Syntax is the study of sentence structure and the rules of grammar. A sentence like “The had a home cat” makes no sense, even though the same words can be re-arranged through the rules of syntax to “The cat had a home.” Semantics is the meaning of sentences. For example, there is a significant difference between ”I was reading a book” and “The book was reading me.” Finally, pragmatics takes semantics further as the study of the meaning of sentences within a context. For example, saying “let’s wrap this up” may have very different connotations if said during a meeting or on Christmas Eve. [44]

Natively, computers communicate using highly structured data such as numbers on a spreadsheet. Computer code reveals the inner binary nature of machines—in effect, their language. If humans wish to seamlessly communicate with machines, then either we all need to learn how to program machines and interpret their outputs or we need to teach machines the subtleties and complexities of communicating with humans using our rich languages.

Applying the principles of computer linguistics to powerful new AI approaches to learning, NLP has created a range of sophisticated technological solutions over the last several years. The rapid proliferation of NLP across our mobile devices and home systems, such as Apple’s Siri and Amazon’s Alexa, has started the process of lowering previous barriers to interacting with technology using human language. While some of the systems may still be notably clunky, the investment in this technology, coupled with its potential to seamlessly communicate with humans, will see this branch of AI continue to mature and develop new applications.

When we talk to each other, subtle factors such as the variability of tone and volume on certain words, or those awkward periods of silence, are sometimes more telling than the actual words spoken. This provides a window into the subconscious, of what we actually mean. Our brains are naturally wired to detect clues hidden in conversations, such as signals of age, engagement, and trustworthiness. The true latent power of speech recognition is the ability to add this additional structured data to the text, providing greater richness and insight on intent and leveraging machine learning to continually improve the predictive power.

Providing the digital worker with the ability to listen can unlock a wide range of use cases across the enterprise environment. Front-office operations may be able to leverage speech recognition integrated with automation to perform live customer analysis such as caller categorization, NPS (Net Promoter Scoring), and complaints analysis. By listening to a customer conversation, a digital worker could augment the call advisor in real time with predictive information feeds, such as identified product information, and upsell or cross-sell opportunities.

The digital worker could automatically notify a senior supervisor if it identified a high-probability risk to the customer relationship that required more skilled human intervention. Banking and financial services could integrate speech recognition and automation on incoming calls, with the ability for the digital worker to provide real-time fraud risk scores based on active conversations and, if required, to immediately block accounts. The machine learning algorithm would be able to continually refine its rulesets to identify the best markers of fraud based on correlations of actual cases.

Where NLP can be considered the reader, with the machine ingesting what has been said or written and assigning meaning and a response, NLG is the writer. It provides the inverse process, allowing the machine to turn structured data into written text. Despite how information is often shared in organizations, our natural form of understanding is not through highly structured data such as spreadsheets or graphs, but through stories and the rich supporting narratives they provide. This is how we have digested information and shared knowledge since the first cognitive revolution some 70,000 years ago.

NLG, or Natural Language Generation, is the process of taking highly structured data inputs and transforming them into meaningful written language, the inverse of NLP. While the emergence of the data-driven enterprise has undoubtedly provided many benefits such as deep customer insights, data-driven performance, and decision-making, the sheer volume of data across the enterprise has led to data blindness. The true value of the data, the information which is critical to decision-making, can become latent. Even worse, as demonstrated by many political parties, it can be interpreted as what you “want” it to say.

Given the significant volumes of hours spent by most organizations in annotating reports, a digital worker skilled in NLG could access required systems, such as MI (Management Information) or analytics systems, daily or weekly. It could run relevant queries to create the data component of the report, such as graphs or tables, generate the body of text, and issue the reports by email to a list of users. Exception reports could be triggered by defined thresholds such as low inventory levels, invoking the digital worker to prepare and disseminate the supporting information in advance of a management meeting.

This approach supports standardization of the reporting, which in turn supports structure familiarity, which in turn leads to better human interpretation and essentially better decision-making. The digital worker can annotate within supervisor-taught constraints or, leveraging the true capabilities of AI, extrapolate this dataset with historical data, and curate a narrative that predicts outcomes. The digital worker provides an unbiased view across all reporting in the enterprise, supporting true data-driven management and decision-making.

A chatbot is essentially a computer program designed to simulate real-time human conversation, typically supported through a web interface such as a company’s website. Chatbots have evolved over the last few years, predominantly from growth in the digital channel. Supporting front-office operations such as contact centers, chatbots provide an opportunity for customers to raise queries or orders, negating the need for long call wait times and the irritating accompaniment of on-hold music.

In contrast, modern AI-driven chatbots leverage NLP to understand incoming written content and NLG to generate a response. This has significantly improved the conversational capability of these systems, providing users with the ability to interact with them on a more natural linguistic basis. Machine learning supports both immediate conversational dialogue and also historical memory. Algorithms can study previous interactions and find patterns in questions to determine whether previous responses resulted in positive or negative outcomes. This provides a mechanism to continually refine the conversational accuracy and, ultimately, provide a rapid resolution to a customer’s request.

Alongside these technological advances, the social rise of messaging solutions such as SMS and chat services on platforms such as Facebook have created a generation for whom online text-based communication is the preferred medium. The growth of chatbots allows organizations to create customer experiences more similar to chatting to a friend or colleague online, rather than an online web-based FAQ. For the automated enterprise, this communication method provides the opportunity to improve front-office customer adviser services and deal with unexpected volume spikes, allowing customer calls to be balanced with asynchronous chatbot channels.

Organizations have also started to see the advantages of deploying chatbots internally, allowing specific functions such as HR to deal with common queries, such as when the next payment cycle is or how many days’ holiday entitlement is remaining for an employee. By integrating software automation with chatbots, your organization can create a digital worker with the ability to communicate, understand, and importantly action requests from internal or external customers across the enterprise. In doing so, rich end-to-end self-service capabilities can be created which can be fully automated and delivered at significantly lower cost to the business.

Intelligently automated chatbots can also be extended to serve suppliers with information. Based on internal analysis within its purchasing department, a global manufacturer of electrical components identified that just over 60% of requests received from its suppliers related to just a few types of queries, such as information on invoice payments and an update on existing stock levels across the group’s three warehouses. A chatbot solution was integrated with its automation software following an initial trial with a single supplier. Since suppliers already had access to the corporate intranet, an online purchasing chatbot could take requests for information, such as a copy of the latest purchase order. The chatbot would translate this request into a set of structured data commands for the automation solution, including information such as the supplier’s name and invoice details.

The digital worker would then access the corporate ERP system, locate a copy of the latest purchase order, and both append it to the live chat session and email it to the nominated supplier’s email account. To de-risk the delivery project, new services were gradually applied to the purchasing chatbot automation over a few months, alongside a supportive change and communication plan to ensure good adoption across the supplier network. The service provided suppliers with accurate and timely information 24-7, while the company benefited from a significant reduction in the volume of requests that required manual responses. As a result, the business was able to release some subcontracted resource costs.

For many sectors, such as financial services and banking, the COVID-19 pandemic of 2020 created a surge of queries. In Ireland, the BPFI (Banking & Payments Federation Ireland) reported an average increase in calls of 400% [45] as customers anxious about the threat of being furloughed, or losing their jobs, contacted their mortgage lenders to request mortgage holidays. Like most organizations during this period, the companies themselves were faced with significantly reduced headcount in contact centers, with many employees having to transition immediately to working from home.

In response, a European bank was able to quickly enhance its existing chatbot service to respond to the increased demand for such queries. Using the company’s existing website, a chatbot was able to hold conversations with customers to resolve queries. In the case of a request for a mortgage holiday, the front-end chatbot, supported by the backend automation, validated a customer through a series of security questions and reviewed the payment holiday request based on a customer’s existing credit status. Following a set of defined business rules, the automation could immediately approve, decline, or request a follow-up with the customer, with each of the responses confirmed with a text message. Although at the time the organization internally had emerging chatbot capabilities, given the immediate crisis and the need for creative solutions, the senior leadership team provided the required support and sponsorship. It quickly mobilized a joint team working with an automation solutions partner to create a powerful and relevant case for intelligent automation. It took less than two weeks from conception to go live, significantly accelerating the bank’s ability to service customers compared to its competitors.

A solution like this would typically take several months to develop given the traditional method of requiring integrations to be built between a number of different IT systems. Integrating the chatbot with process automation capabilities provided a mechanism to significantly reduce build times, delivery cost, and associated risks.As these solutions accumulate greater amounts of historical data, these intelligent automations can become even more powerful by utilizing machine learning capabilities to identify patterns of customer-specific behavior. For example, an intelligent automated chatbot in financial services could recognize that Mrs. Smith typically checks her credit card balance and her last few transactions once every few months. As soon as Mrs. Smith has logged in, the chatbot could ask whether she wanted to check her balance and accordingly provide the last few transactions.

Larger analysis on data sets could suggest that customers who check their balances more regularly during the month of November may want to increase their credit limits. The intelligent chatbot, supported by automation, could check whether or not the customer would be entitled to a credit limit increase and if so, by how much, without any further approval. It could proactively suggest this as part of the next conversation dialogue with the customer. Intelligent automation provides endless opportunities to differentiate services, create compelling experiences, and transition customer relationships from being transactional and reactive to predictive and value-generating.

Sentiment analysis is the field of study that analyzes opinions, sentiments, evaluations, attitudes, and emotions in written language. [46] Organizations traditionally sought customer feedback on their products or services using questionnaires, interviews, and email surveys. Today, the growth of digital commerce and social media has created an endless stream of online chatter, with customers posting product reviews or using platforms such as Facebook and Twitter to vocalize customer delight or frustration. Having previously dismissed this chatter as noise, many established brands have had to quickly learn how to adapt to these new modes of engagement, sometimes following public relations disasters. Sentiment analysis is a subset of NLP that uses linguistics and psychology to understand public opinion, conduct nuanced market research, understand customer experiences, or monitor brand value. On a simplistic level, the NLP algorithm can deconstruct a sentence into constituent parts and score them on whether the sentiment is positive or negative and how strong it is, using negators or intensifiers respectively. By using a machine learning approach, the algorithm continually refines itself using assigned probability scores for negators or intensifiers of text. By integrating sentiment analysis with automation, the digital worker is enabled with emotional intelligence; with the ability to digest large conversational volumes and immediately provide quantitative feedback. For a new product launch, the digital worker could immediately search internal and social feeds to understand feedback and reviews from customers, upload sentiment scores into a centralized analytics system, and report daily on summary scores or support in sharing positive and negative sentiments on different product features. Sentiment analysis can be integrated into a chatbot digital worker to continually sense levels of customer satisfaction as part of the ongoing conversation.

In areas such as contracts, the Harvard Business Review estimated that inefficient contracting causes firms to lose between 5% to 40% of value on a given deal. [47] The use of document analysis on areas such as contract management provides the ability to significantly reduce commercial risk and raise levels of standardization. For example, by training on a number of different contract types, NLP algorithms can identify patterns of compensation or confidentiality clauses that do not follow existing standards. In turn, they can flag these specific items as high risk and initiate a digital worker to action specific risk mitigation activities, such as requesting changes to the contract by the legal department.

Document analysis has the capability to use technology to review large numbers of contracts. Since organizations typically hold contracts for several years, these could provide the initial basis for a large training data set for supervised or unsupervised machine learning models to be generated and applied. This, in turn, could be used to generate new contract templates that incorporate the best of previous learning to significantly improve the level of contracted risk exposure. Structured data extracted from third-party supplier contracts, such as monetary amounts and legal entity names, can be ingested into centralized data stores to support downstream analysis.

When defining new contracts, intelligent document analysis with automation can allow an optimized sequence of approvals to be set up, based on key data. That could mean intelligent document analysis flagging up instances where contract values are greater than certain thresholds such as £1m, or cases where complex legal text in the document could expose the company to high levels of financial risk.

Many organizations still conduct a large part of their business using paper documents. Whether commercial contracts, invoices, or purchase orders, the “paperless office” for many organizations is still very much a future state. However, the growing recognition that the data within their organizations is one of their key assets and digitization at the point of entry is vital has led to an upsurge in OCR technologies. OCR (Optical Character Recognition) is essentially the process of digitizing physical text, so that unstructured text in paper format can be converted into electronic data, a critical precursor to digitalization. The technology dates as far back as 1914, before the First World War, when physicist Emanuel Goldberg invented a machine that could read characters and convert them into telegraph code. [48] Modern OCR technologies have greatly improved the ability to recognize large volumes of text and increasingly complex documents due to the application of artificial intelligence.

Intelligent OCR tools use a combination of machine learning, computer vision, and pattern recognition to increase the probability of correct matching, removing the limitation of previous OCR tools that used the document’s actual layout to support in structuring the data—for example, the OCR solution assuming that for all invoices, the bottom right-hand corner contained the invoice total. Large multinationals typically deal with hundreds of different invoice formats, with additional variabilities such as different languages. Intelligent OCR uses the entire document to support a greater probability of matching and structuring. It recognizes that there is another area of the document that contains a numeric total, and because this is the largest of a series of other numbers, it has a higher probability of being the invoice total. Used in a supervised training mode, these new OCR tools work with humans to initially provide recommendations. When the recommendations are accepted by the human, this reinforces further matching.

As the machine learning algorithm continually learns from larger data sets and is reinforced by human supervision, the probability of matching increases significantly, leading to greater efficiencies. Intelligent OCR is rapidly becoming a commodity technology, with many RPA platforms now either providing their own native OCR capabilities or integration options to connect with third-party OCR products. The primary differences between current OCR tools currently are found in specialist areas such as handwriting recognition. OCR with automation can support common financial processes such as invoice processing. Once the OCR engine digitizes the information, the solution can accurately label the data items, such as identifying the supplier name, invoice total, specific items, and prices. An automation can check the invoice for any discrepancies, such as the incorrect price of a part as compared to the company’s internal pricing system. If the document passes all of the required checks and business criteria, the digital worker copies the relevant information into the organization’s accounting or ERP system for onward payment processing, seeking manual approvals based on invoice value. Brexit, the UK’s decision to withdraw from the European Union, will present many UK businesses with significantly increased levels of bureaucracy in order to continue trading with their European customers and suppliers. For example, current legislation suggests that after leaving the EU, UK companies will be required to complete customs declarations for goods imported from or exported to members of the EU community.

The number of customs declarations could rise substantially for some companies. For instance, a large UK manufacturer of construction equipment, which imports a wide variety of parts from EU states, estimates that its volumes will go from 15,000 to 75,000 declarations per year. To deal with the increased workload, it has invested in OCR technology intelligently integrated with automation. The technology not only recognizes declaration forms but automates a large part of the transactional processing and onward digital storage of these documents. Even with volumes expected to increase fivefold, the internal finance tax team has sought to invest in technology as opposed to adding additional headcount.

The automated enterprise will see a fundamental shift in the execution of work, moving more towards collaboration between employees and their digital counterparts. Smart workflows coordinate this complex orchestration between humans and emerging cognitive digital workers, providing the opportunity to drive efficiencies across complex end-to-end processes. Work can be allocated based on the skill levels required. For example, a digital worker could automatically process new bank account requests based on good credit scores while employees are involved in cases that require more investigation, such as an existing customer with a poor credit score requesting a new account. While the information age brought with it unparalleled levels of greater customer choice and variability, internally it created increasingly complex organizations and industries, which can significantly benefit from intelligently automated smart workflows. Constrained by government bureaucracy and policies, healthcare has been one of the biggest laggards when it comes to technology. Yet perversely, as demonstrated by the COVID-19 pandemic, it is the industry that has the most potential to gain from digitalization.

In the UK, if you wish to make an appointment to see your medical general practitioner (GP)—a key gateway service to accessing the wider NHS (National Health Service)—many local GP practices require you to call between certain hours in the morning. If all appointments for the day have been booked, you need to call back the next morning. If you manage to get an appointment and the doctor recommends blood tests, a separate appointment with a nurse will be required.

If you require an update on the outcome of the tests, you’ll be advised to call back between two and four days later. And if the blood tests show something unexplained, a repeat of the earlier process is required to secure a second appointment with your GP. Let’s say that following this appointment, you are advised to see a lung specialist at the local hospital. On arrival at the hospital, and following an initial consultation with the pulmonologist, you are advised to take a scan. This, in turn, means you’ll be referred to the hospital’s radiography department. It is a long and winding road, and it is abhorrent that in today’s developed economies, processes like the above still exist—even more so that they exist in such critical sectors.

Technologies like smart workflow integrated with automation provide the capability to coordinate complexity between systems, processes, and people. Within healthcare, these digital tools can orchestrate flows of data, augmenting key human touchpoints such as skilled medical practitioners with technology in order to improve utility, and raise the level of patient care. That could involve scheduling initial appointments online, triaging patients at higher risk using AI analysis based on clinical and patient data, automating services with blood test centers, and even coordinating parking at the hospital.

With the growing proliferation of mobile devices, AI models trained on data sets, and personal IoT devices such as health monitors providing real-time data feeds, technology provides a significant opportunity to transform the healthcare sector. Given the variability of clinical pathways, data, and IT systems, together with concerns about patient confidentiality and data security, innovation in the sector will be driven by ecosystems of new and existing players. They will include incumbents, government authorities, and technology partnerships.

The business process landscape across many enterprises today is complex and changing, unfortunately leading many organizations to have inaccurate supporting process maps. Successful automation is dependent on an accurate, detailed understanding of the current business process. Slight variations, such as differences in the way employees or teams execute a specific business process, can have a significant impact on the estimated benefits of the automation. Understanding the process in detail within short timescales can be achieved with the use of process mining tools. Typical process mining solutions use two different mechanisms: server-side or desktop. Server-side solutions collect log files on enterprise systems such as SAP (a market-leading Enterprise Resource Planning software solution) to trace the journey of the user, as provided at the server level. This provides a view of how they have been using the system. Desktop process-mining tools use software installed on the desktop to capture the user’s activity at the local level, before uploading this data to a central solution.

Once the existing business process has been analyzed, the software may support the ability to simulate changes and their impact on KPIs such as efficiency, productivity, and the time taken to complete the process. Some process mining tools also support the automatic generation of process documentation, reducing the time spent during the early phases of an automation project. Such tools are invaluable as de facto sources of truth when there are differences between the activities undertaken by employees and what’s expected, as defined in the process documentation by the business.

Analytics has been one of the key technologies at the heart of digital. Adopted by organizations since the emergence of the information age, analytics has fundamentally changed the way businesses manage and operate. The recent emergence and growth of big data, in the form of unstructured and structured data such as pictures, tweets, and online purchase history, has helped to usher in the new era of advanced analytics. Advanced analytics builds on the foundations of understanding data, and is subdivided into three main categories: descriptive, predictive, and prescriptive analytics. Descriptive analytics uses data to provide information on what has already happened in the organization, such as total sales last quarter, cost to date on inventory spend, and so on. Insight is generated by aggregating historical data, as this has typically been a good predictor of future performance.

One of the key developments in descriptive analytics capabilities over the last decade has been the shift of information access from database experts to business users. This has created a growing percentage of employees across the enterprise who are information savvy. Given a small amount of training, these employees are comfortable accessing corporate data sources for information using “drag and drop” interfaces, removing the need to understand complex database languages such as SQL (Structured Query Language).

Predictive analytics has been at the forefront of recent developments, thanks to the same key factors that have contributed to the growth of AI: big data, processing power, and machine learning capabilities. Predictive analytics uses many different techniques such as data mining, data modeling, and machine learning to analyze current data to make predictions. This allows organizations to transition from being reactive to proactive, supporting the ability to predict impacts to the business, from changes to consumer buying behaviors to the threat of cybersecurity breaches.

Prescriptive analytics remains an emerging discipline and is the least mature of these subsets. As a natural extension of predictive analytics, it aims to suggest a range of decision-making options and determine the potential outcomes of each of those decisions. For example, in the energy sector, oil companies are using prescriptive analytics to determine the key factors affecting the price of the commodity in order to implement the optimal hedging strategy.