Voice AI Contact Centre for a Panama Health Insurer
Designed and built a clickable cloud contact centre for a Panama City health insurer's member services line: an AI voice agent that says it is automated, verifies callers and handles claim status and card replacement, an agent desktop with live transcript, suggested answers, secure card payments and AI call summaries, and review tools for every AI-handled call.
Most of our work is under NDA. Per agreement, we can share the work but not the company details - all names and identifying information shown here have been changed.
Dashboard, agent desktop from ringing to wrap-up, call log, call detail with 7 tabs and AI conversation review. The desktop runs a full call: live transcript, complaint signal, secure payment and a summary sent to the CRM.
The caller keys the card on the phone keypad, the agent sees each field only as dots, and the recording shows Paused for payment until the slide-over closes. Only the last four digits, the authorisation code and the pause times are kept.
Caller asks for a person, low confidence twice, tool error, vulnerable member, complaint signal or failed ID&V. The agent receives the intent, the ID&V result and a handoff summary, so the caller does not start again.
12 AI only, 14 AI then agent and 2 agent only, across 7 queues and 8 intents, with 18 members, 14 agents, AI turns, tool calls, transcripts and payment sessions.
Seguros Corotú's member services line, about 62,000 calls a month in Spanish and English, ran on an ageing on-premise phone system with a menu seven options deep. Four in ten callers pressed 0, mostly to ask whether a reimbursement was paid or to replace a lost card. Agents re-keyed member numbers into three systems, wrote notes from memory and heard card numbers read aloud, reviewers listened to about 1% of calls, and complaints raised on the phone were often never logged.
Cayuco runs the whole line. An IVR understands what the caller says, an AI voice agent discloses that it is automated, verifies identity and handles claim status and card replacement, and every other call reaches an agent with its context. The agent desktop adds a live transcript, sourced suggestions, keypad card payments with automatic recording pause and an AI summary the agent confirms before it is written to the CRM.
A health insurer's contact centre handles sensitive health data, card payments and complaints on the same line, in two languages, under Panama's data protection law and insurance supervision. Adding AI to it only works if every automated step stays inside those rules.
A "press 1, press 2" menu seven options deep sent four in ten callers to the 0 key. Most of them only wanted simple answers that an automated agent could give, if it could verify who was calling first.
Under Law 81 of 2019 health information is generally treated as sensitive data. Claim details may only be shared after identity verification, and a voice agent must never read diagnoses, procedures or provider names aloud.
Agents took premium payments by listening to card numbers, which put recordings in PCI DSS scope. The security code must never be stored after authorisation, so card data has to stay out of the agent's ears and the recording.
Reviewers listened to about 1% of calls, picked by hand, and complaints raised on the phone were often never logged, so the insurer could not show how it handled them. Members can escalate to the Superintendencia de Seguros y Reaseguros.
Mapped 10 roles from agent and team leader to conversation designer, compliance officer and the caller, 42 screens (40 staff and 2 member mobile pages), the phone journey from greeting to secure payment as a script, and the data model for calls, AI sessions, ID&V attempts, transcripts, payment sessions, evaluations and complaints. Wrote down the rules the design follows: Law 81 and Executive Decree 285, health data as sensitive data, PCI DSS v4.0.1, recording notices, AI disclosure, the Labour Code and human review of every AI output.
Built 5 key screens on a production dashboard kit: the dashboard with 5 KPI cards, intraday offered vs forecast and needs-attention items, the agent desktop with member panel, streaming transcript, complaint signal, suggested answers, secure payment and wrap-up, the call log with handled-by filters, call detail with the path of the call, waveform, AI turns and tool calls and the AI draft vs final summary, and AI conversation review with tags and Add as test case. All actions run on local state.
Integration with cloud telephony (such as Amazon Connect or Twilio Flex) on local SIP trunks, speech to text and text to speech in Spanish and English, a language model via Amazon Bedrock, DTMF masking (such as PCI Pal or Sycurio), the payment gateway, the CRM, the core policy and claims system, the card bureau, SMS and WhatsApp, retention-locked recording storage and Entra ID SSO. Plus the remaining screens for QA and calibration, complaints, workforce planning, intents and releases, compliance and the member callback pages, and legal confirmation of the Law 81 and labour rules.
The voice agent introduces itself as an automated assistant in its first sentence and offers a person at any time. When it hands off, the agent's banner shows the intent with its confidence, ID&V verified by AI and the handoff summary.
The AI agent checks the member number or cédula, date of birth and an SMS one-time code, then reads only status, dates and amounts. For the detail it sends a link by SMS. Card digits and cédula are masked in the transcript.
Secure payment pre-fills the amount due, shows card number, expiry and security code only as received or waiting, and pauses the recording on its own. The call detail draws the pause as a grey block and the payments tab holds no card data.
Suggested answers cite their knowledge base article or record and are rated by the agent. A complaint signal needs Log complaint or Not a complaint with a reason, and the AI call summary reaches the CRM only after the agent confirms it, with the changes shown on the call.

Designed and built the staff console for a trilingual AI support agent at a Colombo telecom operator: a dashboard of containment and satisfaction, every conversation with its tool calls and guardrail events, a live monitor where team leads take over a chat, and a handoff queue that makes sure no customer passed to a person is lost.

Designed and built a clickable feedback and NPS platform for a national airline based at Addis Ababa Bole: event-triggered surveys in six languages, translated and tagged comments, detractor follow-up cases with SLAs and goodwill limits, and alerts when a score drops or a topic spikes at a route or station.
Enterprise Software Engineering - Human + AI
Contact Us
© 2026 - Copyright. All Rights reserved Oceanlab
Location
Global Headquarters
21 Merchant Road, 3rd Floor, Singapore 058267
Main corporate office and leadership operations
North America Office
548 Market Street, PMB 12345, San Francisco, CA 94104
Client operations and business development
India Office
402 Shivalik Shilp II, SG Highway, Ahmedabad 380015
Engineering and delivery center
Enterprise Software Engineering - Human + AI
Contact Us
Locations
Global Headquarters
21 Merchant Road, 3rd Floor, Singapore 058267
Main corporate office and leadership operations
North America Office
548 Market Street, PMB 12345, San Francisco, CA 94104
Client operations and business development
India Office
402 Shivalik Shilp II, SG Highway, Ahmedabad 380015
Engineering and delivery center
© 2026 - Copyright. All Rights reserved Oceanlab