Developing chatbots in the enterprise: both for the customer and employee

The advancement and development of new technologies surrounding natural language understanding, processing and construction is growing up rapidly.

This makes the chatbot space “very exciting,” says Anshuman Singh, head of digital for Mindtree in Europe.

On top of this, given the proliferation of a lot of out-of-the-box tools for chatbot building, it’s become easier to develop and produce a chatbot; although organisations will quickly learn the possibilities are finite and limited.

Indeed, building a bot is significantly more complex when you’re building one for the enterprise.

Mindtree plays at this intersection of making chatbot experiences work at an enterprise level

When it comes to enterprises, the chatbot is a different beast entirely to a bot developed for say, ordering a pizza. There are a number of extra factors that need to be considered.

First it’s important to understand that “bots work on a basic notion of mapping utterances to intents,” explains Singh. By that he means if you have bot for a restaurant, for example, and it picks up on the word dinner, the bot will quickly identity a next best action of presenting the customer with a menu. “Now most bot technologies only support a finite amount of intention utterances, somewhere in the range of 40,” says Singh.

When you get into more complex (enterprise) scenarios, these intention utterances increase dramatically; and it becomes a much more challenging problem to solve.

A bot interface in an enterprise requires “an orchestration of multiple mother/super boards to make it work,” says Singh.

HR, chatbots and robotic process automation: key for digital transformation

Organisations must modernise their HR function to succeed in digital transformation efforts. How? Success relies on software automation — chatbots and robotic process automation — and continued human involvement. Read here

The enterprise chatbot challenge(s)

Then the classic challenge of maintaining context emerges: for example, is the bot maintaining that thread of conversation, so as not to ask the same question again? Or regarding authorisation: for example, an employee can ask the bot about their salary, but should they be allowed to know their boss’s salary — the answer, of course, is no — so how can the bot identify who is talking?

When you get into larger enterprises, you have challenges around languages and nuances of languages, which becomes interesting because most bots are primarily developed with English as their main language.

There are customers who are fairly convinced about the value bots will bring in, but not open to taking a risk or executing the use case that is consumer-facing. They often start internally, with the aim of cost saving and better employee satisfaction

Challenges then emerge on the experience side. In the past few years, most enterprises have invested heavily in training user experience designers.

“When you’re designing for a bot, some of the fundamental design principles still apply,” says Singh, “and suddenly you are in the remit of scriptwriting or copywriting, rather than screen designing.”

The two are completely different challenges; how do you give a bot a personality? How do you ask the right kind of questions? How do you design and script the entire interaction?

What is a chatbot? How to make a chatbot? Why chatbots fail? All your questions answered

What is a chatbot? How do you make a chatbot? Why chatbots fail? In this article, we answer all these questions with the help of Howard Pull, strategic development director at MullenLowe Profero. Read here

Integrating the enterprise chatbot

To overcome the above challenges, organisations should follow a number of processes.

Before integrating a bot into enterprise operations, businesses must identify use cases and choose whether their bot will be customer or employee facing.

That’s the first question that needs to be answered: “should the first bot an enterprise develops be customer-facing or should it be employee-facing,” asks Singh. “What is your risk appetite?”

The second focuses on product management or placement, and this depends on what type kind of customers a business deals with. For example, many of Mindtree’s customers may choose to automate some of their internal service desks, whether it’s the IT or HR service desk to handle internal queries.

“It’s important to pick up use cases that will deliver the maximum amount of business benefit, either in terms of taking a transaction cost away or improving convergence” — Singh

So, a quick roundup: the first step in integrating a chatbot is deciding whether it will be internal or external, the second is to pick up specific use cases and identify what kind of audience you are catering to.

The third step surrounds architecting those capabilities. At the moment, “there are different platforms which offer different kinds of interaction paradigms,” says Singh. “So, the Facebook messenger bot works differently to the Skype bot and so on and so forth. Some of them support a decision tree and some of them are more open-ended — these are true NLP bots and they’re able to understand what the customer is asking as opposed to just presenting a finite set of choices.”

The history of the chatbot: Where it was and where it’s going

Chatbots are not new. The software has been around for decades, and now its application is beginning to impact the enterprise — both internally and outward facing. Read here

Enterprise chatbot examples

Mindtree have developed a voice-based bot that helps enterprises query financial performance, “which in turn hooks back to your standard management information system [MIS] and gets you an answer,” says Singh. This type of bot, for example, helps identify pipeline funnel movement and the revenue for this year. “Senior execs,” continues Singh, “can ask this question to Alexa, which is hooked up to the MIS system.

“That dramatically improves the consumption of information in any organisation because there are so many systems, so many spreadsheets that people are either looking at the wrong spreadsheet or they can’t be bothered to find that link which will give them the answer. So voice-based interaction has helped one of our customers to get the right kind of answers.”

Mindtree itself has an intranet bot called MACI, which handles most of the common transactions of the company’s 20,000 employees

The second example Singh provides comes from a customer who was spending millions in outsourcing their IT and HR service desks. “The trouble,” explains Singh, “is that most of these existing vendors are incumbents and have so much invested in their current service operation that they will not bring in such ideas to the customer.”

To combat this, Mindtree picked the top 10 use cases, both on the internal IT and HR side, and implemented them via a bot on the Skype for business — which is your standard messenger tool for the enterprise. Users could then ask questions and get queries answered. This includes everything from; “applying for leave, getting back my leave entitlement, finding out when the next public holiday, approving a purchase order or getting the latest status on an order,” says Singh.

“We are able to automate these and therefore cut down significant costs in terms of what they were dishing out in terms of support desks.”

There are a few other customers of Mindtree’s who are looking at using bots for simple transactions; in terms of a running order status, reminding someone about an upcoming payment or nudging a customer if they’ve left something in their shopping basket — these are the bots of the finite variety that are being used to execute customer-facing use cases.

Avatar photo

Nick Ismail

Nick Ismail is a former editor for Information Age (from 2018 to 2022) before moving on to become Global Head of Brand Journalism at HCLTech. He has a particular interest in smart technologies, AI and...

Related Topics