Self-service is becoming essential, but the best strategy is to guide the customer down the preferred method without making them feel stuck without the option to engage with a human.
As a growing number of companies turn to artificial intelligence (AI) and automation to drive customer self-service beyond its traditional supporting role, it remains crucial to get the fundamentals right, execute correctly, and provide a seamless path to human escalation in order to prevent additional customer frustration.
Customer service itself is not a new phenomenon; it has just got a whole lot easier as technology has evolved. Before the digital age, customers were given the ability to independently solve problems by reading the printed manual that came with the product – though not that many of us did!
Then came the early analogue and digital interfaces to customer self-service, with examples such as telephone banking, which used Dual-Tone Multi-Frequency (DTMF) tones to return balances and perform basic levels of transactions. The introduction of mobile-based GSM enabled the use of SMS and USSD codes to perform some transactional self-service. Enter the age of 5G, high-speed internet, mobile applications and e-commerce, and there is not much we cannot do to address our needs through self-service.
In a report titled “State of the Connected Customer”, Salesforce states that 61% of customers would rather use self-service channels for simple issues and that an average of 54% of customer issues can be solved by organisations that use them. The reality is that companies delivering a service or product should already have some kind of self-service facility in place. The real differentiator requiring attention and consideration is how the self-service is delivered to meet customers’ expectations, using the next generation of tools available, which now include AI functionality.
Self-service in customer service
Within the context of customer service, self-service is about providing the customer with a digital interface, or communications channel of their choice, to find answers to questions about a business, their products and services, and carry out transactions without any human intervention. This type of self-service usually comprises three distinct pillars: Knowledge, Transactional, and Community:
- Knowledge-based self-service: Digital interfaces such as web portals and frequently asked questions. The introduction of Large Language Models (LLMs) has helped transition this from a static list toward being able to address conversational queries, while omnichannel has extended this functionality into messaging services such as WhatsApp for Business.
- Transactional self-service: Usually provided by secure portals where customers are authenticated and can manage their personal information, track orders, download invoices, submit claims, and make account changes, for example.
- Community self-service: Tends to consist of business-moderated customer forums where end users can troubleshoot their issues and share knowledge with one another.
Over the past few years, South African companies in specific industries have made significant strides in the implementation of self-service functionality for their customers, either through web-based interfaces or mobile applications.
Functionality provided by the apps of local financial services providers compares favourably against global standards. In the insurance sector, self-service is currently used for policy retrievals, claims submissions, and agentic-based call routing to replace complex IVRs. Internet Service Providers utilise portals that allow customers to manage password and account resets, perform feasibility, manage and track internet service orders, and change account and invoice information.
Advancing self-service with artificial intelligence
The introduction of AI has been very effective in advancing self-service. Let’s use the example of a case where a customer wants to reset their password for an account: traditionally, the customer might have had to turn to the company website, with articles that provide all the steps and links needed, or a web chat with an IVR-style (interactive voice response) menu of options to choose from. All of these will have been delivered with a rules-based experience and will have likely resulted in a few “I don’t understand your request” errors for the end user as well.
Add in AI, and the experience becomes more seamless: the customer could engage with a WhatsApp-based virtual agent through text or voice and be assisted while using natural language. This would be powered by LLMs, which take the customer’s request and, using automation tools integrated into backend systems where the account and password reside, facilitates verification (perhaps through a one-time PIN) and helps the customer reset their password. This moves the self-service process from simply providing the customer with more information on what to do to actually helping them accomplish what they wanted.
AI is also being used in a very practical way to enhance self-service: the use of virtual receptionists with the ability to not only offer self-service via the phone but also – and most importantly – direct customers through to the right person or department without the headache of going through multiple layers of menus found in a traditional IVR. A great example is Telviva’s digital agent, Viva, which not only uses company information to assist customers from a sales and support perspective but also connects callers to the right people and teams within the organisation while supporting the use of natural language in multiple South African languages.
Considerations before implementing AI-enabled self-service
Seamless access to real-time company data is mandatory, particularly when transactional interactions are taking place. When looking at instances such as banking, foreign exchange rates, and package tracking, we can see that access to real-time information is not a new requirement, but this is improving with the addition of a more conversational interface.
In the world of customer service, where a knowledge base is used as the reference point to provide support to a customer, there are five key considerations around knowledge bases for business ahead of implementing self-service:
- Structuring: The knowledge base is the master source of truth that various digital interfaces will be trained on, and as such, it is crucial that this data be structured correctly. Knowledge bases should be structured as per vendor guidelines so that keywords, metadata, headers and more all work toward making responses as accurate as possible.
- Security: The knowledge base should have different layers of security to ensure that only public-facing documents are made available to the self-service engine. This is a crucial component of the ISO27001 security accreditation process, which requires that all documents receive a security level classification.
- Document versioning: Proper versioning ensures that multiple different documents with slight variations or updates are not added to the knowledge base, which helps prevent virtual agents from providing incorrect information or hallucinations by allowing them to draw their own conclusions.
- Detailed ticketing: Resolutions – even those achieved through traditional support means – must be documented in tickets and then converted into knowledge base articles so that the learnings can be used to further enrich self-service channels. However, care should be taken to ensure that personal or specific information is not added to such articles.
- Restricted access: The virtual agent or customer service interface should be prevented from accessing public information and instead rely solely on the data from its own knowledge base. This controls the output and ensures that the LLM is only used for its conversational ability to better engage with customers.
While using proprietary company data helps avoid hallucinations, it will still show up in a poorly structured or limited dataset. As such, it is crucial to ensure that all parts of the business contribute to this dataset. Researching the most frequently asked questions, or tickets by category, and writing articles about these particular topics will ensure the customer has a good experience and has access to the latest information.
The human in the loop
While self-service options can help in handling simpler engagements, the customer might, during the course of an engagement, divert toward a more complex request. Self-service should never attempt to manage these requests, and clear rules and guardrails must be put in place to ensure that such escalations are handed over to a human agent. Some people may also prefer to only deal with humans as a matter of preference.
A sure way to an unhappier customer is by not allowing an option to interact with a human, no matter how simple the business may believe the customer’s request may be. Even worse, this may force customers into turning to competitors who are prepared to engage on their terms.
This is where a good omnichannel solution, such as Telviva Omni, comes into play by bringing together various communication channels, tracking the customer journey and interactions end-to-end, and facilitating seamless handovers between virtual and human agents. This means that the customer is not forced to redo identity verification checks and repeat themselves to the human agent.
This means that human agents primarily become specialists who deal with more complex engagements while self-service tools handle standard, repetitive requests. As such, agents will require advanced emotional intelligence (EQ) to handle complex queries from often frustrated or irate customers, alongside sharp root-cause analysis and critically verifying AI outputs for accuracy. Without a script to fall back on, they will have to think like investigators, connect dots across various business departments, and solve non-standard problems.
Human agents will also be required to curate and enhance the quality of the knowledge base to continually improve the performance of self-service tools that are intended to not only help customers but also assist human agents using AI assistant tools for support with complex enquiries. Going forward, they will require digital literacy in prompt generation and in knowing how to query an AI assistant effectively during a live call and have the ability to perform critical fact-checking to validate AI suggestions before presenting them to a customer. They will also increasingly need basic technical writing skills as they transition into knowledge base authors.
The new metrics for success
Businesses currently measure First Call Resolution (FCR) – ensuring the ticket logged is responded to once and closed at the same time with a positive resolution – as a key metric to determine the performance of the customer support team. Going forward, organisations are looking at Zero Call Resolution Rate for measuring how many interactions are successfully handled by various self-service tools.
This will require a strong, independent Quality Assurance (QA) engine to determine this with a level of accuracy, as it goes beyond a traditional deflection marker, which is the measure of interactions handled by customer self-service versus human agents. Here, it is crucial that there is segmentation between human and virtual agent customer satisfaction (CSAT) ratings in order to avoid a blended view and a drop-off in virtual agent satisfaction ratings.
The reality is that until recently, self-service has played a supporting role to the more traditional methods of customer service and support. The customer had the choice of going the legacy way (calling customer support or sending an email) or using digital self-service (mostly online FAQs at the time). However, with the introduction of conversational AI tools, especially voice agents, customers in the future will find it difficult to differentiate between the blurred line of traditional legacy support and modern self-service. They might not even be given a choice once both begin to “feel” the same.
This could be really good or really bad, depending on how well the business executes its self-service strategy. It’s good when done properly because it means that customers have access to 24/7 customer support and service. On the other hand, if not done properly, it can result in customers being stuck in an endless digital loop of frustration. The guardrails of what can and cannot be done with self-service, even when delivered in a conversational format, must be maintained so that handover to a human happens as required, aligning with the concept of “digital when you want it, human when you need it”.
Combine the efficiency of digital self-service with the personalised touch of a human in the loop to experience the best of both worlds: let your customers access information digitally whenever they prefer, and connect with your knowledgeable team for tailored assistance when needed. Chat with us to get started.