The Advantages of Chatbots: All You Need to Know
Chatbots have been with us for 2 decades now in various forms, initially leveraging very large databases of answers to questions that were matched with rules, that transitioned into intent matched through technologies like NLP and NLU, onto intent that was determined on large volumes of training data, and now AI, where large language models (LLMs) and their attendant capabilities to summarize, pinpoint match, cluster, and, of course, generate answers, means that bots are now at a point where they could provide the digital front door to company websites, to online shopping portals, and to organizations that manage information dissemination. But will they be the tool that delivers a multitude of advantages and provides a digital face to the world?
This article will be the first of a series on digital tools that deliver platforms for automating and lifting human involvement out of lower-value work and onto more impactful work. So, we will walk through how expectations are moving toward how well conversational AI is being mobilized and why organizations should still bother with the bot, even if the generative powers of LLMs, like ChatGPT, appear to be able to sweep all comers before them.
But first, the advantages of Chatbots and thus the drivers that organizations are using to justify business cases The table below provides the common advantages of a chatbot, its drivers, and some of its disadvantages.
| Features | Benefits |
|---|---|
| Provide Customer Support 24/7 | Customer Service accessibility. Lower costs to provide the same manually over the same time windows. |
| Answer FAQs automatically | Intent determination is key, however, a better experience is received at a much lower sustainable cost. |
| Enabling the team to offer their best service | Chatbots create more time for agents to tend to conversations that need human touch and intuition. Improved customer satisfaction on both counts leads to greater customer lifetime value. |
| Sales can be made directly through omni-channels | A transaction-capable bot on social channels, mobile apps, as well as the web channel, and adjusting conversations to suit, enables 24/7 sales across channels, providing incremental revenue and cash flow. |
| Pro-actively support the purchase decision | Providing premium content, and recommendations and using other sales techniques in the purchase journey increases revenue and cash flow. |
| Mitigate abandonment in the customer journey | Chatbots can provide informed input at all steps in the purchase journey, preventing late-stage abandonment for primary reasons, like excess charges at checkout. |
| Reduce Stress for Employees and customers | By handling a portion of the call and chat interaction pool, Chatbots reduce volume pressure on employees and (indirectly) customers. This translates into direct lower resourcing, but more importantly, it directs human effort where it is most needed. |
| Customers can organize through the chatbot | Customers can graze across channels, reviewing, selecting, and booking trials and appointments to try in person, all facilitated in the same conversational thread across channels by the bot. |
| Extend the brand | As the potential digital front door, chatbots can create brands, extend brands, and support brand style without losing appropriate conversational tone across channels. |
| Multi-Lingual Support | Chatbots can detect language, perform translations, and determine intent and service in the original or selected language to sustain multi-lingual interactions, removing the need for manual translation. |
| Disadvantages | Reasons & Mitigation |
|---|---|
| Escalating away from the Chatbot | Persistent questioning by the bot before an escalation to a live agent is frustrating. Radio buttons presenting the availability of Live chat handoff should be offered routinely throughout the journey. |
| Setup cost & time, reliance on programming skills | Traditionally, chatbot setup and maintenance have proven costly, time-consuming, and required specialist skills. Low-code program languages, the use of LLM techniques, and business user training deliver more supportable solutions. |
| Emotionless Support | Supporting efficiency without personality does not suit all stakeholders. Bot conversations need context, asynchronous engagement, ‘small talk’, and human crafting of responses to enable a trusted conversation. |
| There is a limit to functionality. | Bots cannot cope with every possible conversational path. Integrations for personalization, conversational design, feedback, and learning tools enable ever-growing bot conversational fluency. |
The emerging view for chatbots is that LLMs, like ChatGPT, will replace conventional bots. The generative powers of the LLM and its successors, enable the consumer, student, and professional to have a ready base of knowledge. So informed chatbot users will be the norm, even in the short term.
In this context, when organizations use chatbots now, the premise to start designing the bot is the intention to work with LLMs and not as a separate tool. In later articles, we will explore how the LLMs are being used to provide context and control on ‘guide rails’, so the messaging from all organizational channels is managed and consistent.
If we return to how to take advantage of chatbots, then there are a range of considerations that work toward achieving the advantages. Primarily, this is good conversational design and practice. We shall endeavor to cover with the following suggestions that help good conversational design and practice:
Frequency & Presence
To take advantage of the 24/7 availability and omnichannel presence, a bot needs to surface across a range of channels, delivering content appropriate for the channel, and appearing across multiple pages to meet the expectations of modern digital lifestyles. To support this, the bot architecture needs to support a multi-bot framework, where centralized control can manage any manner of domains, business units, channels, brands, and content relevant to the channel.
Context Management
In order to conduct a normal conversation, recognizing the context of the users’ prompt leads the bot to respond, not just in what is matched for intent in a knowledgebase to deliver the response (via NLP, NLU, or other matching tools), but to logically build on prior conversation prompts and responses, or context, to form a natural conversation. You may expect all bot vendors to deliver this capability, but prior responses, even in an asynchronous conversation, are an emerging expectation, not always delivered by the marketplace.
Context is even more broadly relevant with the blending of content from LLM into responses the bot can offer. LLM responses, uncontrolled and delivered through a bot run the risk of messaging counter to organizational direction, policy, product plans, pricing, or other commercial factors the organization shares privately with customers directly. Managing how LLMs can provide context, which is managed by the organization before releasing on bots by chosen channel, is an emerging challenge and will be treated in our forward series.
Conversational design by selected tone and language of the delivered response, appropriate to the channel is the next key capability to provide. Bots generally have personas, so differing language tones between channels are both a challenge for consistency, but also food for marketing and their key need to be involved in content authorship and publishing. A key expectation for a bot in online retail, government services, wagering, and financial services is for the bot to enable recommendations and to accompany bot conversations with immediate navigation to pop relevant web pages for the user to inspect, whilst the bot offers tips and conversation bridging to enable next step information and carry the conversation across the digital journey.
Personalization
Layering over these capabilities is the further expectation for personalization, the ability to have personal details interwoven into the conversation. In this respect, the bot can be managed to seek account or log-in credentials to assist the customer through the conversation whilst preserving both security & privacy. The challenge for the bot in the conversation becomes managing content relevant to the customer’s questions, personal details, and conversational health, either prompting the customer with additional information to move the conversation forward or gracefully enabling an escalation to a live agent to hand off to enable conversation development and closure.
Achieving this capability speaks to the need for integration into single or multiple sources of data, depending on the organization’s management of customer data. This is not a small matter and often leads to weightier considerations of capital in both investments. Simpler and versatile programming codes like Python lend themselves well to the dexterity needed to juggle the ‘few balls in the air’ required to conduct moderately complex conversations. The advent of low or no code, integration as a service, also supports this complexity by enabling integration from diverse legacy systems that make up the ‘whole of customer’ view.
So, for this week, the above suggestions are but a few of the key aspects of conversational design needed to support bots that address the emerging expectations of conversations in modern digital journeys.
Next time we will tackle the prospect of automating voice operations, so that predictable conversations can, in fact, be conducted by machines, with only some involvement from humans…. and still sound as if a live ‘human is in the loop’ … and without resorting to TTS. You can also learn more about chatbots by getting in touch with us and scheduling a demo.