As the writer Elbert Hubbard once said, “One machine can do the work of 50 ordinary men. No machine can do the work of one extraordinary man.”1 Your company may be filled with dozens, hundreds, or even thousands of professionals with extraordinary potential, but they are often spending their time doing mundane, time-consuming tasks when they should be applying their potential to higher-level, human cognition challenges. Exponential technology makes that leap possible, creating the opportunity to transfer the “machine work” to machines and augment and expand the role of the professional.
As the future of work unfolds, demands on professionals to increase efficiency and add value are mounting. In turn some professionals are seizing the opportunity to fundamentally change how they do their work. These Exponential ProfessionalsTM (ExPros) of the future are being powered by a range of exponential technologies.
Programs can be developed to do tasks of varying levels of difficulty, freeing workers from time-consuming, repetitive tasks so they can focus on deeper, more value-added work. For a technology to be considered “exponential,” it must be developed so the speed significantly increases each year, and/or the cost is significantly reduced. A wide range of exponential technologies, illustrated in the graphic below, are rapidly advancing, with disruptive impact on industries and all aspects of our lives. Let’s explore some of the major exponential technologies and their potential impacts.
What exponential technologies can do.
In addition to freeing up capacity for people to work on higher-value tasks, exponential technologies automate and drastically reduce errors. Some of the businesses where we’ve implemented these technologies have reduced errors to as little as 0 percent, and lowered costs to fractions of the previous cost. Consider one type of ExPro—actuaries. If the actuaries responsible for pricing insurance plans are busy cleansing data and running models for most of their day, they can’t spend their time revolutionizing their pricing model’s inner workings, optimizing the value of benefits being provided, and helping their clients receive the care they need. But with the increased capability provided by exponential technologies, actuaries can go beyond the work of today and offer enhanced insight into how the products they design can help people live longer, healthier and more financially secure lives.
These technology advancements are particularly important in the wake of new regulations, such as IFRS 172 and FASBTI,3 that are putting unprecedented pressure on the actuarial profession. Given that the technology exists and is proven to be able to drastically improve quality and reduce time allocated to meeting regulatory requirements like these, why not take advantage of it?
Exponential technologies can also augment professionals’ capabilities to create new kinds of value. In the health care field, for example, exponential technologies can unlock insights that physicians may not be able to identify with traditional methods. A Nature Research Journal study4 demonstrates the positive impact of incorporating AI into radiology workflow, specifically inter-cranial hemorrhaging (ICH). Researchers collected 46,583 head CTs (~2 million images). They trained a deep convolutional neural network (CNN) on 37,074 studies and evaluated 9,499 unseen studies. They implemented these models to re-prioritize “routine” head CT studies as “stat” on real-time radiology worklists if an ICH was detected. The algorithm reduced the average time in which a radiologist diagnosed ICH patients from around 8.5 hours to just 19 min (a 96 percent improvement), demonstrating the positive impact of incorporating AI into radiology workflow. This result is particularly significant because research shows that nearly half of all ICH mortality happens within the first 24 hours.
In an example from the world of HR, Deloitte designed and deployed 10 AI assistants within a client’s HR department to support the work across several processes: workforce forecasting, performance appraisal administration, promotion, and posting administration and officer promotion. For another client, Deloitte developed a cognitive virtual agent which aimed to resolve inconsistent HR services that lacked tracking and reporting capabilities. The goal was to centralize services while keeping them high touch and personalized. The virtual agent handled five different pilot conversations that immediately led to enhanced customer experience as well as cost savings due to the agent handling 50 percent of inquiry volume.
Understanding different types of exponential technologies and how they can add value
Exponential technologies take several forms.
Data Wrangling is the process of mapping and compiling raw data into a more succinct and usable format for downstream use in reporting and analytics. Deloitte implemented a data-wrangling solution for a contract model validation process to automate data processing (consolidation, formatting, transformation, duplicate resolution) and validation (reconciliation, consistency and reasonability checks) around the model input. The solution resulted in better-quality data feeding the models, reduced time spent, and was reusable for other applications.
NLG—Natural Language Generation ingests structured data (i.e., spreadsheets, databases) and creates structured text designed for humans to read. An NLG tool can be used when producing an audit summary memo. After professionals complete their testing and validation workbooks, the results are fed directly into the NLG tool, which can produce a summary memo. The team would then review the memo and make updates where necessary. In the medical world, physicians can use NLG to generate detailed, accurate medical notes through spoken or written sentences.
NLP—Natural Language Processing tools ingest structured and unstructured data such as emails, tweets, PDFs, etc., and output data designed for machines to use (Internet of Things). For example, NLP tools can be used to extract information from millions of pages of contracts with varying formats and terms and convert it into a neatly formatted, concise spreadsheet that can be easily analyzed and fed into the models. This process replaces the manual input of data and can save a significant amount of time.
Machine Learning uses statistical techniques to give computer systems the ability to progressively improve performance on a specific task with data, without being explicitly programmed. In an insurance example, a cognitive reserving solution developed with machine-learning algorithms and Tableau data visualization software facilitates review of claims data and easily finds the underlying drivers of variances in claims experience.
CNNs—Convolutional neural networks and DNNs— deconvolutional neural networks are two similar types of artificial intelligence that are designed to mimic the way that a brain would operate—by running small simple tasks that with each iteration allow the neural network to gather new information and “learn.” How can this be used? Improving medical testing is one area. There are more than 2 billion chest X-rays performed worldwide per year.5 In one study,6 the accuracy of one algorithm, based on a 121-layer CNN, in detecting pneumonia in over 112,000 labeled frontal chest X-ray images was compared with that of four radiologists, and the conclusion was that the algorithm outperformed the radiologists.
NLG—Natural Language Generation ingests structured data (i.e., spreadsheets, databases) and creates structured text designed for humans to read. For example, an NLG tool can be used when producing an audit summary memo. After professionals complete their testing and validation workbooks, the results are fed directly into the NLG tool, which produces a summary memo. The team then reviews the memo and makes updates where necessary. In the medical world, physicians can use NLG to generate detailed, accurate medical notes through spoken or written sentences.
Implications of exponential technologies
Exponential technologies present exciting opportunities to expand the role of professionals while increasing accuracy and efficiency. This transformation is happening all around us— many S&P 500 companies are rapidly automating their processes to streamline their business for simplicity, achieve digital transformation, increase service quality, improve service delivery, and contain costs. A rapid response is critical. Greater reliance on exponential technologies demands new types of talent, new training programs, and rethinking or reapplying professional standards. There is both a need to trust machines and what they can do and a need to avoid over-reliance on machine output to the extent that human wisdom, experience, and intuition are ignored.
By letting machines and people do what each does best, we make room for superjobs and other forms of higher-level work that tend to be more rewarding for the people who do them and for the organizations that enable themand the customers they serve.
Darryl Wagner is a principal with Deloitte Consulting LLP, where he leads the global Actuarial & Insurance Services practice as well as our Exponential Professional™ offering helping workforces and organizations achieve their full potential by leveraging emerging technology and workforce options.
Sourabh Garg is an Actuarial and Insurance Solutions senior consultant within the Human Capital practice of Deloitte Consulting LLP.
Callum Humphrey is an Actuarial and Insurance Solutions business analyst within the Human Capital practice of Deloitte Consulting LLP.
3FASB, Accounting Standards Update No. 2018-12, Financial Services—Insurance (Topic 944): Targeted Improvements to the Accounting for Long-Duration Contracts, August 2018, https://www.fasb.org/jsp/FASB/Document_C/DocumentPage?cid=1176171066930&acceptedDisclaimer=true.
4Titano, J. J. et al. Automated deep-neural-network surveillance of cranial images for acute neurologic events. Nat. Med. 24, 1337–1341 (2018).
5Wang, X. et al. ChestX-ray8: hospital-scale chest X-ray database and benchmarks on weakly-supervised classification and localization of common thorax diseases. Preprint at https://arxiv.org/abs/1705.02315 (2017).
6Lindsey, R., et al. Deep neural network improves fracture detection by clinicians. Proc. Natl. Acad. Sci. USA 115, 11591–11596 (2018).