Newcastle City Council: AWS Contact Centre Services (Amazon Q and Contact Lens)
Cloud-based contact centre platform using Amazon Q and Contact Lens to support staff, transcribe calls, and provide service insights.
1. Summary
1 - Name
AWS Contact Centre Services (Amazon Q and Contact Lens)
2 - Description
Amazon Q and Amazon Contact Lens are AI-enabled features within the council’s AWS Amazon Connect contact centre service: Amazon Q supports agents with real-time prompts and approved information, while Contact Lens transcribes calls and provides interaction analytics to help improve service quality and staff training; the tools are used to support quicker, more consistent customer service, and all decisions remain with council staff.
3 - Website URL
https://aws.amazon.com/connect/ https://aws.amazon.com/q/ https://aws.amazon.com/connect/ai-agents/ https://aws.amazon.com/connect/contact-lens/ https://docs.aws.amazon.com/connect/latest/adminguide/contact-lens.html
4 - Contact email
digitaltransformation@newcastle.gov.uk
Tier 2 - Owner and Responsibility
1.1 - Organisation or department
Newcastle City Council
1.2 - Team
Customer Experience and Change
1.3 - Senior responsible owner
Assistant Director - Customer Contact, ICT and Digital Transformation Newcastle City Council
1.4 - Third party involvement
Yes
1.4.1 - Third party
AWS and PwC
1.4.2 - Companies House Number
AWS – 09162820, PwC OC303525
1.4.3 - Third party role
Amazon Web Services provides the hosted contact centre platform and associated AI-enabled services. PwC supported system design, configuration, and initial implementation.
1.4.4 - Procurement procedure type
Framework call-off
1.4.5 - Third party data access terms
AWS operates strictly as a data processor on behalf of Newcastle City Council. Council data is not used by AWS to train models. Access to data is limited to service operation and is controlled through council-managed, role-based permissions.
Tier 2 - Description and Rationale
2.1 - Detailed description
This transparency record covers the AI-enabled components of the council’s AWS Amazon Connect contact centre service—Amazon Q in Connect and Amazon Contact Lens—used by contact-centre agents and supervisors to support customer interactions. Amazon Connect routes calls/chats to the right team, Amazon Q provides agents with real-time, approved knowledge prompts and suggested responses, and Amazon Contact Lens automatically transcribes conversations and generates interaction analytics (e.g., key topics/trends) to support quality assurance and staff coaching; Amazon Lex may also be used to power IVR/chat menus for self-service and triage. The scope is limited to inbound/outbound contact-centre interactions handled through Amazon Connect and the associated supervision/quality processes; the outputs are advisory and do not make or automate decisions about residents (e.g., eligibility, enforcement, or case outcomes). Limitations include that prompts, transcripts and analytics can be affected by audio quality, accents/background noise, incomplete information and model confidence, and the tools are not used outside the contact-centre context or as a sole basis for decisions—staff remain responsible for verification, follow-up and any actions taken.
2.2 - Benefits
Amazon Q (in Connect) aims to provide agents with real‑time, approved knowledge prompts and suggested responses during calls/chats, helping staff locate the right information faster and respond more consistently. Expected benefits include: shorter average handle time and after‑call work, improved first‑contact resolution, fewer transfers/escalations, better compliance/consistency in customer communications, and improved agent confidence (including quicker onboarding for new starters). Amazon Contact Lens automatically transcribes calls and generates interaction analytics (e.g., key topics/trends and QA signals), giving supervisors better evidence for quality assurance, coaching and service improvement. Expected benefits include: improved QA outcomes, wider coverage of call monitoring, faster identification of recurring issues and training needs, reduced manual note‑taking, and better reporting/audit trails.
2.3 - Previous process
Legacy telephony systems with manual routing and limited analytics / intelligence.
2.4 - Alternatives considered
On-premise and basic hosted solutions were considered but did not provide sufficient resilience or staff-support capabilities.
Tier 2 - Deployment Context
3.1 - Integration into broader operational process
Amazon Connect is embedded in day‑to‑day contact handling for council services (phone and digital). Amazon Q in Connect supports agents during live interactions by surfacing approved knowledge, suggested responses and prompts (e.g., questions to ask, next steps, signposting and escalation guidance). Agents use this information to resolve enquiries more consistently, reduce unnecessary transfers, and capture accurate notes/actions, but staff make all decisions and can ignore suggestions.
Amazon Contact Lens supports post‑interaction and supervisory processes by producing call transcripts and interaction analytics (e.g., themes/topics, QA/compliance indicators and coaching insights). Supervisors and team leads use these outputs to prioritise quality reviews, identify training needs and recurring issues, and improve scripts/knowledge content and service processes. Outputs are advisory and do not automate eligibility, enforcement or case outcomes.
3.2 - Human review
All outputs are reviewed by council staff and treated as advisory.
3.3 - Frequency and scale of usage
Used daily across council contact services.
3.4 - Required training
Staff complete system, data protection, and responsible AI use training before access.
3.5 - Appeals and review
N/A
Tier 2 - Tool Specification
4.1.1 - System architecture
High-level system architecture (Amazon Q and Amazon Contact Lens within our Amazon Connect deployment):
-
Channels & routing (Amazon Connect): Residents contact the Council via telephone and/or digital channels that terminate in Amazon Connect (hosted in UK AWS regions). Amazon Connect contact flows provide IVR/self-service steps (where configured), capture interaction metadata, and route the contact to the appropriate queue/agent in the agent workspace.
-
Live agent support (Amazon Q in Connect + Contact Lens real-time): During a live call/chat, Amazon Contact Lens can generate real-time transcription and interaction signals. Amazon Q in Connect is embedded in the agent workspace and uses this interaction context (and configured/approved knowledge sources) to surface suggested responses, next-best actions, and draft summaries with citations. Outputs are advisory and support the agent’s work; agents remain responsible for what is communicated and actioned.
-
Post-interaction analytics (Amazon Contact Lens): After the interaction, Contact Lens produces transcripts and analytics (e.g., key topics/trends, sentiment and QA/compliance indicators). Where configured, post-interaction redaction is applied to reduce exposure of sensitive data before transcripts/artefacts are stored and viewed. Supervisors use these outputs for quality monitoring, coaching and service improvement within the Connect environment.
-
Data, security & integrations: Interaction artefacts (e.g., recordings if enabled, transcripts, Contact Trace Records and analytics) are stored in Council-controlled AWS services (commonly Amazon S3) in UK regions, with retention and access controls defined in the Council DPIA. Access is via Azure AD SSO with role-based permissions enforced through AWS IAM; data is encrypted in transit and at rest (AWS KMS) with audit/monitoring via CloudTrail/CloudWatch. Integrations with Council line-of-business systems (e.g., CRM/case management/knowledge bases) are implemented via APIs/AWS Lambda where required.
Supporting documentation (public): https://aws.amazon.com/connect/ ; https://aws.amazon.com/connect/contact-lens/ ; https://docs.aws.amazon.com/connect/latest/adminguide/contact-lens.html ; https://aws.amazon.com/q/ ; https://docs.aws.amazon.com/connect/latest/adminguide/what-is-amazon-connect.html
4.1.2 - System-level input
The system receives audio from telephone calls and text from digital chat interactions between residents and council contact centre staff. Interaction metadata such as call timestamps, routing information, and agent identifiers is also generated as part of normal contact handling.
4.1.3 - System-level output
The system produces call routing outcomes, real-time prompts for agents, call transcripts, and analytical outputs such as detected topics and trends. These outputs are presented to staff through the contact centre interface and are advisory only.
4.1.4 - Maintenance
System configuration and monitoring are carried out by Newcastle City Council. AWS maintains underlying infrastructure.
4.1.5 - Models
AWS-managed pre-trained machine-learning models are used for speech recognition, transcription, summarisation and analytics. Models are not trained on council data.
Tier 2 - Model Specification
4.2.1. - Model name
Not applicable. The models used within AWS Contact Centre Services are AWS‑managed, pre‑trained models embedded within Amazon Contact Lens, and Amazon Q in Connect.
4.2.2 - Model version
Not applicable. Model versions are managed and updated by Amazon Web Services and are not exposed to Newcastle City Council.
4.2.3 - Model task
The models perform speech recognition, natural‑language understanding, transcription, summarisation, and interaction analytics to support contact‑centre operations. They do not perform decision‑making tasks.
4.2.4 - Model input
Model inputs include audio from telephone calls, text from digital chat interactions, and associated interaction metadata generated during contact‑centre use.
4.2.5 - Model output
Model outputs include call transcripts, detected topics and trends, analytical insights, and real‑time prompts or information presented to contact‑centre agents.
4.2.6 - Model architecture
Not applicable. Model architecture details are not disclosed, as the models are proprietary, AWS‑managed services.
4.2.7 - Model performance
Not applicable. Model performance testing, evaluation, and benchmarking are conducted by Amazon Web Services as part of service delivery and are not directly controlled or assessed by Newcastle City Council.
4.2.8 - Datasets and their purposes
Not applicable. Newcastle City Council does not provide or manage datasets for training, validation, or fine‑tuning of these models.
2.4.3. Development Data
4.3.1 - Development data description
Not applicable. Newcastle City Council does not supply, create, or oversee datasets used to develop the AWS‑managed models.
4.3.2 - Data modality
Not applicable.
4.3.3 - Data quantities
Not applicable.
4.3.4 - Sensitive attributes
Not applicable. No council data is used in model development.
4.3.5 - Data completeness and representativeness
Not applicable.
4.3.6 - Data cleaning
Not applicable.
4.3.7 - Data collection
Not applicable.
4.3.8 - Data access and storage
Not applicable.
4.3.9 - Data sharing agreements
Not applicable. No development data is shared by Newcastle City Council.
Tier 2 - Operational Data Specification
4.4.1 - Data sources
Telephone call audio, digital chat text and interaction metadata.
4.4.2 - Sensitive attributes
Personal data and special category data may be present depending on the interaction. Automated redaction reduces exposure.
4.4.3 - Data processing methods
Not applicable.
4.4.4 - Data access and storage
Data is stored in UK AWS regions with strict access controls and retained up to 12 months.
4.4.5 - Data sharing agreements
Data is shared only with AWS as a contracted processor under strict contractual terms.
Tier 2 - Risks, Mitigations and Impact Assessments
5.1 - Impact assessments
A Data Protection Impact Assessment (AWS Contact Centre DPIA v6) has been completed. A Telephony Privacy Notice explains call recording and transcription. https://new.newcastle.gov.uk/budget-policies-performance-data/policies/privacy-policies/telephony-privacy-notice#:~:text=we%20collect%20information%20when%20you,Privacy%20Policy%20for%20more%20detail
5.2 - Risks and mitigations
Risks include privacy concerns, transcription inaccuracies and potential bias, particularly for people with speech impairments, strong accents or where English is not a first language, as well as over-reliance on automated insights. Mitigations include human oversight of decisions, data redaction, testing and monitoring, staff training, and governance controls.