A common question asked by businesses looking to implement a Conversational AI or chatbot solution like Iris in their organization is, ‘where do I begin?’.
The general benefits of rolling out an AI in the business across text and voice channels are generally known; the need to solve a problem, desire to be more efficient or offer a better customer experience.
But how to make this happen? Ensure it’s successful and does what you require from the beginning?
Keeping in mind some projects are bigger than others with many moving parts. Here are five foundational questions to ask before you begin implementing a chatbot of your own.
1. Who is the Audience?
Who is going to be using what you build is the first and possibly the most important question of all. It will guide structure of the dialogue, how you intend to achieve your objectives, the persona of the chatbot and the underlying capabilities the solution will need.
As an example, the way you speak to potential customers can be dramatically different to existing customers. So if you’re looking to speak to both you need a way of differentiating them, how are you going to do this? Now we’re now getting into the capabilities.
At the very least, defining your audience will influence the personality or lack there of in the final solution.
2. What are the objectives?
Having a clear understanding of your business objectives from the beginning of the project can help maintain alignment and focus to ensure the chatbot is successful.
The crucial step here is to define your business objectives as well as your customers objectives when they use the chatbot to interact with you. What do your customers want and how can you get it to them in the most efficient manner?
3. What persona will you adopt?
There’s no right or wrong when it comes to the personality your bot will adopt, the advice here is commitment. Select whether you will give the chatbot a personality (and even a name), speak as if the business has a voice of its own or be completely neutral.
Get this slightly wrong and it can be challenging to keep your users engaged, get it right and you’ll create an immersive experience.
The personality of the bot will guide the tone, voice, language and cadence. This is more art than science and a hard balancing act to get right. But, don’t forget you should always be looking to test, learn and iterate. The personality can evolve over time.
4. What are the key capabilities?
Once you have defined the audience, objectives and personality, the technical capabilities you need to deliver on them come into play.
Generally we go for a keep it simple for the first time out rule. Ask yourself: what messaging experience can I create without going too technical? Build a simple first version of your bot then test it with the target audience, are they receptive?
Limiting dependencies we sometimes call it, a Conversational Marketing Lead Gen bot is a good example, sounds easy right? Say you want to capture an email address, but to make this useful you also need a name too. How do you validate the email? Store the information once you’ve capture it? Then how are you going to use all these emails? How long will you store them? Etc. See how a simple ‘let’s capture an email’ can expand the capabilities quite quickly?
Start simple, do the least amount of work and potentially use a roadmap to rollout in phases.
5. How do I measure, test and learn?
While objectives are the things you want to achieve, goals for the chatbot should be measurable and timely. I would like to see ‘X’ number of people use it over ‘Y’ period of time, with ‘Z’ number of people making all the way through the experience.
Make a plan to review these goals frequently when you launch, slowly scaling back the reviews once you have been live for a while. These check-ins will give you key milestones to test different outcomes and review their progress.
As we mentioned above, your bot should never be set-and-forget but an always evolving and improving experience over time for customers.
Each of these five questions can be explored further and we always do when we work with our customers. Hope they have been useful in giving you a starting point to build a bot.