In the past four years, Opus Research has tracked exponential growth in the number of “Intelligent Assistants,” positioned as chatbots, message bots, and automated virtual agents. We’ve chronicled at least 100 success stories at banks, airlines, phone companies, cosmetics purveyors and online merchants. It is now time to address the challenges confronting executives at companies—perhaps the majority of implementations—that are neither successes or failures; just older, wiser and more experienced. Today, we stand at the dawn of the third decade of a new millennium. A self-selecting group of customer care professionals and contact center operators reap benefits from cloud-based technologies. Communications Platform as a Service (CPaaS) and Contact Center as a Service (CCaaS) providers led by Amazon, Twilio and Vonage and bolstered by Salesforce, Microsoft and IBM offer unprecedented levels of “openness” and flexibility along with consumption-based pricing.
Each platform provider rounds out its offer with a “marketplace for microservices.” They give enterprise employees, charged with designing chatbots or automated intelligent assistants, the ability to shop around for the AI-infused resources like analytics, enterprise search, natural language processing or dialog management that they need to bring competitive advantage or foster conversational bonds with customers or prospects. As a result, enterprise developers are not locked into any single service provider. They can lean on Amazon Lex, Microsoft LUIS or Google DialogFlow for basic understanding.
Then they are free to use IBM Watson to search and retrieve relevant answers from an “unstructured data” that resides in Word files, PDFs, PowerPoints or databases that are under the control of CRM, ERP, order entry, reservations or any number of automated processes. They can also expand to any other service-compliant system to tap into heretofore unknown power and operational capabilities.
Choice can quickly evolve into too much of a good thing.
With so many combinations and permutations of resources, and so few companies with long-term experience, success strategies are just beginning to emerge. Here are some words to the wise for companies contemplating the do-it-yourself approach to chatbots and intelligent assistants in the age of CPaaS.
Start small, and define the “human/bot” balance
Early adopters share a common approach. They “start small” with high-impact proofs-of-concept, then scale-up in a controlled fashion by promoting use across more channels and interaction types. This approach calls for development environments, control panels and dashboards that enable business unit managers – the ultimate subject matter experts (SMEs) – to supervise the training of automated intelligent assistants, then monitor and fine-tune their performance and suggest the most promising path forward, showing the organization in the process how to optimize their operations.
The emerging approach and the “solution stack” that has evolved to fulfill on newly discovered needs have revealed some unanswered questions; primarily, “Who owns the conversational contact center?” or, indeed, should bots be treated as agents in the next-generation contact center? Thus far, the most common role for intelligent assistants is that of triage and intelligent routing. This application calls upon the bot to perform tasks that it does best, i.e. rapid recognition of the purpose or intent of the call followed by “intent matching,” which starts with determining whether the IA knows the answer and, if not, where to route the request in order to help an individual complete his or her task.
The best SME for training an intelligent assistant (IA) was an experienced, successful customer care agent; ie, the person most familiar with and engaging to loyal customers. As the role of IAs expands from customer support to bolster the objectives of sales, marketing, operations, and human resources (not to mention crossing channels from traditional chat to “whatever the customer wants to use to contact us”), the mix of skills among development teams broadens along those dimensions and beyond. Experienced brands, for instance, have added professional script writers from Hollywood to give their automated IAs the personalities or personas that best support their brand and messaging. That was unheard of when the first order objective of IAs was to divert tasks from live chat or voice calls.
Get ready for conversations to support more transactions
Taking the CPaaS-based approach to IA development opens the door for brands to think expansively about resources that accelerate the route to transaction completion. Most microservices marketplaces give brands a number of options for what Opus Research calls intelligent authentication: ideally a zero-effort replacement for the cumbersome processes of matching a password or PIN to a claimed identity and getting down to business.
In addition, vendors like Twilio have put an emphasis on establishing trusted communications links, providing high confidence that, on the one hand, an individual is whom he or she claims to be and, on the other hand, that the individual is in contact with a verified representative of the brand. To transform transactions into conversations is a good start, but firms are also discovering that their IAs must integrate information in highly-dynamic databases, such as flight status or product inventory.
After the transaction, the same holds true about order status, to answer such a common question as, “Where’s my package?” The openness and flexibility of CPaaS is more conducive to adding the connectors necessary to support new features and functions of the next IAs. All of these must be secure and trustworthy by design and placed in the proper context of each conversation as it is created.
May you survive and thrive in interesting times
According to Wikipedia, there is no Chinese curse that corresponds with the English adage: “May you live in interesting times.” Yet here we are. Brands feel compelled to bring chatbots or IAs into their customer care or digital marketing plans. CPaaS providers are more than happy to present them with a plethora of technologies and solutions, initially to speed up the time it takes to launch a bot and then to supervise its training and performance.
To survive and thrive in this environment, Opus Research developed this short check-list:
→ Overcome informational silos: Answers or recommendations rely on speedy access to conversational intelligence, i.e. conversation history, documents (product manuals), data (warehouse inventory) or processes (like travel booking systems) that fulfill the customer’s request with both dynamic and static information – across any channel and for any event.
→ Augment, don’t replace humans: IAs can perform triage functions by understanding the purpose of a contact and connecting to the right resource, making agents more productive, employees more efficient and customers more pleased. The value-add of context and intent from IA makes conversations far more powerful. Integrate with existing performance measures: in many ways, IAs are like digital employees and their performance must be monitored, optimized, and reported against corporate KPIs.
→ “AI” is not always the answer: Successful companies can leverage the investment already made in rules-based solutions and other scripting investments for past Interactive Voice Response, webchat, FAQs and other systems. This applies as companies plan tactics for voice-first, mobile and messaging based intelligent assistance – but also extends to other channels and opportunities.
In these interesting times, it’s easy to get lost when evaluating all the options. To counteract the confusion, the first order mandate is to choose the options that make the best use of existing personnel, IT and contact center assets and to remember that platforms are the easiest, most dynamic way to facilitate the necessary solutions.