Collecting information from your customers

Always ask

If your action asks for the same type of data in more than one step, use the Always ask for this information setting to prevent the assistant from making incorrect assumptions. For example, you might have an action in which one step asks for a hotel check-in date and another step asks for the check-out date. If you skip asking, the assistant can mistake the check-in date for the check-out date.

To require that a step is always used in the conversation with a customer:

1. In the customer response, click the Settings icon to open Customer response settings.

   

2. Choose Always ask for this information, regardless of previous messages.

3. Click Apply.

Never ask

There might be some situations where you need a step to never ask a question because you anticipate redundant questions in the conversation.

To set that a step is never asked in the conversation with a customer:

1. In the customer response, click the Settings icon to open Customer response settings.

2. Choose Never ask. Collect information from previous messages.

3. Click Apply.

Options

An options response presents customers with a list of choices to select from. How these options are presented depends upon how your customers connect to the assistant. In the web chat integration, the options are shown as clickable buttons (for 4 or fewer options) or as a drop-down list (for more than 4 options).

There are two ways to create the list:

• Enter a list of options and synonyms
• Generate a dynamic list from a variable

Confirmation

A confirmation response presents customers with the choices of either Yes or No as clickable buttons. Use this response type when the customer's response must be either Yes or No.

The following customer responses are recognized as Yes:

• yeah
• yea
• yup
• sure
• positive

The following customer responses are recognized as No:

• not
• nope
• nay
• negative

The default setting for Confirmation is Always ask for this information. For more information, see Skipping steps, always asking steps, or never asking steps.

Regex

A regex response collects a text string that matches a pattern that is expressed as a regular expression. Use this response to capture a value that must conform to a particular pattern or format, such as an email address or telephone number.

You can specify multiple regular expressions for a single response. For example, you might define multiple regex patterns that match part numbers from different vendors that use different formats. Input text for the response is recognized if it matches any of the regex patterns you specify.

To add a regex response:

1. In Define customer response, click Regex.

2. Click Edit response.

3. In the Regular expression field, select one of the predefined regular expressions, or click Define custom regular expression to write your own.

   To use a predefined regular expression, select one of the following:

   • Email: An internet email address (for example, user@example.com).
   • Phone number: A ten-digit US phone number (for example, 800-555-1212 or (800) 555-1212).
   • URL: A correctly formatted URL for an online resource, optionally including the protocol (for example, example.com or https://example.org/index.html).

   For examples of other common patterns, see Example regex patterns.

   For more information on regular expression syntax, see Syntax.

   Only English characters can be used in a regular expression. If you need to use other characters in a regular expression, you must represent those characters in Unicode.

   watsonx Assistant uses the Google RE2 regular expression library to match regular expressions at run time. Regular expression syntax can vary between implementations, so ensure any regex patterns that you write conform to the RE2 syntax.

4. If you want to specify multiple regex patterns for the response, click Add regular expression to add another field in which you can select or define another regular expression.

   When you specify more than one regular expression, a Name field is displayed for each one. Use this field to give each regex pattern a unique name. You can use this name in subsequent step conditions to identify which regex pattern was matched.

   

5. Test your regular expression by typing example input in the Test field. If any text within your input matches the regex patterns you specified, the matching text is listed in the Assistant recognizes: field.

   

   The Test feature in the step editor uses a browser-based regex engine to find matches in your test input. At run time, the assistant uses a different regex engine that might have different results, especially with complex patterns. Before deploying your assistant in production, always use the assistant preview to test any step that uses a regex response.

6. Click Apply when you're finished editing.

7. By default, customers can't change topics when an utterance matches the pattern in a regex response. If you want a customer to be able to digress with a regex response, click the Settings icon, then enable the toggle Allow customer to change topics before evaluating a regex response. This gives the customer a chance to change topics before the assistant checks the regex pattern. For more information, see Allowing your customers to change the topic of the conversation.

You can save your configured regex response for reuse in other steps. For more information, see Saving and reusing customer responses.

Number

A number response collects a single numeric value.

The customer can specify the number value in either numerals (100) or words (one hundred). Negative and decimal values are recognized.

Date

A date response collects a specific calendar date or a range of dates. The assistant can recognize dates that are expressed in various formats. Valid examples include:

• Today
• Friday
• Now
• 10/30/2020
• October 30th, 2020
• October 30th

Time

A time response collects a single time or a range of times. The assistant can recognize times that are expressed in various formats. Valid examples include:

• 12:45PM
• 10:30
• 6am
• Now
• at 10
• from 5pm
• 4 o'clock
• half past 4

Currency

A currency response collects a currency value, including the amount and the unit. The assistant can recognize currency values that are expressed in various formats. Valid examples include:

• $10.00
• 20 cents
• five dollars
• 500 yen

Percent

A percent response collects a fractional value that is expressed as a percentage. The assistant can recognize a percentage that is written by using either the percent symbol (%) or the word percent). Valid examples include:

• 15%
• 10.5 percent

Free text

A free text response collects any arbitrary text string. Use this response for capturing any text, such as a name or address, or special instructions to be passed along. Valid examples include:

• 123 Main St.
• John Q. Smith
• Please add extra sauce

The free text response type has these default settings:

• Customer response collection behavior is Always ask for this information and can't be modified. For more information, see Skipping steps, always asking steps, or never asking steps.
• Change conversation topic is disabled. If you want a customer to be able to digress and change topics while entering a free text answer, enable the toggle Allow customer to change topics during a free text response. For more information, see Allowing your customers to change the topic of the conversation.

You can enable watsonx.ai in your assistants to intelligently recognize the text strings in a free text response with multiple action variables. For more information about using watsonx.ai to gather information, see Information-gathering.

Creating a saved customer response

To create a saved customer response:

1. In Actions, click Saved responses.

2. Click New saved response.

3. In the Name field, specify a descriptive name for the saved customer response configuration.

4. Configure the details of the response as required. A saved response can be created only as an options response type. For more information about this response type, see Customer response types.

5. Click Save. The saved customer response now appears on the Saved responses page.

6. In the Type of response field, choose Options or Regex.

You can also edit or delete any existing saved customer response. Keep in mind that any changes that you make apply to all instances of the customer response in any step that uses it. If you delete a saved customer response, any steps that use that response become invalid and must be corrected to use a different response type.

For the options and regex customer response types, you can also create a saved customer response based on the customer response configuration within a step. If you configure a customer response in a step, click the Save response for reuse icon and specify a descriptive name for the saved customer response. (This isn't available if you use the dynamic setting for an options response.)

Uploading saved customer responses

If you have many saved customer responses, you can upload them from a comma-separated value (CSV) file than to define them one by one. If you are migrating entities from the classic experience to saved customer responses in watsonx Assistant, see Migrating intents and entities.

1. Collect the saved customer responses into a CSV file. Save the CSV file with UTF-8 encoding and no byte order mark (BOM).

   The required format for each line in the file is as follows:

   <savedResponse>,<value>,<synonyms>
   

   Where <savedResponse> is the name of a saved customer response, <value> is a value for the saved customer response, and <synonyms> is a comma-separated list of synonyms for that value. For example:

   genres,science fiction,sci-fi,SF
   genres,historical fiction,HF
   genres,young adult,YA
   genres,autobiography
   genres,biography
   genres,fantasy
   locations,Adams Street
   locations,Central
   locations,South End
   

   Uploading a CSV file also supports patterns. Any string that is wrapped with / is considered a pattern, as opposed to a synonym. For example:

   ContactInfo,localPhone,/(\d{3})-(\d{4})/
   ContactInfo,fullUSphone,/(\d{3})-(\d{3})-(\d{4})/
   ContactInfo,internationalPhone,/^(\(?\+?[0-9]*\)?)?[0-9_\- \(\)]*$/
   ContactInfo,email,/\b[A-Za-z0-9._%+-]+@[A-Za-z0-9.-]+\.[A-Za-z]{2,}\b/
   ContactInfo,website,/(https?:\/\/)?([\da-z\.-]+)\.([a-z\.]{2,6})([\/\w \.-]*)*\/?$/
   

2. Go to the Saved responses page.

3. Click the Upload icon .

4. Select a file from your computer. The maximum CSV file size is 10 MB. If your CSV file is larger, consider splitting it into multiple files and uploading them separately.

   The file is validated and uploaded, and the system trains itself on the new data.

Using saved customer responses in steps

After you save a customer response, it becomes available as a response type for any step. To use a previously saved customer response in a step:

1. In the step editor, click Define customer response.

2. In the list of customer response types, click Saved to see the available saved customer responses.

   

3. Click the saved response that you want to use.

To remove a saved response from a step, click the Delete icon. Removing a saved response from a step affects only the step that you are editing. It does not delete the saved response or remove it from any other steps.

To edit a saved response from a step, click Edit response. Keep in mind that if you edit a saved response, your changes affect all steps that use the response. If you want to edit the response only for the step you are editing, click the Unlink from saved response icon. After you unlink a response, any edits you make affect only the step you are editing; they do not affect any other steps, nor are they applied to the saved response.

After you unlink a response, you cannot relink it. If you want to return to the saved response without your edits, delete the response, and then read the original saved response. If you want to make your edited version of the response available for reuse, save it as a new saved response.