Making your AI business case too big to fail (Part 1) |

Our Blog

where we write about the things we love

13

Aug

Making your AI business case too big to fail (Part 1)

You’ve bribed the boss, sweet-talked the CFO and whipped the tech-heads into a frenzy. Congratulations! You got your AI project off the ground. Now what do you do?

This is where the real fun begins: designing your AI services.

Designing your bot
Photo by rawpixel

How hard can it be? It’s just a person asking a computer a load of questions, right? Remember that you’re creating an experience. Taking a little time to consider what you want that experience to be will set your AI up for success and in this two-part post we highlight some fundamentals of good design.

We’ve used a chat bot in our scenario but the principles can be used for any AI that has direct contact with people.

Preparation, preparation, preparation

Before you start designing there’s a bit of prep to get done.

Where should the bot live?

Ironically the first place to start is also the latest buzz-phrase appearing on every business’ call sheet: user experience (UX). Commonly referred to as the ‘colouring-in department’ UX ensures your bot meets the needs of the people using it and (equally important) adds value to your business.

People think focus means saying yes to the thing you’ve got to focus on. But that’s not what it means at all. It means saying no to the hundred other good ideas that there are.
Steve Jobs

Your team will have a lot of ideas on what a bot could do and what channel it needs to live in. To give you confidence you’re putting it in the right context, you need to understand what your user’s current experience actually is. Don’t second guess this.

One way to understand this is to create a user journey map. A journey map visualises the user’s processes, needs and perceptions and helps you identify where the bot can add value. Value could be smoothing the user’s journey to a desired goal or improving business operations (preferably both). They’re quick to do and can steer decision making across multiple aspects of your digital ecosystem, not just the bot. Plus they are a great way to focus your team on people’s needs.

Creating the bot journey map

Journey maps also minimise the risk of the bot having any negative effects on the business. What’s good for customers may not be good for your bottom line and journey maps enable stakeholders to identify this early on.

Essentially to create one you need to choose one of your design personas (or ideally involve real users), take a scenario they typically face, and then break that scenario down into the individual tasks they do to complete their goal.

For each mini-task it becomes very clear very quickly where their experience is lacking and where a bot could add value.

Each journey map will be different depending on your design persona and the scenario, but in general a good place for your bot would be:

  • at pain points in the user journey
  • where rapid response or transactional conversations are appropriate
  • outside standard office hours when human support is costly or resource heavy

Define the scope of the bot

This step in the design process is essential. Where does interaction with the bot start and end? Is there a point where handover to a human is appropriate? What systems or data does the bot need to tap into to help the user complete their goal?

A few things will steer bot scope. The task you are attempting to solve, the audience using it, and the technology the bot taps into to complete tasks. The key is to consider the first two together.

The task and the person

Your journey map will help you identify tasks that either add value for the user or streamline business processes. When assessing value to the user be considerate of what value actually is for them and how it is currently gained. For instance, if speed of service equals value and the existing mechanism of achieving this is already pretty good, then your bot has to do something pretty special to get people to switch from their usual behaviour (unless micro-improvements are critical).

Before you add this task to your bot’s scope consider the person expected to do it. Will they be comfortable doing it through a bot? If bots are already part of their daily digital digest or they are the right demographic, your bot may be natural extension of their current behaviour. It’s not impossible to get non-tech users to take to bots but if this is your first time, pick the low hanging fruit.

One final plot twist is context. Where are you expecting people to engage and on what device? This could change everything. Trying to find out which platform your train leaves from in a busy terminal is different from browsing for a new pair of jeans at your desk.

Technical integration

There are plenty of bot frameworks that will allow you to create a basic bot in a matter of hours and launch on your channel of choice. Tools like QnA Maker and LUIS from Microsoft enable you to quickly give your bot a heap of prior knowledge and the ability to learn and understand us.

But what about all that data about your customers you’ve been collecting? If integrated correctly, giving your bots access to data sources and applications can increase their value enormously. In fact cross-system integration can be what distinguishes you in a busy market.

There is a consequence to this though. The more systems and tools you integrate with the more costly and complex the build will be. If you’re a bot virgin then focus on doing one or two things really well rather than integrating every system you have in one go. There are several reasons for this:

  • having a large number of features will complicate the build and drive up costs and timings
  • bigger projects have more points of failure
  • until people engage with the bot you don’t know how they will interact with it or what they will try and use it for
  • confidence in the bot’s ability to perform a single task will encourage people to use it again when a new feature is released

A key strength of bots is their ability to learn. Rather than guess what the next feature should be and gamble precious resource on hunches, get the bot to tell you, through real data. Securing future funding for small incremental enhancements is easier if you can prove business value with each release.

Also, bot framework providers like Microsoft are constantly enhancing and expanding their bot toolkits. Make sure your technology partner knows what new out-of-the-box (OOTB) features are coming up. It could stop you spending budget on capabilities that are just around the corner.

Choose a bot platform

Speaking of bot frameworks your final prep is to decide which one to use. There are plenty out there but the runaway leaders to date are the Microsoft Bot Framework, Botkit, Chatterbot, and Chatscript. All are easy to use and are getting better all the time.

Our chosen platform is the Microsoft Bot Framework. There are several reasons for this:

  • you have all the tools you need in one place (NLP, Q&A content, remembering conversations, receiving requests and triggering responses). This means you can focus on design and the experience rather than connecting all the pieces of the bot brain together
  • the Microsoft Bot Framework is a platform itself which other providers often support in addition to their own bot services
  • it supports many programming languages and platforms which makes it ideal for deployment across multiple channels without the need to customise a bot for each (e.g. Facebook, WhatsApp, Skype, Messenger, Slack)
  • it’s got huge backing at Microsoft. Always an advantage to know your bot framework provider is going to be around for a while and that new tools are going to keep rolling off the production line :)

Ok, so that’s the prep done. Part two of this post covers your bot design, its personality and the experience you want to give your users.

Posted by: Adam Ford, Practice Lead, Digital & Experience Design | 13 August 2018

Tags: customer experience, bot, AI


Blog archive

Stay up to date with all insights from the Intergen blog