An overseas education consultancy agency unified 7 different contact points — WhatsApp, Instagram, Messenger, written web chat, live voice web chat, a phone line, and human handoff when needed — into a single AI assistant infrastructure. The assistant writes directly to the CRM, the phone line is set up as a separate, per-minute billed service, and live testing confirmed correct responses in 5 languages (Turkish, Farsi, Azerbaijani, Russian, Japanese). Below is the setup and three measured figures.
The problem: every channel its own team, its own delay
In overseas education consulting, prospective students reach out through different channels — Instagram DM, WhatsApp, the website, phone — often in different languages at different hours. When channels are managed separately, two problems emerge: messages from the same prospect across different channels look like separate conversations, and keeping up with every channel at once during peak periods (enrollment season) strains the team.
The setup: one AI brain, seven contact points, one CRM
The system built connects WhatsApp, Instagram, Messenger, written web chat, live voice web chat, and a phone (PSTN) line to the same AI assistant logic; complex or sensitive situations are handed off to a person. The assistant writes the information it gathers from every channel directly into the CRM, so a prospect's record moves forward the same way regardless of which channel they wrote from. The phone line was set up under a different model than the other channels — a separate, per-minute billed service. Before every deployment, an automated test suite (over 6,000 scenarios) runs to make sure new features don't break behavior across channels.
Measured result 1: 7 channels, one assistant
WhatsApp, Instagram, Messenger, written web chat, live voice web chat, the phone line, and human handoff — 7 contact points were connected to a single AI assistant infrastructure and a single CRM, instead of separate systems for each.
Measured result 2: verified correct responses live in 5 languages
Live measurement confirmed the assistant produces correct responses in Turkish, Farsi, Azerbaijani, Russian, and Japanese, and that a region-specific exception rule (e.g. different information for a particular state) showed up correctly across all of these languages.
Measured result 3: quality is enforced by an automated test suite
Over 6,000 automated test scenarios run before every live deployment; one deployment round measured all five services going live simultaneously with zero critical errors (traceback/critical).
Next step
To discuss how requests coming from different channels — WhatsApp, Instagram, web, phone — could be unified into a single AI assistant for your business, visit the WhatsApp AI assistant service page, request a short discovery call through our contact form, or call +90 850 840 08 23 and talk live with Azra, our voice assistant.
Frequently asked questions
Where do the numbers in this case study come from?
From the project's live deployment and measurement records: the channel count comes from the system's own scope, language verification comes from real live sample conversations, and the test count comes from the automated test suite's own result before deployment.
Does every channel give the same response, or does it vary by channel?
The underlying knowledge base is shared, but per-minute billed channels like phone can have extra constraints applied (e.g. knowledge-base size); consistency across channels is protected by automated tests.
Which situations get handed off to a human?
Complex, sensitive, or out-of-scope requests (special-case evaluation, complaints) are automatically routed to the relevant team member.
Can this setup be adapted to another industry?
Yes, a similar omnichannel AI assistant structure can be built for any sector needing multi-channel, multilingual customer contact — consulting, tourism, education, healthcare; the channel count and languages are adjusted to fit.