The ultimate guide to Conversational Process Automation (CPA)

How is CPA different from chatbots?

A chatbot is a computer program that simulates human conversation through voice commands or text chats or both. Today, for the insurance industry as for others, chatbots offer an innovative customer experience in substitution or at the side of traditional communication channels (e.g. phone, email, forms). Their interface is similar to most common messaging applications such as Whatsapp, Messenger, with the substantial difference of a robot (bot) to chat with.

CPA mainly focuses on processes requiring integration and advanced business logic, oftentimes customer-facing. The aim is to keep interacting with customers through an intuitive and personal front-end solution whilst allowing the customer to self-serve in a business process. This isn’t possible with a generic chatbot or a webform. Phones and web chats aren’t feasible either because of the time they take. In other words, CPA augments chatbot capabilities to enable a personal conversational experience into high-value processes at scale and execute tasks that a human would normally do.

On top of this, CPA is powered by a platform designed to analyse conversations and data captured by the chatbot in order to generate knowledge and critical insights about the performance of the process and about the customers subject to it. For example, CPA platforms can highlight which type of customers are selecting a particular coverage option, which time of the day customers are engaging the most, which are the most recurring requests, and where customers dropped the conversation.

The platform often includes machine learning (ML) capabilities to enable continuous improvement. It means that CPA is powered by AI but with different technical skills than conversational AI, as the focus is more on the optimisation of the business performance instead of the conversation itself. In other words, “intelligence” is applied to generate insights based on real customer data - both demographics and behavioural- that inform businesses over the changes to be made in the process, or in the chatbot conversation itself.

To deepen, read the blog How to avoid chatbots' failure with Conversational Process Automation

How is CPA different from Conversational AI?

As the name suggests, Conversational AI mainly focuses on the communication between computers and humans and can be defined as the brainpower that makes machines capable of understanding, processing and responding to human language.

In order to achieve its purpose, Conversational AI platforms make use of various technologies (e.g natural language processing - NLP, machine learning - ML, predictive analytics etc.) aimed at offering the best conversational experience, learning from it and improving it over time.

While Conversational AI focuses uniquely on the performance of chatbot conversations, CPA is primarily concerned with the execution and optimisation of the business processes it covers. In other words, CPA brings the actual capabilities to execute the demands of the customers that require actions outside of the chatbot conversation itself, linking them with backend systems and processes.

Besides, as a side effect, CPA helps businesses to be more performant by nimbly adapting to customer needs through the gathering of structured data and continuous improvement based on analytics and insights available through dashboards.
In summary, CPA foresees the use of conversational interfaces to gather and structure human input, therefore it overlaps with Conversational AI. Also, the uniqueness of CPA is to expand chatbot capabilities by linking them with existing processes, fully or partially automating parts of them. Therefore, it also overlaps with business process automation (BPA).

To deepen, read the blog How Conversational Process Automation and Conversational AI differ

How is CPA different from RPA?

RPA is an application of technology aimed to improve business processes through automation. It normally leverages robots that automate the tasks of existing processes. This ultimately reduces process inefficiencies and improves their performance at a fraction of the cost. The business processes targeted here are middle and back-office processes.

Such robots also eliminate unnecessary human manual input in various steps of the business process. Despite this, RPA has a glaring limitation: very often business processes start with customer-facing interactions. This often translates into unstructured data that RPA is not able to handle.

CPA was born more recently to solve this problem: it focuses on the automation of front-end customer-facing processes. This often implies the use of expert chatbots to enable a personal conversational experience at scale.

Objective number one of CPA platforms is to interpret and understand customers’ unstructured input, structure it and wrap it into a format digestible to robots (e.g. RPA itself) or back-end systems directly (e.g. insurance platforms or other automation platforms).

To deepen, read the blog CPA + RPA: Enable self-service for customers

How is CPA linked to other technologies?

The automation ecosystem is wide with various technologies designed to execute particular tasks and achieve specific goals. In the image below, we consider the world of customer-facing interactions between individuals and businesses in relation to automated solutions.

CPA is automating conversations and processes at the same time reconciling the conversational realm to the business process automation one.

To deepen, read the blog Demystifying automation in insurance

Evolution of insurance channels

Insurance is a highly transactional business process between a customer and a company. Whether a customer is about to enquire, buy, manage an insurance policy or make a claim, the type of interaction boils down to two categories:

1. The request of information from customer to insurance company. Usually, this includes questions regarding the policy at all stages of the product. Traditionally, the interaction has been managed through voice, where the information is exchanged through conversation between the two parties.
2. The request of information from insurance company to customer. Usually, this involves the collection of data from the customer such as personal details or details about the insured item. The interaction is more written in this case and takes the shape of a structured form.

Originally, the conversation between a customer and an insurance company was managed through face-to-face interactions. To address the challenges related to the need for physical presence and limited availability during office hours, insurers evolved through the setting up of call centres. Call centres can provide a good interaction but they saturate quickly causing customers to wait. Web live chat appeared in the market in the last few years as a solution to limit saturation at a lower cost. Yet, they suffer from the same problem since the number of simultaneous conversations for an agent is still limited.

While the data varies for every customer, the collection process is very similar. This is why insurance companies set up forms at different stages (e.g. getting a quote or making a claim). With the digital revolution, the paper form has evolved into webforms and web portals. These provide short-term cost-savings opportunities by automating processes that were once performed manually. Yet, they provide a dry, impersonal, and untailored customer experience. Web forms lack not only the business logic required for the execution of insurance processes but also have difficulty in delivering a user-friendly customer experience across devices from laptop to mobile.

CPA appeared in the market to solve the current challenges and limitations within the insurance industry. Leveraging technology to deliver automated, intuitive, and personal conversations, the idea behind CPA is ultimately to offer a pleasant experience at scale. It aims to ensure that every conversation between a business and a customer is personal at a fraction of a call centre cost.

40 years ago insurance leaders decided it was time to push their business to the next level by enhancing operational efficiency. Today besides the constant need for operational improvement, customer expectation is raising more than ever. CPA imposes itself as the natural evolution by offering the chance to take two pigeons with one stone.

Example for Quote & Buy

Within insurance, quote & buy processes are mainly managed through webforms. The main problems related to such digital solution are:

1. Lack of responsiveness. Webforms provide a dry and impersonal customer experience whereby customer details are collected in bulk, independently from the customer type or input, resulting in a clunky one-size-fits-all experience. Often customers are asked questions that are irrelevant in consideration of the input they just entered in the previous fields.
2. Poor multi-device capability. The webform experience is designed for desktop devices, still working for tablets but not for smaller devices such as mobile phones. On top of that, most of the time the design is not responsive and not adapting to the device size causing the mobile customer a very frustrating experience.
3. Low detail of analytics. A traditional webform journey is usually divided into 4 steps, each one corresponding to a different webpage (e.g.details, quote, policy, payment). Few data can be created out of this digital asset. For instance, the conversion rate and the percentage drop of potential customers will be available only between one page and the other.

In contrast, CPA offers a simpler and more personal experience through expert chatbot conversations. The customer enters the insurer’s website and finds the chatbot to interact with. The experience is conversational and natural, though a messaging-like interface, adapting to the customer input and providing clarification of unknown terms including insurance jargon. Depending on the level of automation, payments can be securely performed through the chatbot.

While transactions are executed, expert chatbots fill the CPA platforms with data from each conversation. When the insurer logs in, he will be able to see the result of the quote & buy process with performance details. The platform allows to analyse conversations in depth and highlights improvement opportunities based on customer demographics and behavioural patterns. Through it, it will be possible to see how the chatbot is performing according to demographics, or how individuals are interacting with it, where they are dropping the conversation and suggest reasons why. After insights are generated, AB testing can be easily set up with the ultimate goal of performance optimisation. This process opens the door to further personalisation, with dedicated and more performing journeys according to demographics or device and ultimately, conversion rate optimisation.

See how Spixii partnered with Duck Creek and Blue Prism to offer an end-to-end quote & buy solution.

Example for Policy Administration

Most of the policy administration tasks (roughly 80%) are quite simple and directly manageable by every policyholder. If the insurance company allows self-serve, such tasks can be easily performed by customers, without any input from its operators and without exposing policyholders to unnecessary queues over the phone.

Among the most common tasks, there is the update of policy details, request of policy details or next payment to perform. As an example, let’s say a policyholder recently changed his residence address. With CPA implemented into the insurer’s system, the policyholder can initiate a conversation with a chatbot from the insurer’s website and ask to change the address related to his policy. If the expert chatbot is connected with the policy system, such policy update will be performed without any input from operators. If instead, the request from the policyholder is more complex, it can be routed to a human, who will find details about the conversation along with the task to perform.

See how Spixii partnered with Duck Creek and Blue Prism to offer an end-to-end quote & buy solution.

Example for Claims

There are multiple steps within a traditional claim process including notification of loss, validation of the claim, assessment adjudication and payout. Because the process is mostly managed by claim handlers, the process will require long processing times often increased by lack of efficiency across the pipeline, increasing policyholders anxiety and reducing their opinion over their insurance provider. How can CPA help with improving this process?

An expert chatbot is implemented within the insurance company website and manages conversations with customers 24/7/365. When integrated with the insurer’s policy system (usually through API), the chatbot is able to authenticate the policyholder. Different levels of security can be applied from simple authentication to two-step systems with email verification. The policyholder engages in a slick and simple conversational experience and is able to complete his first notification of loss (FNOL) with all the information related to his claim, including media assets such as photos, videos or documents.

Direct feedback from the policyholder can be collected both in qualitative (e.g. verbatim feedback) and quantitative format (e.g. TNPS - Transactional Net Promoter Score or CSAT - Customer Satisfaction).

Depending on the insurance business’ ecosystem, further automation can be applied. The expert chatbot, when powered with CPA processes, can directly open claims within the insurer’s claim system but is also able to recognise simple “fast-track” claims and potentially settle them with or without human supervision.
Therefore through CPA, the insurer can potentially route complex claims to human, simpler ones to touchless journeys and possibly extend the conversation with a group of customers to offer additional information or services.

Simultaneously, the conversational data gathered by the expert chatbot are stored in the CPA platform and ready to be analysed. The performance of the chatbot is now visible through graphs: behavioural patterns are identified, conversations can be analysed at a granular level according to demographics and feedback processes in a single space. The conversational experience can then be reiterated according to such data-driven insights.

See how Spixii partnered with Duck Creek and Blue Prism to offer an end-to-end quote & buy solution.

Example for Customer Service

Similarly to Policy Administration, Customer Service usually consists of very simple and recurring queries (roughly 60% of requests). Normally, such questions are either answered within website pages (e.g. FAQ systems) or delegated to call centres. Most of the time customers and policyholders end up queuing on the phone decreasing the insurer’s esteem.

CPA can employ expert chatbots within insurers’ web pages to deal with most of the customer queries. The chatbot is able to provide instant answers and of course available 365/7/24 when call centres are not. When queries are complex, they can be routed to a human operator that can better deal with them.

CPA and data protection

Insurance businesses collect and process a huge amount of data. While it can be used to inform decision-making, optimise the performance of processes, mitigate risk and meet increasing customer expectations, data is also critical for regulatory purposes.

Insurance is one of, if not the most regulated industry. Therefore, good data governance and practice is expected to be in place as part of every insurer’s compliance program. This not only means alignment with general data protection (GDPR) guidelines but more strictly to the insurance distribution directive (IDD) as well.

In order to comply with the high standards defined by such regulators, it is mandatory for every insurance business to have a coherent data and analytics strategy along with a solid data governance framework with transparent and fully auditable data management standards and controls.

At Spixii, the CPA platform is based on a fine-grained access control (FGAC) framework to guarantee security at a granular data level. Each data item has its own access control policy with fully customisable access to it. Such a strategy empowers insurers to control and set different data access for their employees or third party companies according to their function and their level of responsibility.

When it comes to data storage, Spixii CPA platform adopts different data strategies:

1. Data minimisation: Spixii CPA reduces sensitive data to collect and store to the bare minimum
2. Data anonymisation: all the information collected through Spixii expert chatbots is unidentifiable to an individual or entity
3. Data retention: Spixii CPA features data policies with deletion of sensitive data with totally flexible timings according to its client needs

On top of this, Spixii CPA platform applies data segregation with the FGAC framework to secure each insurer's data in a separate database.

A thorough and deep comprehension of data regulations was required to bring the Spixii CPA platform to an enterprise level. But also the ISO 27001 certification, periodically renewed by the Company, made a great impact in increasing the security of the software solution and data protection.
In particular, the ISO standard recommends performing scans of vulnerabilities, penetration tests, static and dynamic security analysis of the application and continuous review of the OWASP reports.

More details of Spixii certifications are available on the Compliance and Security webpage.

Addressing FCA’s priorities with CPA

Vulnerability is one of the most discussed topics in the insurance industry and one that is demanding the most drastic changes for insurance firms.

According to the FCA’s Financial Lives 2020 Survey, there are 27.7 million adults with characteristics of vulnerability in the UK, meaning more than half of the adult population (53%), giving reason to FCA’s urgency about protecting vulnerable customers.

In the table below, FCA’s high-level vulnerability principles are reported together with the concrete ways in which chatbot technology and Conversational Process Automation (CPA) can uphold those principles and meet requirements.

Michael Sicsic, Spixii advisory board member and former head of supervision at the FCA said:

Firms struggle most with design and organisation change necessary to identify, track and use the most relevant customer information. Communications to vulnerable customers needs to be tailored –with different trigger points and different wording according to circumstances. The more information a customer shares, the better the organisation is able to respond, and mitigate the impacts of a customer’s degree of vulnerability.

At the same time, relevant data on vulnerability must be systematically collected and recorded so it can be used by frontline staff. Chatbots and Conversational Process Automation (CPA) can support insurance firms to achieve both goals, the best customer and business outcomes.

As a tangible example, in 2018, Spixii designed a multi-award winning expert chatbot system called Zara for Zurich UK to help them manage their claims. Anne’s entire testimonial using the expert chatbot can be viewed below where she also explains the context of filing for the claim.

To deepen, read the blog How insurers can meet the needs of vulnerable customers with CPA and chatbots