Viva.com is a European tech bank for businesses, founded in Athens and now operating across Europe with local acquiring in 23 countries. Its support team runs on Intercom and staffs 24/7 cover in English and Greek. When a merchant writes in because money has moved and they do not know why, the clock on that conversation is not set by a support policy.
When the Question Is About Money That Already Moved
“Somebody has stolen money from my account.” That is a real message Viva’s support team receives, and it is the reason the deployment has a hard rule written into it. A message like that cannot be handled well by an assistant that is merely polite and quick. It has to be recognised for what it is and put in front of a person.
Most of the volume is not that. It is merchants asking to raise a daily or per-transaction limit, checking why a payment behaved the way it did, or asking something whose answer depends on whether they are the account admin or an employee on it. The same question from two different roles has two different correct answers.
Language is the other constraint. Viva staffs round-the-clock support in English and Greek. A merchant writing in another language at two in the morning is outside that cover, and the honest options are a fallback or a wait — which is exactly the gap the deployment was built to close.
And the work sits inside Intercom, not beside it. Viva’s agents already move between the helpdesk, the back office, a CRM and Microsoft Dynamics on a single case. Anything that added a fifth place to look would have made the job worse, not better.
Sofia, Inside Intercom
Open runs as an AI agent named Sofia inside Intercom, on web chat and email. Intercom stays Viva’s system of record and its agents keep the console they already use. Nothing was migrated.
Sofia handles daily and per-transaction limit-increase requests, the highest-volume repeat ask on the account, collecting and routing the request rather than leaving the merchant to describe it twice.
She also takes conversations back mid-thread. When a human agent reassigns an Intercom conversation to Sofia, the assignment event hands it over and she resumes the reply — so a case can move between agent and AI in either direction instead of only escalating one way. Custom actions call Viva’s own endpoints, configured per environment.
Security is the deliberate exception. A merchant reporting stolen funds is transferred to a live human agent and the case is retagged for manual handling. Non-English enquiries outside Viva’s English and Greek cover are offered an English handoff rather than an automated guess.
The Results
Open’s AI agent now handles the majority of Viva’s inbound merchant support:
| Metric | Impact |
|---|---|
| Automation rate | 62% |
| Tickets resolved | 58,392 in 90 days |
| European markets | 23 |
| Channels covered | Web chat and email |
The 62% figure is scoped: inbound contacts only, across the two channels integrated at the time — web chat and email inside Intercom — measured over a trailing 90-day window, 58,392 of 94,231 tickets. It is not Viva’s overall support automation rate, and it should never be quoted without that sentence attached.
The Impact
62% automation on inbound web chat and email means Viva absorbs merchant volume across a growing European footprint without scaling headcount proportionally with each new market. Adding a country adds a regulatory ruleset, not a support team.
For a merchant, the change is narrower than the number suggests. A limit request is handled while they are still in the conversation, and a report of stolen money reaches a person instead of a bot. Those are different outcomes, and the deployment is built to tell them apart.
