IQ
Introduction
IQ is CX's Gen AI assistant service, providing Live Suggestions, Spelling, Prediction, Rewrites, Evaluation, Summarization and more.
The service spans not only the agent desktop, but also services supporting contact flow implementation.
Further features will be added to IQ with each CX release.
Service Overview
CX IQ utilizes AWS's Bedrock service to access AWS-hosted Large Language Models (LLMs) to support various Generative AI use cases.
Bedrock runs in your AWS account and you have complete control of your data. All services (for example Bedrock Agents) are deployed via CX's orchestration layer - you do not need to deploy these manually.
Features
IQ is made up of a set of independently controllable features, each enabled from the NeonNow Admin app once the Ready Checklist is complete:
- Correct - automatic correction of spelling, grammar, and punctuation, ensuring text maintains a consistent, error-free level of quality.
- Extract - condense emails, task information and more into immediate, actionable summaries.
- Refine - rule-based rewrites to keep messages clear and on-brand, including Evaluate for tone and language feedback.
In addition, IQ powers a range of always-on agent assists such as Live Suggestions, Spelling, Prediction, Rewrites, Evaluation and Summarisation.
IQ features are currently in Beta. Behaviour and configuration options may change between releases.
Requirements and LLM Model lifecycle
CX IQ orchestrates and utilizes AWS Bedrock agents. Model lifecycle is a key consideration of utilising LLMs, and this is handled by CX.
To ensure continuous support, the context of a 'current' model and a 'new' model exists. As new models become available, these are phased in and made available for enablement. When this happens, an "A new version of IQ is coming" panel is shown at the top of the IQ page, listing the new models so you can transition smoothly.
How a new model becomes available depends on its provider:
- Anthropic models require a one-time use case to be submitted in AWS Bedrock before they can be used (see the Models Enabled step below). Once that initial use case has been submitted and approved, future Anthropic models are automatically enabled - you do not need to submit a new use case each time a model is phased in.
- Other providers (for example OpenAI) do not require a use case and are enabled immediately, with no further action required.
Generative AI Usage & Pricing
AWS Bedrock pricing is documented here.
Each CX IQ operation will consume 'tokens' as a standard measure of LLM usage.
As a rough order of magnitude, to evaluate a short paragraph of text using Refine's Evaluate:
- input tokens = 400
- output tokens = 500
Total cost incurred = 400 tokens/1000 * $0.008 + 500 tokens/1000 * $0.024 = $0.0032 + $0.012 = $0.0152
Setup & Readiness
To enable IQ, log in to the NeonNow Admin app and review the Ready Checklist. This guides you through the required steps, showing how many are complete (for example 2/3 completed). Each item displays a health state of Healthy, Unhealthy or Unavailable, and provides a Fix action where attention is needed.
The three checks are as follows:
1. Service Role Permission
The existing CX service role must be extended to allow orchestration of Bedrock Agents, as well as to allow self-awareness for role version checks. It is recommended to copy the role policy as shown in the NeonNow Admin app into an inline policy of the role. An example is shown below for informational purposes only (please ensure the policy is copied from the NeonNow Admin app).
{
"Sid": "CheckOwnRole",
"Effect": "Allow",
"Action": [
"iam:GetRolePolicy",
"iam:ListRolePolicies"
],
"Resource": [
"arn:aws:iam::<AWS Account>:role/<unique-id>-client-apiExeRole"
]
},
{
"Sid": "CheckOtherCWRoles",
"Effect": "Allow",
"Action": [
"iam:ListRoles"
],
"Resource": [
"arn:aws:iam::<AWS Account>:role/"
]
},
{
"Sid": "PassIQServiceRole",
"Effect": "Allow",
"Action": [
"iam:PassRole"
],
"Resource": [
"arn:aws:iam::<AWS Account>:role/neonnow-iq-service-role"
]
}
2. Granting Bedrock Model Access
IQ relies on more than one LLM, so this check covers all of the models IQ requires - not just a single model. Hovering over the 'Models Enabled' item lists the models, each with its own state, and the check is only healthy once every required model is enabled. The same per-model view appears in the "A new version of IQ is coming" panel when newer models are being phased in.
Models are accessed through AWS Bedrock, and each one has to be enabled inside your own AWS account before IQ can use it.
The CX service role cannot enable models for you - it does not hold the AWS Marketplace permissions (aws-marketplace:Subscribe, aws-marketplace:Unsubscribe, aws-marketplace:ViewSubscriptions) that AWS requires to grant access to a foundation model.
Enabling a model is therefore a one-time action performed by a person in your AWS account who has those Marketplace permissions. Once a model has been enabled in the account, every user and role - including the CX service role - can invoke it without needing Marketplace permissions of their own.
What you need to do
AWS documents the full process in Request access to models . In short:
- Request access to (enable) each required model in your AWS account. Access is granted at the account level, so this is a one-time step per model, performed by someone with the AWS Marketplace permissions listed above.
- Make sure the prerequisites are in place: the AWS Marketplace permissions, a valid payment method on the account, and - for Anthropic models - a completed First Time Use (use case) form.
- Enabling is effectively instant. When a model is first invoked AWS subscribes the account automatically in the background (usually within a couple of minutes); for Anthropic, access is granted as soon as the use case form is submitted and approved.
The simplest way to enable a model is to open it in your AWS account: in the Amazon Bedrock console, pick the model from the model catalogue and open it in the playground. Running a single prompt there triggers the account subscription, which enables the model for everyone in the account (including IQ). See Request access to models for the details.
The NeonNow Admin app helps you here too: hover over the Models Enabled item to see the tooltip listing every model and its state, so you can tell which still need enabling. Selecting the Enable button opens each model's page in AWS, each in a separate tab, where a use case may need to be submitted.
For Anthropic models, submit the following text in the use case:
- Company Name: enter your company name
- Company website URL: enter your company website
- What industry do you operate in?: select your company's industry from the drop-down
- Who are the intended users you are building for?: select 'Internal users'
- Describe your use cases: we will be using the model to summarize, rewrite, and predict text for our contact centre application CX which integrates to Amazon Connect, and uses various AWS services including Bedrock agent.
Use cases are normally approved within 10-15 minutes.
You only need to submit a use case once. After your initial Anthropic use case is approved, future Anthropic models are enabled automatically as they are phased in. Models from other providers (for example OpenAI) are enabled immediately and do not require a use case.
3. IQ Service Role
A new IQ Service role must be created (neonnow-iq-service-role) to allow orchestration of Bedrock Agents. This is the role referenced in the main service role extension described above. If the role is not present, select Fix to open the Update Your IQ Service Role modal. Use the copy action to copy the IAM policy shown, then select Update AWS Role to launch the CloudFormation that creates the new Service Role.
Select the I acknowledge that AWS CloudFormation might create IAM resources with custom names. tick box and select Create stack. Once the role is created, review the NeonNow Admin app readiness - all steps should now show as ready. The Service Role is shown below for informational purposes only.
{
"Version": "2012-10-17",
"Statement": [
{
"Action": [
"bedrock:InvokeModel",
"bedrock:InvokeModelWithResponseStream"
],
"Resource": "arn:aws:bedrock:*::foundation-model/*",
"Effect": "Allow",
"Sid": "AmazonBedrockAgentBedrockFoundationModelPolicy"
},
{
"Effect": "Allow",
"Action": [
"bedrock:InvokeModel*",
"bedrock:CreateInferenceProfile"
],
"Resource": [
"arn:aws:bedrock:*::foundation-model/*",
"arn:aws:bedrock:*:*:inference-profile/*",
"arn:aws:bedrock:*:*:application-inference-profile/*"
]
},
{
"Effect": "Allow",
"Action": [
"bedrock:GetInferenceProfile",
"bedrock:ListInferenceProfiles",
"bedrock:DeleteInferenceProfile",
"bedrock:TagResource",
"bedrock:UntagResource",
"bedrock:ListTagsForResource"
],
"Resource": [
"arn:aws:bedrock:*:*:inference-profile/*",
"arn:aws:bedrock:*:*:application-inference-profile/*"
]
}
]
}
If the IQ Service Role details cannot be loaded, the Admin app shows a "something went wrong" message. Refresh the page and try again.
Output Locale
Once the Ready Checklist is complete, select the Output Locale. This determines the text locale (including spelling, dates, and numbers) returned from the various IQ services.
Enabling & disabling features
Each IQ feature is controlled by its own toggle on the IQ page.
When you enable any feature, a confirmation modal asks you to acknowledge that IQ-generated content may not always be accurate and that your organisation and its users are responsible for reviewing and using it appropriately. Generated content may contain errors or omissions, users remain responsible for verifying it before acting on it, and sensitive or personal data should not be entered into IQ prompts unless appropriate.
Only Refine (including Evaluate) provisions a dedicated Amazon Bedrock agent. When you enable Refine, the confirmation modal includes an additional information panel noting that IQ will create a new Amazon Bedrock agent and that any usage will generate additional charges.
Disabling a feature behaves differently depending on whether it has an agent. Refine shows a turn-off confirmation modal warning that IQ will remove the Amazon Bedrock agent previously created for it. The other features have no agent to remove, so they can be turned off directly without a confirmation.