AI Assist Configuration Options

configuring your ai assistant applies to admins setting up or editing an ai assistant section create/edit flow → step 2 configuration overview the configuration step is where you control how your ai assistant behaves during live interactions it is divided into three sections platform integration, general configuration, and advanced configuration you can expand or collapse each section using the chevron on the right of the section header your changes are not applied until you click save and next or publish you can save a draft at any time without affecting a live assistant platform integration this section connects your platform to the assistant it must be completed before the assistant can go live refer to the platform integration guide for setup steps specific to your platform general configuration general configuration controls the assistant capabilities that are visible to agents during live interactions each setting can be turned on or off independently using its toggle smart replies what it does surfaces ai generated reply suggestions to agents in real time, based on the customer's message turn this on to help agents respond faster and more consistently when disabled, no suggestions appear in the assistant panel when smart replies is enabled, two additional settings appear max words per smart reply controls the maximum length of each ai generated reply suggestion enter a number to cap reply length leave the field blank if you want no limit default 20 words tip shorter replies work better for quick acknowledgements for complex queries, a higher limit gives the ai more room to be helpful smart reply feedback 👍 👎 allows agents to rate individual smart replies as helpful or not during live conversations this feedback can be used to improve suggestion quality over time when enabled, agents see thumbs up and thumbs down buttons on each suggestion customise bad feedback reasons (optional) you can add up to three custom reasons that agents can select when they mark a reply as unhelpful — for example, "irrelevant", "wrong", or "outdated" each reason has a 15 character limit these fields are optional; if left blank, agents can still give feedback without a reason live transcript what it does shows a live, real time transcript of the call in the assistant panel while an agent is on a voice call note this setting only applies to voice interactions it has no effect on chat or messaging interactions this toggle is currently disabled for chat only setups if your platform supports voice, contact your administrator to enable it real time sentiments what it does analyses the customer's tone during a conversation and displays sentiment signals (positive, neutral, negative) in the assistant panel this helps agents gauge how a customer is feeling without having to ask, so they can adjust their approach in the moment disposition suggestions what it does at the end of a conversation, the assistant suggests a call disposition (outcome category) based on what was discussed when disposition suggestions is enabled, two additional settings appear disposition notes when enabled, the assistant also generates a short summary note to accompany the suggested disposition this gives agents a starting point for call wrap up rather than writing from scratch allow agents to edit disposition notes when enabled, agents can edit the ai generated disposition note before submitting it this is recommended if you want agents to verify accuracy or add context before the record is saved tip if you want agents to use ai generated notes exactly as written without modification, turn this off advanced configuration advanced configuration covers settings that affect data handling and language behaviour these are typically set once during initial setup and rarely need to change mask sensitive data what it does automatically redacts sensitive customer information — such as card numbers, id numbers, and personal identifiers — from transcripts and smart reply suggestions this is enabled by default and recommended for all deployments to protect customer privacy and meet data compliance requirements add rules if your business handles specific types of sensitive data beyond the standard set (full name, mobile number, card numbers, etc ), you can define custom masking rules using the + add rules option this lets you extend masking to business specific data formats important disabling this setting means sensitive data may appear in transcripts and be visible to agents only disable if you have a specific reason and understand the compliance implications language what it does determines the language in which smart replies and transcripts are generated and displayed when enabled, the assistant uses your configured language setting for all ai generated content in the panel note the languages available are determined by what your transcription engine supports accuracy may vary depending on audio quality and the specific language model in use contact support if you need a language that isn't currently available saving your configuration button what it does save and next saves your configuration and moves to the next step (deploy and publish) save as draft saves your changes without publishing if the assistant is currently live, it will be taken offline you will need to publish again to make it live remember if your assistant is currently live and you click save as draft, it will stop responding to users immediately see save as draft — what happens to a live assistant for more details