A bot is an asset in your conversational strategy, but it is also a project to carry through at launch and throughout its life. This article details the right questions to ask yourself (and answer!) to approach it in the best possible way.

1. What will your bot be used for?

Determine the purpose of your bot. There is no shortage of examples and application possibilities! Increase your sales, relieve your teams, answer visitors at any time of day? It's up to you. Ideally, this purpose will also take account of your brand strategy as a whole and of your technical and human resources.
Determine the monitoring indicators that will help you steer this project: CSAT and NPS are important indicators for a large number of iAdvize users. However, if your bot's purpose is to help reduce the number of calls or emails, you may want to choose a productivity indicator instead, such as a lower email count or resolution time.

2. What are your current conversations about?

Creating a bot is a great time to take stock of your activity on iAdvize, or of the activity of the department concerned by its deployment (customer service or pre-sales):

Conversation content: what are the main topics raised by your visitors? Can your agents answer them easily with canned responses (in this case, maybe a bot could do the same thing), or do they spend time detailing your products or answering complex questions?
Your satisfaction scores: what are your CSAT scores and NPS? If you use the filters in your Conversation report, which conversations get the highest scores, and conversely, which get the lowest? How much do these performances matter to you and your team?

3. How busy are your respondents?

What are their average first response time and average handling time?
How much demand are your agents under: are they close to saturation, or on the contrary very available?
Are they entirely dedicated to iAdvize, or do they also handle calls and emails?
What are their working hours?
Are skills evenly distributed across the team, or does everyone have their own specialty?

All of this will probably affect how you want your bot to work: fully autonomously, or rather alongside your respondents.

4. Can your bot rely on an API?

The subject is technical, but it can be of great service to your visitors:

Do you have an API whose content could be useful to your visitors?
If so, is it an external API (such as delivery tracking), or one that you own?
Do you have all its documentation and, if necessary, access and a key?

5. Who will be your bot master?

Even if your iAdvize team contains multiple people, ideally you should designate one person in charge of the bot.
This person will have an accurate view of the steps in its scenario, of how it is designed, and of its history. This will improve your efficiency!
Also consider the workload the bot may represent in their schedule, as with any other project! Depending on its complexity, allow at least half a day to a full day per week for one month for the launch.

6. Multiple choice or intent detection?

A bot can rely on multiple choice questions only ("Multiple choice" card), or also use the "Intent detection" card, which understands your visitors' free-text answers (Natural Language Understanding, or NLU). This card only appears if at least one intent exists in the bot's language.

We recommend keeping it simple with multiple choice questions if:

It is your very first bot project
Your release deadlines are very short, or you lack resources and time (intent detection takes more time to set up and maintain: creating and training intents)
The bot will have simple tasks and goals,
And will handle few topics (multiple choice trees are more rigid than intent detection)

You can add intent detection later if you wish. It has other advantages:

It lets you create more varied scenarios covering many topics
It offers a smoother experience to your visitors, who ask their questions in their own words
For this reason, it is also an excellent source of customer insights: by analyzing your visitors' wording, it shows you phrasings or needs you may not have spotted!
But it also requires more preparation to configure the artificial intelligence before launch, then more time to maintain it throughout the project.

7. Your bot in iAdvize

A bot is both a respondent and an extra cog in the overall mechanics of iAdvize. So it's good to think about:

Your engagement rules. For example, if your bot is used for pre-qualification and self-service, would you like to assign all your targeting rules to it? Would you like to test it on an engagement campaign before extending it to your entire site? Or would you like to adjust its hours according to those of your team?
Your routing: will your bot make a lot of transfers, or do you hope it will handle many conversations independently? Will it have to “collaborate” with a team of respondents who all have the same skills, or who each have specific expertise? All these aspects can affect its operation and are discussed in detail in our article on routing.
The number of bots you have: if it seems appropriate, you can also divide the tasks assigned to your bot between two bots to optimize its work (this point is explained in our article).

8. Get creating!

Now you have your roadmap in mind, it's time to get started!

if your bot uses intent detection, first activate its language in Shopping Assistant > Intents ("Activate" button), then create your intents
create the steps in your scenario
and write your cards in the bot builder

We wish you every success in creating your bot