Take time to identify and plan a high level roadmap of key scenarios you expect your chatbot to be used in. Don't try to implement too much with the first iteration of your conversational AI application.
Start with a narrow set of scenarios to begin with. This will make it quicker for your to release updates. Iterate on each update and focus on the niche areas that drive the biggest impact.
Use the following 4 steps to hell you identify key scenarios:
Start by exploring existing datasets, knowledge and software within your current IT portfolio. Do you provide support or helpdesk capabilities? What are the high volume or traffic areas on your website? What kind of data does your business, customers or users regularly interact with?
Consider if surfacing any of the above in a conversational AI channel unlock hidden business value or make things better for your customers or users.
After completing the analysis of existing IT portfolio, shortlist what you believe are the high-value processes or datasets. Things to consider when arriving at this shortlist include:
Consider how your conversational AI solution might address these. During this step, niche down on the processes, use cases and user journeys your solution will need to handle.
The output from this step will form the initial scenarios that your conversational AI application will improve.
Use the datasets and high-value processes you have discovered to prototype the key scenarios your conversational will be deployed in. An example of a Key Scenario might be:
| Existing Data | Salesforce CRM, Telephone and IVR records, customer queries, Google Analytics |
|---|---|
| High Value Process | Customers regularly calling to query account data |
| Key Scenario | FAQ for Top 5 Customer Queries |
| Why this a Key Scenario | Free up agents on our help desk to handle more complex queries |
The output of this step will generate the foundations for the logic and language model that your conversational AI solution will need to implement.