From Pilot to Production: What a US Insurer's Voice AI Rollout Teaches MENA Enterprises
A US health insurance agency took voice AI from an after-hours pilot to a full-time agent during its busiest season. Here's what MENA banks, insurers and healthcare providers can learn from how they did it.
How eHealth, a US health insurance agency, moved its voice AI deployment from a small pilot to a full-time agent handling its busiest period of the year.
The problem was one every MENA contact centre will recognise. During Medicare open enrollment, a window of about 54 days, eHealth handles roughly half a million calls. Despite having a website, 70% to 80% of its customers buy over the phone. Even with long shifts and weekend work, callers were waiting up to six hours to reach a licensed agent.
eHealth deployed Regal's voice AI agent to prescreen callers before passing them to a human. According to data Regal shared with CX Dive, the agent now answers every call immediately, has received 77% "exceptional" satisfaction ratings, and delivers a 27% higher purchase rate than the outsourced human screeners it replaced.
The results matter less than how they got there.
Start small, start after hours
eHealth didn't launch with a full-time agent. It began in February 2025 with an after-hours pilot that only handled calls arriving when the contact centre was closed. What it learned there shaped the full deployment.
This is the right model for MENA enterprises too. After-hours calls are low-risk to automate, since the alternative is usually voicemail or nothing. They give you real conversations to learn from before the agent touches your peak hours.
Regulated industries need hard rules, not just smart AI
Insurance and healthcare come with strict rules on what can be said in a call. One example from eHealth: a lengthy privacy statement that must be read in full, without interruption. Callers kept cutting in, which forced the agent to start over.
The fix was a fixed, scripted segment that isn't generated by the AI and can't be interrupted. Once the statement finishes, the agent checks the caller is still on the line and moves them to a human.
This is the pattern MENA banks, insurers and healthcare providers should insist on. Generative AI handles the natural conversation. Mandatory disclosures, consent language and regulated wording stay locked and scripted, in Arabic and English, exactly as your compliance team approved them.
Expect more than one person on the line
eHealth also found that some customers call on their own behalf, while others call with power of attorney, sometimes with two people speaking at once.
In MENA this is the norm, not the edge case. Adult children call for elderly parents, and family members join the call midway. Your agent needs to handle multiple speakers, verify who has authority to act, and do it in whichever dialect each person is speaking.
Understanding beats keyword matching
Older voice automation needed to be programmed for specific words. eHealth's agent, built on a large language model, understood a caller who had just lost her husband even though she never said it directly, and responded with care.
For Arabic speakers, this matters even more. People rarely say things the same way twice, and phrasing changes from Riyadh to Beirut to Casablanca. An agent that understands meaning, not just keywords, is what makes dialect support actually work.
Bring compliance in from day one
Internally, eHealth faced the usual fear: what if the AI says something outrageous? Its answer was to make compliance a core stakeholder from the start, alongside quality assurance, telephony, CX managers and trainers. Senior leaders listened to real calls so they could hear what the agent was doing.
As eHealth put it, "for any business that's regulated it's not a plug-and-play solution." Getting it right took iteration, A/B testing, and people simply listening to calls to check they sounded right, even when the metrics looked fine.
What this means for MENA
Every MENA market has its own version of open enrollment: Ramadan, Eid, salary days, insurance renewal periods, travel peaks. These are the moments when contact centres break and customers wait.
The lesson from eHealth is to pilot where the risk is lowest, lock down the regulated parts, involve compliance early, and let the agent earn its way into peak season.
Planning a voice AI rollout in a regulated industry? Book a call and we'll walk through where a pilot makes sense for your operation.