AI-driven chatbots are set to save industry $8 billion per year by 2022, according to Juniper Research. Those gains will be driven by the flexibility, conversationality, speed, and a sense of personal customer engagement that natural language processing gives to chatbots.
But chatbots represent just one component of a comprehensive self-service portal. Immediate and cost-effective service has proven to be impossible through call centers and human chat. Self-service portals are the comprehensive answer to that need for immediacy.
Despite that overwhelming demand for immediate help, more than two-thirds (68 percent) of businesses say live chat and phone agents struggle with the volume of customer requests and only 49 percent of customer problems are solved on the first interaction. Robust self-service overcomes those shortfalls.
Tapping Into the Power of AI
Artificial intelligence (AI) is key to powerful self-service portals — it exponentially increases insight into the customer’s problem and by powers lightning-fast answers.
AI improves enterprise knowledge about the user, understanding of the user’s current need, and the findability of information that addresses that need. And within a sea of knowledge found in so many portals, chatbots and unified natural-language search bars take users quickly to the exact information that they seek — drastically reducing labor costs while improving the quality and speed of information provided.
Chatbots and intelligent search bars are two of the fast-evolving channels within the online self-service portal. Forrester Research also includes websites,mobile apps, web chat, messaging, and email in its “lower-cost digital self-service channels” that constitute the self-service portal. Those channels are fed by content sets: Typically, knowledge bases, FAQs, user communities, order history records, and download centers, all supported and unified in presentation by intelligent search.
From Old Tech to Super-Automation
Telephone customer-support centers are notoriously expensive, and they frustrate customers. Call center deflection is a long-standing practice of providing users — both employees and customers — with problem-solving and satisfying alternatives to the telephone.
Human agents and old methods of automation (such as touch-tone entry prompts that serve recorded voice solutions to an array of common requests) are costly. Contact center software provider Aspect Software estimates that the average cost of a customer service phone interaction — including frontline labor, managers, call-management software, hardware, utilities, and real estate — is around $35 to $50 per interaction.
Text chat, on the other hand, averages $8 to $10 per session. Other research by Forrester finds companies spending 30 percent less for web chat, which allows reps to communicate with multiple clients simultaneously and include explanatory content that can’t be conveyed via telephone.
Evolution of Self-Service Tools
Call center deflection already saves an average of $8 million a year for a large enterprise; as an added bonus, fully 80 percent of customers prefer them, according to a survey of customers of the 25 top U.S. utility providers in February 2018 conducted by ForeSee, the provider of voice-of-customer solutions.
Over the years, many mature self-service channels have evolved and energized the self-service portal:
- Tightly scripted chatbots
- Conversational natural-language chatbots
- Google and Alexa voice skills
- Virtual customer assistants
- Intelligent search bars
- Interactive knowledge bases
The tools that you select will be the ones that make your self-service support portal most effective. To ensure an effective portal, you must:
- Optimize the user experience
- Engage and empower users
- Reduce and reallocate labor from support to analytics
- Choose your metrics for success
Candidate metrics for success might include the ratio of visits to tickets, the number of callbacks, the number of interactions per ticket, the ratio of new users to existing users, standard web engagement metrics, search analytics, trends in the average cost per incident by support channel.
Self-Service Done Wrong
Despite growing customer desire for quality self-service, adoption has been hindered because overall customer experience is lagging. Forrester’s Customer Experience Index for 2017 found that year-over-year customer experience quality is worsening, not improving. And part of that is attributable to inadequate self-service.
We have known for years about the pitfalls of inadequate online support, which leads to inadequate call deflection.
A Nuance Communications survey, for example, found that 71 percent of consumers would prefer to use a virtual assistant over static web support. But those consumers ran into trouble. Before recent advances in AI-driven self-service, more than half of consumers reported being unable to resolve their issues on the web.
Those who did find a solution, said it took too long. Those who didn’t find a solution, felt frustrated at having to call a live agent.
To realize both ROI and high customer acceptance from full self-service portals, enterprises need guidance from those who have succeeded with automation.
Unified cognitive search “opens the door for personalization,” which is a major step forward for self service.
Veritas Maps Path to ROI
Like other companies following a de-merger or spinoff, data management firm Veritas found itself forced to re-engineer systems that had been cloned from its former parent company. With systems and data scattered across multiple sites, Veritas struggled to
- Reduce call-center expenses
- Filter internal search results by user
- Exclude rivals’ search results
- Coordinate knowledge management and support resources
- Make search smarter and more intuitive for users
Perhaps the biggest challenge of all was to present users with related Q&A, knowledge base items, product documentation, case data, downloads, and searches — each with its own vocabulary — in a unified view of related items derived from different sources.
Veritas valued the benefits of Solr open source search. When Google Search Appliance was discontinued, Veritas needed an alternative. It eyed Lucidworks Fusion for search capabilities that promised to unite disparate content sets, understand Veritas’ data, and predict users’ intent to boost search-result relevance.
“Fusion gave us the ability to index content from different systems and present the unified results,” says Joe Kugler, senior manager for knowledge management and self help experience at Veritas. “But if a user clicked on those results, they would actually end up going to that source system. Our new services layer, kind of like magic, was able to connect all those things and provide the right data flowing through the APIs [instead of jumping from system to system].”
Fusion provided what Google could not: A unified, filtered, and organized view of seemingly disparate but closely related content sets. Having a taxonomy that knitted together the different terms and rules for the same ideas was critical. Fusion’s Taxonomy API unified Veritas’ content collections and enhanced Veritas’ site search.
“Fusion was able to bring all that together into a consistent taxonomy,” Kugler said. “If you don’t have a taxonomy office or something similar, you end up with one standard for your software downloads and another for your product names, your documentation has something different again, and your price book is different.”
Fusion also provided Veritas with natural-language search across its content sets, so users can skip the phone and find the right help and supporting data swiftly.
Unified cognitive search “opens the door for personalization,” which is a major step forward for self service.
Personalized search, of course, tailors each user’s view of support and knowledge-base content, further speeding the user’s quest for quick, accurate, and comprehensive answers. Veritas is still in the early stages of personalization, using software downloads and maintenance contracts to entice users to create profiles and log in.
Personalization won’t work well if silos aren’t configured optimally.
Based on his past IT roles and present work at Veritas, Kugler says, “I think a lot of companies wrestle with where do knowledge management and support belong — in separate silos or together — and I think it’s a huge advantage to have those two entities together. Then you’re not dealing with contention for resources and all kinds of stuff. But if you don’t have a taxonomy that lays over the whole thing, you almost can’t solve it,” Kugler says.
By leveraging open source search and analytics technologies, natural language processing, cutting-edge AI, and strong taxonomy, Kugler says, Veritas was able to slash support costs by 12 percent and reduce support-case longevity by 30 to 40 percent while pleasing users with an industry award-winning self-service portal. The portal gets rave reviews within Veritas, as well; its services layer allows non-technical staff to make changes to the portal on-the-fly, allowing IT staff to focus on genuine technical challenges.
Going forward, Kugler expects to reap additional ROI as personalization is fully implemented and as machine translation makes Veritas’ content available in nine languages.
AI Benefits Across Silos
Before Veritas’ application of AI across the breadth of its self-service portal, a number of enterprises were developing AI-driven chatbots to provide specific brand-defining services more economically. These implementations offer a starting point to extend AI-powered search and automated chat broadly across the user experience.
- Capital One Financial’s Eno chatbot interacts with bank customers through text message. It provides account information and assists with payments from customer smartphones. The chatbot builds customer engagement without human labor costs.
- Expensify CEO David Barrett says the company’s Concierge chatbot, which guides new users through expense-report setup and proactively troubleshoots impending issues, has helped reduce banking problems by 75 percent and that the website has quintupled the number of its free trials.
- Amtrak’s Julie chatbot helps customers find information on reservations, rewards, station locations, and routes. Among other revenue boosts, the chatbot is credited for $1 million in annual savings related to customer service email costs.
In all these cases, AI-driven chatbots have combined with improved search to unify information and tools from different silos into a concise user view.
Best practices in call center deflection
Whether you obtain a self-service portal off the shelf, or develop a custom portal in-house, you are going to want many of these key features – and the requisite technologies to support them:
- User profiles for personalization. Lucidworks Fusion gives users contextual, personally relevant search results and recommendations through integrated AI.
- Up-to-date marketing and sales information
- Knowledge and learning center
- Helpdesk and FAQs
- Online marketplace
- Collaborative space
- Forum community
- Chat services
- Service requests (and high-value transactions, if necessary)
Quality chatbots alone require:
- Machine learning to interpret a large volume of past requests and correlate them to past solutions
- Natural language processing to interpret current requests
- Logic to map current requests to past equivalents
- APIs to provide the detailed solution to the user
Identifying and implementing these technologies requires experience — and data to train and experiment with. For some enterprises, past live chat services provide valuable training data for an advanced chatbot or virtual assistant.
Refining the Completed Self-Service Portal
AI-driven self-service naturally causes a shift in enterprise staffing, from old-fashioned support to analytics.
A report by Forrester Research offers five ways to measure digital-to-phone deflection and address the reasons why users resort to telephone support.
- Map customer journeys and flag triggers for phone calls
- Analyze cross-channel leakage analysis – for example, from FAQ to phone or knowledge base to chat. Include transcripts.
- Map phone numbers to specific digital touchpoints, to identify successes and failures
- Bridge web- and mobile-app-initiated calls with call identifiers
- Connect web journey and phone data for intent determination, to know where specific cases are occurring on the phone and web at the same time, to better ascertain and predict intent.
Here’s a macro example of the measures done by individual enterprises: About 55 percent of contact center decision makers surveyed by Dimension Data in 2017 (PDF) expected to see a significant decrease in phone volumes over the next two years, while more than 75 percent of them believed that full automation of customer service interactions will rise.
Don’t Save Money, Make Money
AI-driven self-service drives profits higher, and those profits, properly reinvested, drive even more improvements in customer satisfaction.
A 2016 study by Fifth Quadrant, commissioned by SugarCRM, found that:
- 55 percent of business decision makers believe chatbots improve key KPIs, especially online sales rates, customer satisfaction, and operational efficiency
- 77 percent of consumers believe that immediate online help would increase their likelihood of completing transactions online more often
- 85 percent of businesses believe that immediate online help would improve online sales conversation rates
A 2019 study of 314 marketing professionals by Evergage and Researchscape International found that major retailers, such as Carhartt and Zumiez, are successfully using machine learning to boost online sales conversions.
The retailers use machine learning to consider individual shoppers’ past session activity and purchase history to customize the user experience in real time at every step of the shopping journey. Zumiez’ personalized AI recommendations increased conversions and boosted spending 2 percent per order. Carhartt’s instant recommendations generated a 5 percent increase in click throughs to its product description pages and increased conversion rates seven times over the previous triggered email solution.
Powered by AI, self-service portals become not only a customer-support tool, but also a terrific sales opportunity.
With $8 billion in industry gains anticipated in the next two years from AI-driven chatbots, and further gains foreseen from applying cognitive search to the other tools of a self-service portal, now is the time to apply the insight and speed of AI to user support as well as sales.