AI Voice and Digital-Service Agent.

Old Mutual Corporate's contact centres handle a high volume of routine enquiries. The opportunity was to use conversational AI to serve customers around the clock, improve capacity and allow human agents to focus on interactions that need judgement or empathy.

As AI Solution Architect in the AI and Automation team, I designed a text-and-voice agent for contact-centre and WhatsApp channels. The architecture combines governed large language models, a retrieval-augmented knowledge base and integrated tools to answer common enquiries and complete approved actions.

I produced the high-level architecture, business context, cost model, total cost of ownership and return-on-investment case for executive and architecture-board review. I then developed the detailed design for technical approval and continued working with engineering through implementation and production readiness.

The AWS-based solution uses Amazon Bedrock models, LiveKit and Amazon Nova Sonic, with supervisor, human-in-the-loop and observer controls. Its target is to address the six most common enquiry types, relieve up to 60% of current contact-centre workload and reduce associated operating expenditure by an estimated two-thirds.

• Designed a governed AI agent spanning voice and WhatsApp channels.
• Presented the architecture and investment case to executive stakeholders.
• Combined knowledge retrieval with secure tool integration for customer service.
• Designed for human oversight, operational control and production observability.
• Targeted 24/7 service, up to 60% workload relief and substantial operating-cost savings.