West Corporation

Posted on October 27, 2017 by West Corporation 



5 Key Steps to Build a Good Bot

By Aaron Fisher, Vice President of Innovation

Unless you’ve been living unplugged for the past two years, you’ve almost certainly heard the buzz around exciting developments in artificial intelligence (AI), machine learning and bots.

5 Steps to Making a Good Bot - Terminator

5 Steps to Making a Good Bot - Mind Bots

What people think I do.

So what is a bot? When people hear the word, some are concerned we’re on our way to building Skynet, while others think we’re learning to control computers with our minds.

Fortunately or unfortunately, depending on your point of view, we’re not doing anything quite that exciting yet. Today’s developers leverage AI to create a good bot with the goal of improving customer experience (CX).

These are the kind of bots we’re creating at West. While we certainly didn’t meet all our expectations right out of the gate, we learned five important lessons to making a good bot from our first bot pilot experience.

A long time coming

But first, a little history. While bots may be a hot topic today, they’re hardly a new concept. Theoretical AI was first conceived in 1950, and bots have been around for 30 years.

But bots haven’t become mainstream until recently thanks in part to investors who have poured billions into AI and machine learning over the last decade to make a good bot smarter. Now, developments like Apple’s Siri and Amazon’s Alexa have brought bots into millions of homes.

From intelligent IVR to agent-assisted texting, automation and self-service remain the way of the future. And bots are poised to be next in line.

Getting on the same page

Before getting into what we’ve learned about making a good bot, let’s make sure we’re on the same page. If you’re like me, you’ve seen all sorts of different words and definitions for AI. So here’s how we define it:

Artificial Intelligence is any work performed by machines that simulates human intelligence.

Bots fit into this definition. For our purposes, a bot is a software application that can interact with users via a conversational interface in order to perform useful work. But we’ll break them down further into two categories.

Chatbots are simplistic applications that perform Q&A and non-complex tasks. These are easier to develop with a short time to market, but they have limited error handling, which constrains them strictly to their original functions and could create a potentially frustrating user interface (UI).

Enterprise bots, on the other hand, complete more complex tasks or perform analytics to solve problems. They use complex error handling that mirrors existing CX strategies, and some bots’ multichannel design enables users to select their channel of choice or use multiple channels to complete a single goal. Enterprise bots are more complex to develop and maintain, but they will enhance your CX strategy and provide an experience your customers have grown to expect.

Lessons Learned

At West, we strive to create enterprise bots that enhance a client’s CX strategy. We’re getting involved in the bot revolution by piloting a bot with one of the nation’s largest utility providers. The bot lets customers start and stop their service without speaking to an agent.

The pilot has had its share of successes and struggles, but we came away with five important lessons to remember when developing a good Enterprise bot:

1. CX goals are paramount.

When we started the pilot, our utility partner had a five-year strategy and roadmap. They wanted to transform into a customer and energy services provider, which is uncommon among utility companies, who have regulated profits.

The utility wanted to weave together digital and CX strategy and add a disruptive type of technology with a cross-channel, self-service experience. As an AI solution could understand who their customers are, what they want to do and how they want to communicate, we developed a strong partnership based on shared CX and technology goals.

Before investing hundreds of man hours in development, look for a technology partner who shares your vision. Your communication technology partner will have far more experience with enterprise bots and can design your new application with specific business and consumer goals in mind to please the people who matter most, your customers.

2. Start small and find your best use case.

Rather than experiment across every audience and interaction channel, test bots with a single audience segment or interaction type. Then analyze, train and expand from there.

With our bot, we first considered a solution to handle outage communication. But this was already effectively handled through other channels, and we didn’t want to add technology for the sake of adding technology.

So we found a common problem: starting or stopping service, which had previously been achieved through automation with limited success. The majority of these transactions were handled by live agents, so a good bot could create an opportunity to reduce call center costs.

Many attributes make up a good use case, but here are a few criteria to consider:

  • Moderate call volume. Direct your bot to handle a moderately complex task with consistent call volume to gain plenty of data.
  • Power users. Find loyal customers who are more likely to experiment with new technology.
  • Actionable changes. Consider a solution that uses automated technology to reduce expenses or achieve CX goals.
  • Data feedback. Data is the engine that drives a good bot. Think about how your bot will collect data, which you can use to improve its functionality.

Take time when identifying the right use case. Call centers and business functions with lagging technology are good places to look.

3. Create a vision.

Don’t stop with what a good bot should do. Think about how it should do it. Our first bot lacked an end-state vision and essentially became a glorified IVR. We took a step back and created the video below to see what the bot would look like in a perfect world. This became our driving vision.

When piloting a bot, consider how customers will use the product before the development phase begins. That gives you the chance to consider the idea through the public’s eyes and develop a customer-centric product from the ground up.

4. Fail fast and recover faster.

People who use your bot will experience failures. It takes time and training to get a bot to function at a high level. Designing a good bot challenged our established UI tool, which we use to design IVR and drive applications. Our bot had its share of failures, but by adjusting our UI process and keeping the bot running, we’ve made tweaks and continued to collect important data to make the technology even better.

Here are a couple tips to get you off on the right foot. Consider iterative development lifecycle instead of a traditional waterfall approach. Gather data to support regular sprints, and ensure all developers have easy access to this data. Evaluate all your design and development processes and make sure they’re the best fit for the goals you have in mind.

5. Get up to speed.

Yes, bots are awesome, but they may still require some adjustment. Develop a strategy of how you’ll engage your customers and employees once the bot is in place. You have to get people to use it. Otherwise, there’s no data to help it improve.

We suggest you treat your new bot like a call center agent. Start in an offline environment. Then move it into limited production with close monitoring before releasing it in full production. Start promoting with current customers before promoting to potential ones. Loyal brand advocates will likely be excited to help you test a new product that could make their experience even better.

And don’t forget to use the data you’ve gathered to optimize your bot to fit how your customers interact with it. Train your bot like you would a new employee.

Not So Simple

Anybody can build a bot, but that doesn’t mean it will provide the level of service customers expect. Even Facebook has scaled back its ambitions after its AI bots hit a 70 percent failure rate.

AI is built on the premise of perpetual trial and error, so don’t be afraid to fail. Bringing AI into CX isn’t a one-and-done implementation. It’s a dynamic exercise in ongoing improvement to continuously strengthen the relationships brands build with each and every individual consumer and interaction.

Think big, but start small. And before deciding what kind of bot will bring you the most ROI, begin developing your vision with this assessment to evaluate CX across your organizational structure, data and analytics, metrics and technology. Click here to see what your results mean.

It’s an exciting time to be in the CX business. Technology is driving change, and we’re leading the way. It can be hard to keep up, but you’re not alone. Call or text West Interactive Services at 800.841.9000 and learn more about how a good bot can revolutionize CX for your business.

Aaron FisherAaron Fisher has been with West for 23 years, assisting in various capacities across the organization including in client services, business design, information technology and client engagement. Aaron graduated from Northwest Missouri State University and currently serves as vice president of innovation at West. Over the years, Aaron has been responsible for the health and growth of a $30 million business portfolio, led the development of natural language processing and advanced speech recognition projects, planned and implemented new organizational initiatives and is the primary liaison for multiple Fortune 50 clients. His continual leadership and success in an ever-accelerating business environment has firmly established him as both a leader throughout the West organization and across the industry.

Leave a Reply

Your email address will not be published. Required fields are marked *