Questions will ask the customer for information when appropriate. Thankful will
save the customer’s answer in the provided memory.
Thankful is smart enough to know when the information it’s about to ask has
already been provided. In these cases, Thankful will automatically skip this
step and not ask the user the question. To override this default behavior,
which is rarely needed, you may check the “Always Ask” checkbox.
When Thankful asks a question of the user, it will set the ticket in the
helpdesk to Pending automatically. When the user writes back in, the helpdesk
will reopen the ticket, at which point Thankful will re-engage and continue the
flow as appropriate.
To help Thankful understand what it should look for, you’ll set the Answer Type
for the question step. The following answer types are available:
- Personal: Extract information directly from a message.
_ Address: Thankful will detect an address and automatically build a
structured representation of it for use in the flow. This saves a
standardized address to the provided memory.
_ Email: Thankful will detect emails in the response and save an array
to the provided memory.
_ Order Number: Thankful will detect order numbers in the response and
save an array to the provided memory. Thankful will automatically prepend a
UK-1024) or append a suffix according to the
_ Phone Number: Returns single phone number. * Person: Returns a single person’s name.
- Nuance: All Nuance answer types use a classifier. They all return arrays.
They can return either  or [“None”] as well as any combo of classes. The
possible returned values expand often as training data is added. For a complete
list, please contact Thankful.
_ Cancel Order Reason
_ Cancel Subscription Reason
_ Contact Reason: Return the reason for contacting from the your
primary, deep learning classifier.
_ Custom Model
_ Return Order Reason
_ Sentiment: May return
Neutral * Track Order Reason
_ Datetime: returns a date in RFC3339 format
_ Freeform Text: Returns the user’s full reply, excluding any prefix
and suffix to the message, such as content beneath the email footer.
_ Number: Returns a single number, in floating point form
_ Option: Extracts keywords. Use comma separated values. It will match
on lowercase forms but the keyword set to the memory will be set to the form
you’ve used in the Options step, capitals included. If you want to add buttons
for chat, surround the option with square brackes, such as
[Yes] would create
the “Yes” button.
_ Product: Extract a product in structured form with Name, Color, Size,
Company and Quantity.
_ Regex: Search using a regular expression and return the first result. * Yes/No: Search for common yes/no responses. Where yes/no responses
aren’t likely, consider a custom model which detects positive or negative
replies to a question.