The individual elements of RPA software are not new. However, it’s the combination of all the features into a single package that works with existing systems which, in many cases, creates a compelling alternative to core-platform integration or replacement. And not only can RPA reduce manual operations costs by 25–50% or more, it does this while improving service and compliance, typically providing an ROI in less than a year.
What about Cognitive Robotics/AI?
There has also been much focus on the potential of Cognitive Robotics/AI, with leading companies developing, for instance, driverless cars and self-navigating drones. While the progress being made in these projects is very impressive, the costs are significant and they pose some interesting challenges.
The equivalents in the financial services are self-optimizing customer service, loan pricing, financial advice, and claims/complaints handling. Designing a good statistical or machine-learning optimization approach is challenging enough, but designing and monitoring one that aligns to legal, regulatory and ethical conduct requirements is exponentially more so.
Moreover, the AI hype often overrides a realistic discussion on human-level-AI progress and expectations. For example, AI pioneer and author Douglas Hofstadter 4 argues that famous cutting-edge vocal applications and translation/game-playing programs do not contain any real AI.
Nonetheless, there are clearly areas where a degree of learning or ‘cognitive’ technology offers a significant advantage, such as processing of paper documentation, understanding speech and detection of fraud. In these areas, there are three standard approaches:
Cognitive Robotics could carry substantially higher costs than standard RPA and so perhaps should be reserved for the highest value processes. However, as a more general solution, it could also form part of a future wave of automation, when financial services organizations are more mature in the deployment of advanced analytics techniques and associated model risk management, and when the technologies are more mature and cost less.
- Adoption of a niche product: Common for highly specialized situations like voice processing and natural language interpretation, or for analysis of legal contracts.
- Adoption of a targeted solution: For example, generic document scanning and intelligent optical character recognition (OCR) solutions for processing paper documents.
- General cognitive robotic platform: Combining an analytics or machine learning platform with RPA.
Digital and robotics: combination benefits
The gains from automation are considerable, but much more is possible when robotics and digital are combined. RPA works with content that is available within a system, so can only automate a claims process once the initial information has been dealt with by agents, which might involve numerous conversations and manual input of information.
But if those preliminary stages are delivered via digital channels (maximizing the extent of customer self-service), then robots can start working faster and across an entire end-to-end process.
The ROIs that can be delivered will significantly outstrip those available from robotics alone, by as much as two-and-a-half times. As robotics take on greater responsibility for an end-to-end process and minimize or even eliminate altogether the amount of human intervention required, potential ROI rises sharply.
Connecting digital with robotics addresses some of the largest inefficiencies in current processes. By working with any legacy system and with a digital adapter sitting on top of the robotics, this can in fact digitize whole new areas of business processes.
This is where one may see the next big wave of opportunity. For example, insurers are likely to be able to digitize support for only 25% of their current products and services, but the combination of robotics and digital expands the scope across a far wider range – and therefore the available savings too.
RPA vs. platform upgrade
The core benefits from Software Robotics are the same as for any automation approach: reduced overall cost; improved speed and timeliness; improved accuracy; improved governance and control; and full audit history. In a sense, these benefits are the same as those typically associated with a core-platform upgrade. However, robotics can deliver these benefits much faster and at lower cost than traditional IT integration projects, for three reasons:
- Use of existing user interfaces: No (or very limited) requirement to change existing legacy systems, something which is often expensive and time-consuming.
- Integration testing costs are significantly reduced: There is no requirement to synchronize releases across all platforms. Robotics works with the core platforms as they are at any given point in time, and contains many accelerators for accessing existing systems and desktop resources.
- The visual nature of RPA tools, which build on existing core applications: Allows process automation to be delivered incrementally using an agile approach (typically, a two-week release cycle). This accelerates benefit realization and improves transparency, thus reducing risk, and also allows for automation of processes that evolve over time.
How is robotics deployed, and what is the target-operating model?
While robotics is based on deployment of a software tool, it should not be treated as an IT integration. That approach generally leads to low adoption and reduced benefit. A far more effective approach is to imagine a virtual workforce, or a set of invisible robot hands, working from a task list and following documented processes.
In a sense, this is is comparable to the deployment of desktop tools: IT provides the platform, and business users make use of the software to add value. For robotics, it should be business users (or staff very close to business departments) automating processes.
Within a large organization, the actual operating model will need to be scaled into a centralized, hybrid or distributed robotics capability, but the principle of keeping ownership and control for process automation close to business users and departments is key to successful adoption, and for protection of business agility.
Robots and people
Robots are a highly flexible workforce that can seamlessly move from any defined task to any other to meet business needs. What robots are not intrinsically able to do, however, is exercise subjective judgment, build empathy or support customers’ emotional needs. They are not able to handle situations that are new and different from the processes prescribed to them. In this sense, they are not a replacement for people.
The real benefits come from the combination of people, core platforms and robotics so that:
- Core platforms support core data records and automate highest value processes.
- Robotics run all the repetitive, standardized processes across separated core platforms, and one-off high-volume processes or rapidly evolving processes that are costly to automate within core platforms.
- People focus on adding value through strategy, building deep customer relationships, managing exceptions, driving change and continuous improvement, and low-frequency activities that are not cost-effective to automate.
Robotics for insurance
We now briefly look at the potential benefits of RPA, AI and Drones for the insurance industry 5, which include: reduced costs for operations, possibilities to offer new services, bespoke products for individuals, fraud detection and prevention, and improved risk assessment accuracy. Some of these are still tentative right now and depend crucially on a rapid advancement of AI.
RPA in Insurance
RPA is already a reality for insurance 6. RPA benefits for insurance include the reduction of a claims documents processing team and of costs more generally. Guttridge 7, discusses the kind of benefits that can be obtained via RPA.
For instance, in less than two years, ten automated processes within the insurance business have been introduced, which has led to processing time reduction (one process by over 90%) and uninterrupted operations with multi-skilled robots working on processes 24 hours a day, seven days a week. Another important benefit has been the lack of human errors in processes.
Insurance applications such as bespoke products for individual clients would require an intelligent virtual agent/broker and a high degree of cognitive computing and it is not really clear whether machines will attain human level interaction capability in the next few years.
Optimists say we already have the technology that is needed for this, but the reality is that human level interaction is still quite limited and we will have to wait and see if the technology really attains credible human interaction capabilities.
When this happens, we could indeed have personal bespoke virtual brokers for tailored life and car driving insurance policies, for example, with an enlargement of the insurers’ services to a much broader population and for a much broader range of risks. Another area where AI could be used is on claims validation. While RPA can considerably simplify the operations around claims management, the approval of a claim still requires judgement and evaluations beyond the RPA grasp.
In this sense a sufficiently advanced AI, having access to the claim-related data via drones, sensors or preferred news channels, could pre-validate or pre-approve claims by verifying the claimant information and data, potentially using drones if further investigation is needed. A hybrid approach could also be used: an AI system augmented with human intervention when needed.
AI methods could also use social data to design fraud indicators that could predict to some extent the risk of a fraud from a given entity. Currently, machine-learning algorithms are being used for fraud detection 8. As AI advances, these algorithms could attain higher predictive power and could become crucial in the management of fraud risk.
Drones and sensors
Drones and sensors offer a number of opportunities and challenges. We already considered the use of drones for claim validation above. Similarly, sensors would measure the insured person/property/vehicle risk sensitive parameters, allowing the insurance company to tailor the insurance offer to the specific client risks, and verifying that the risk profile the client has in mind corresponds to the actual risks measured in reality. Sensors could also create a positive feedback effect on clients.
A client who is aware that her car contains a number of sensors will be more careful in driving. However, it is the combination with AI that could make sensors and drones revolutionary. The possibility to automate evaluation and judgement on claims and policies based on drone- and sensors-collected data would extend the sensors-based insurance approach to a much broader base, with potentially lower premiums for clients and reduction in risk for insurers. Given current limitations of AI this is still tentative, but it is definitely an area where insurers are investing relevant resources 9.
We have highlighted the current challenges and opportunities in applications of robotics/AI to financial services and insurance in particular. Combined RPA and software approaches have been already implemented with considerable benefits in cost reduction and efficiency. More advanced AI applications depend on the general advancements of AI, and human level interaction agents are not there as yet. Nonetheless, we can foresee the great potential for these applications in the insurance sector and beyond.
1 See Frey and Osborne, 2013;
2 See Lamberton et al, 2017;
3 See Lamberton, 2017;
4 Reported in Herkewitz, 2014;
5 See also Cranfield and White, 2016;
6 See Lamberton, 2017;
7 See Gutteridge, 2015, reported in Cranfield and White, 2016;
8 See Guha, Manjunath and Palepu, 2015;
9 See EY2016b;
Adam Cranfield and Dan White (2016). The rise of the robo-insurer. Ninety consulting paper
EY (2016a). Get Ready for Robots
EY (2016b). EY 2016 Sensor Data Survey: Disrupt or be disrupted
Carl Benedikt Frey & Michael A. Osborne (2013-09-17). The future of employment: how susceptible are jobs to computerisation. Oxford University, Oxford Martin School
R. Guha, Shreya Manjunath and Kartheek Palepu (2015). Comparative analysis of machine learning techniques for detecting insurance claims fraud. Wipro research paper. Available at https://www.wipro.com/documents/comparative-analysis-of-machine-learning-techniques-for-detecting-insurance-claims-fraud.pdf, Accessed on October 10, 2016
Guttridge, A. (2015). How RPA is taking service providers to the next level. http://www.blueprism.com/4639, retrieved on Dec 1, 2016
Lamberton, Chris, 2017, Get ready for Robotic Process Automation, Financial Services Insights, https://fsinsights.ey.com/big-issues/Digital-and-connectivity/get-ready-for-robotic-process-automation
Lamberton, Chris, Brigo, Damiano, and Hoy, David, 2017. Impact of Robotics, RPA and AI on the insurance industry: challenges and opportunities, The Journal of Financial Perspectives: Insurance
This is an abbreviated version of the Impact of Robotics, RPA and AI on the insurance industry: challenges and opportunities article published in the Journal of Financial Perspectives: Insurance edition.
Download Impact of Robotics, RPA and AI on the insurance industry: challenges and opportunities here