“If I had asked people what they wanted, they would have said faster horses.”
- Henry Ford
Chatbots have come a long way—from rules-based to AI to human-first hybrid. From serving content to being engines of inbound marketing, demand generation, and account-based marketing.
However just a decade back, few businesses could have predicted their websites needed bots at all. When bots were implemented, it was a case of putting the cart before the horse; bots can only go so far in engaging a customer yet too often they are left on their own sans human intervention and end up turning away prospects and business by virtue of their own limitations.
That probably explains why 40% of first-generation chatbot/virtual assistant applications launched in 2018 will have been abandoned by 2020 (Gartner). Then again, Oracle predicts that by this year, 80% of businesses plan to utilize chatbots. So what's changed? The bots have. Most proactive enterprises have already shifted to AI bots, but there is still a lot left to be desired.
The vast majority of chatbots in use today are not AI which is problematic given that customers are expecting state-of-the-art AI bot implementations at home and in the office. But on the B2B front, except for a few niche use cases where some AI bots are performing well, most of them aren’t living up to customers’ expectations. The Gartner report states: “The market is awash with low-end VCAs and chatbots that deliver poor user experience, create friction and don’t deliver business benefit… Only the enterprise-grade VCAs that create a compelling user experience and deliver truly material business value will survive.”
It's true that AI's self-learning mechanisms take on some of the heavy lifting. But whether your business’ pet dragon is rules-based or AI-powered, a human agent makes a tactical difference.
The following common causes for chatbot fails to make the case for human intervention:
1. Lack of training
Chatbots are only as good as their training. They are trained with training data. That's not as easy as it sounds. Larger enterprises that depend on AI and ML report that they've had to push pause on projects due to issues with the volume and quality of data needed to train their AI bots. For a bot to be effective, it needs to train with massive volumes of data from a range of conversation datasets and then an entirely original dataset based on the specifics of your business which can be gleaned from past conversations. This data needs to be labeled and annotated well for it to be of any use to the bot, and unfortunately, much of the quality issues reported by large enterprises using bots relate to the labeling and annotation of data. Annotation and labeling apart, the approach to training appears to be: use and forget. Without updates on utterances and relevant responses, your bright new bot will end up a fossil sooner than you expect. 81% of respondents said that the process of training AI with data was more difficult than they expected.
2. Ignoring or misreading the intent
Training is also about mapping phrases to the right user intention. Without this, bots are off on the wrong foot and without continuing training, the bot will botch future conversations. For a bot, capturing intent involves peeling layers:
- Tokenization: Breaking up a sentence into words and parts of speech.
- Understanding a word’s meaning: If a word pops up that doesn’t feature in a bot’s lexicon, it has to make an educated guess based on context and syntax or be forced to ask for clarification. When a word happens to have several meanings, bots have to pick the right one or seek more info.
- Understanding syntax, semantics, and pragmatics: Sentence structure can be hard enough for humans. But a bot’s cred depends on identifying the relationship between words and the meaning of a sentence. Pragmatics is an extra layer of context through which a bot can better arrive at the intended meaning of a conversation.
While bots peel these layers, users shouldn’t and don’t care to observe the rules by which chatbots operate. Casual abbreviations, colloquialisms and everyday rogue grammar will stump the best-trained bot.
3. Lack of context
Context is an awareness of the user’s current state and picking up the conversation where it was left off. That requires persistence and memory. However some chatbots are also programmed to limit access to data generated which negates the key purpose of having bots in the first place. A simple solution to this is having a well-integrated MarTech stack. Having your conversational marketing platform speak to your marketing automation and CRM can deliver chatbots more context than just input data.
The language context is another beast to grapple with. One phrase/word can have multiple meanings, and if you add sarcasm to that, the bot hits a wall. Once again a healthy dose of training and AI self-learning can help bots wing it to an extent. And let’s not forget what needs to be done if you want your bot to take on multiple languages. Another round of training and updates.
4. Inability to integrate with third-party software
Chatbots have become a critical tool of conversational marketing because of their demonstrated capability in demand generation, but that potential zooms to the next level with APIs that help integrate with a business’s CRM, backend, and intelligence systems. This access to a business’s processes helps to actively leverage dormant data intelligence while a conversation is underway with a prospect.
And for marketers especially, integrating disparate systems is one of their most challenging barriers to marketing technology success (Ascend2).
Unsurprisingly bots that aren’t too confident with their security testing, forgo API integrations altogether. This only makes it harder for the chatbot to process user queries and create value. But more practically such API integration becomes necessary, especially with marketing automation platforms in order to capture context. This at the very least helps keep track of the conversation thread that is being picked up after a gap. Software integration also helps in rapid lead qualification which is important given that a business has the best chance of qualifying a lead when they reply within the first five minutes.
5. No protocol for escalation
Ever wanted to bypass the bot and get straight to the human agent? Ever failed to find a way and ended up shutting down the chat window? For a B2B bot, the omission of an escalation clause can be enough to push prospects over to competitors.
Insent’s human-first conversational marketing platform bypasses these fails by prioritizing the human agent over chatbot for high-value prospects and programming seamless handoffs from bot to human once fresh prospects have incubated. Moreover, Insent’s secure and seamless third-party API integrations make the most of existing insights and sentiments, even as it feeds in pertinent data.
The future is a human-first hybrid. The future is already here.
Chatbots may be making their way along the hype cycle but consumers are already expressing fatigue. Most consumers covered in the 2019 CGS Customer Service Chatbots & Channels Survey said that AI was simply not as helpful or empathetic as a human agent. The largest percentage (13%) of those surveyed described AI customer service as ‘maddening’.
Just as revealing: The 2018 CGS Global Consumer Customer Service Report showed that half of the respondents looking for a quick answer would choose chat over all other channels, but by 2019, only 29 percent picked that method of communication; and 40 percent of respondents chose phone or voice first.
In the final analysis, bots are meant to assist humans; humans don’t assist bots, though they may well be trained by bots. In conversational marketing, bots are a tool to augment human intelligence, enhance human capacity, and advance the customer experience.
Insent harnesses your pet dragon bots and helps them soar in tandem with your sales reps in converting target accounts into dream customers.
Book a demo today and witness how your website can take wing to become your strongest sales channel.
Insent.ai is a human first B2B Conversational Marketing platform that enables marketers to deliver red carpet experience to high-value prospects.