Disable user input in Now assist panel

[msg2anil]

Hi All, 

I have a requirement to get consent from the user while he starts to use now assist panel. I can give a good message and choice to agree or cancel. but instead of choosing from given options if user starts to type it leads to topic discovery and Now assist panel will start to give him answers. 

How can i restrict user to type in now assist panel (Disable user input) - i want the user only to select from the choice list. 

[Abhishek Pal]

Hi @msg2anil ,

There is no supported OOB configuration to physically disable or hide the text input box for only one step in the Now Assist panel.

However, you can prevent typed text from triggering Topic Discovery while the consent question is active.

Recommended configuration:

1. Open the consent topic in Virtual Agent Designer.

2. Use a Static Choice or Boolean user input control instead of a normal text input.

For example:

Display name: Agree
Value: agree

Display name: Cancel
Value: cancel

3. Open the Advanced configuration of the choice node.

4. Under Conversation switching, disable:

Turn on to let users change the subject

When this option is enabled, text entered by the user is passed to NLU prediction and can trigger another topic or Now Assist response.

Disabling it prevents the user from switching topics while the consent choice is waiting for a response.

5. Branch based on the selected value:

agree
-> Store the consent
-> Continue to the required Now Assist functionality

cancel
-> Display a cancellation message
-> End the conversation

Important limitation:

The composer may still remain visible, but arbitrary text should no longer trigger Topic Discovery from that choice node. The supported solution controls conversation behavior; it does not disable the textbox at the UI level.

Do not use DOM manipulation, CSS, or clone the OOB Now Assist panel component to hide the composer. That would be unsupported and upgrade-sensitive.

If this is a mandatory legal or compliance consent, the stronger design is to collect and store the consent before enabling access to Now Assist.

Store information such as:

- User
- Consent status
- Consent timestamp
- Consent-text version
- Withdrawal status

Then allow access to the Now Assist panel or its skills only after valid consent exists. Do not depend only on an in-memory Virtual Agent topic variable for compliance evidence.

Recommended approach:

Consent choice in the conversation
-> Static Choice or Boolean control
-> Conversation switching disabled

Strict mandatory consent
-> Persist consent before providing Now Assist access

Official references:

Static Choice user input control:
https://www.servicenow.com/docs/r/conversational-interfaces/virtual-agent/va-static-choicelist.html

Now Assist in Virtual Agent supported controls:
https://www.servicenow.com/docs/r/conversational-interfaces/now-assist-in-virtual-agent/now-assist-i...

Hope this helps!

If this response helped, please mark it as Helpful.
If it resolves your issue, please Accept it as Solution.

Kind Regards,
Abhishek Pal

[msg2anil]

Hi @Abhishek Pal -  Thanks for the solution, but unfortunately I'm unable to see the Advanced configuration to disable topic switching.  below screenshot is what i see and we are in Australia upgrade. 

[Abhishek Pal · Accepted solution]

Hi @msg2anil ,

Thanks for sharing the screenshot. You are correct, and I need to correct one point from my earlier response.

Your screenshot shows that this is an LLM-based topic in the Australia release. That is why you do not see:

Conversation switching
-> Turn on to let users change the subject

ServiceNow documents that this option is available only for NLU topic discovery. It is not exposed for an LLM Static Choice node.

The options you see are correct for an LLM Static Choice:

- Enforce user prompt
- Validation
- Additional instructions for LLM
- Hide or skip this node

Important:

Enforce user prompt does NOT disable typing.

It only replaces the prompt generated by the LLM with the message that you configure.

Similarly, Validation can validate the value captured by the input node, but it does not physically disable the Now Assist composer or provide a supported per-node switch to turn off LLM mid-topic discovery.

Therefore, for an LLM topic in the Australia Now Assist panel, there is currently no supported OOB setting to disable/hide the text input box only while this Static Choice is displayed.

For your consent requirement, I would use one of these approaches.

Recommended approach for strict consent:

Use an NLU/keyword Virtual Agent topic for the consent step instead of an LLM topic.

Configure:

Static Choice

Start Using AI
Value: accept

Cancel
Value: reject

Then under:

Advanced
-> Conversation switching

keep:

Turn on to let users change the subject = OFF

This prevents free-text input at that node from being used for NLU topic switching.

After:

accept
-> Persist the consent
-> Continue/start the required Now Assist experience

reject
-> End the conversation or prevent access to the AI functionality

The text composer can still be visible, but the user cannot use that input to switch to another NLU topic while the consent node is active.

If you must keep the current LLM topic:

Use the options visible in your screenshot:

1. Enable Enforce user prompt.

Use a message such as:

Please review the consent notice and select either "Start Using AI" or "Cancel" below. A selection is required before continuing.

2. Under Additional instructions for LLM, add:

Do not proceed with this workflow unless the user explicitly selects one of the provided consent choices. Do not interpret unrelated free-text responses as consent.

3. Use Validation to ensure that the resulting value is only:

accept

or:

reject

4. Persist the consent server-side before enabling the protected AI functionality.

However, I would not treat Additional instructions for LLM as a security/compliance enforcement mechanism. LLM instructions are behavioral guidance, not access control.

If this is mandatory legal/privacy consent, the stronger architecture is:

User opens Now Assist
-> Check persistent consent record
-> No valid consent
-> Present controlled consent experience
-> User explicitly accepts
-> Save user + consent version + timestamp
-> Allow Now Assist functionality

If consent is rejected:
-> Do not invoke the AI capability

This ensures that even if a user types another request, the actual AI functionality remains blocked until valid consent exists.

I would not recommend:

- DOM manipulation to disable the Now Assist textbox
- CSS to hide the composer
- Modifying or cloning the OOB Now Assist panel
- Relying only on LLM instructions for compliance enforcement

So for your Australia screenshot:

You are not missing any configuration.

The "Conversation switching" option is absent because this is an LLM topic. My previous instruction applied to NLU topics.

If your requirement is strict consent before any AI interaction, I recommend either using an NLU consent topic with Conversation switching disabled, or enforcing consent outside/before the LLM experience using a persisted consent check.

Official references:

Static Choice - Australia:
https://www.servicenow.com/docs/r/conversational-interfaces/virtual-agent/va-static-choicelist.html

Mid-topic switching in Now Assist in Virtual Agent:
https://www.servicenow.com/docs/r/conversational-interfaces/now-assist-in-virtual-agent/intent-switc...

Assistant Designer user input controls:
https://www.servicenow.com/docs/r/conversational-interfaces/virtual-agent/va-user-inputs.html

Hope this helps!

If this response helped, please mark it as Helpful.
If it resolves your issue, please Accept it as Solution.

Kind Regards,
Abhishek Pal