How to build a chatbot… that doesn’t get stuck
We all know the experience — you start a chatbot session with a company to get help, the chatbot hits a dead end. You next call the company and talk to an agent, but the agent doesn’t know the answer, tells you to call another number. You call that, hit another dead end, and you give up. We call this the customer service spiral of misery. And just one instance of it will make a customer stop doing business with the company.
So why does it happen? And how can a chatbot help fix the problem? Let’s take a look.
What causes the spiral of misery?
The spiral occurs in larger organizations that have multiple departments with different customer experiences. For example, at a bank the credit card department might have a number for automated reporting of stolen cards, while the mortgage department might have a chatbot and a call center with human agents to sell home loans.
The company gets in trouble when:
- one part of the organization launches a communication channel that handles only a subset of customer questions (an 800 number only for reporting a lost card)
- agents in different departments use different customer service platforms that are disconnected from each other
- teams produce conflicting or inconsistent customer-facing content.
The departments might have the best intentions for their customers but they usually have slightly different goals, have different decision makers with their own budgets and priorities, and are likely to want specialized tools that are perceived as the optimal solution for their needs. They also want to make decisions quickly without having to collaborate (slowly) across departmental lines.
In the end, they offer their customers a mish-mash of ways to solve their problems at different points in the customer journey and on different channels. We equate this to a set of mystery doors on a game show each with different outcomes — some good, some bad, some just dead ends.
And these are the perfect conditions for the spiral of misery to flourish.
How NOT to fix the spiral of misery
Many customer service software vendors want you to think that investing exclusively on their tool suite across the entire customer journey will fix the underlying problems leading to the spiral of misery. They have a valid point: a connected set of tools across all channels would likely produce a streamlined experience.
But here’s the problem with that logic — no single, monolithic suite will truly meet the needs of each department. Not today, not ever. This market is becoming more and more fragmented with best-of-breed tools for increasingly niche categories. So, unless your company adopts an autocratic management system that doesn’t allow decentralized decision making, it will take years of time for your decision makers to agree on tools, and more time to implement them, during which even more players come to market surpassing the tools under consideration. Meanwhile, your competitors have pulled ahead with more modern capabilities.
How a chatbot can fix the spiral
A chatbot is, in its most simplistic form, a natural language interface to get answers to questions. When Watson Assistant launched in 2016, that’s basically what it was — a chatbot service for developers. Over the years, we’ve seen many of our customers pair this core technology with some complex orchestration software to produce a streamlined, personalized, and complete customer service experience. The architechure looked something like this:
We’ve learned a ton about the spiral of misery, about chatbot design, and about system orchestration. Based on that, we now offer a platform that makes building an integrated, end to end solution simpler and faster:
The end result? Watson Assistant can prevent the spiral of misery because it is effectively a one-stop shop for problem solving on any channel that connects to and resolves the fragmented customer experience created by separate departments. It can:
- Repair vague or broken interactions without hitting dead ends
- Cover way more topics by either referencing your existing help content or recommending new topics based on past questions
- Operate in any human language with high accuracy
- Operate on your preferred channels and escalate to agents in any of your preferred contact center tools
- Route to the correct human agent based on the details of the topic and the agent’s skills
- Orchestrate any other system that the assistant needs to interact with to get stuff done
This newsletter would be way too long if I detailed each of these here, so instead I’ve linked the related posts about each topic above. If you want just one post that summarizes everything I’ve mentioned, check this one out!
Hope you enjoyed newsletter post number 1! Stay tuned for more posts to help you build amazing chatbots. (BTW, we like to call them “assistants.”)