Customer support automation, built for a global payments leader.
A global payments leader with 30M+ devices across 150+ countries cut a sixth off every support call, releasing nearly 3,000 support hours a year, worth $166,500.
A global leader in payment and commerce technology.
The company provides payment and commerce solutions to businesses worldwide, with 30M+ devices deployed across 150+ countries. Its technology supports merchants, retailers, and financial institutions across markets.
Where the 10 minutes a call were going.
Every interaction took 10 minutes, much of it spent retrieving information rather than diagnosing the issue.
No single source of truth
Customer records, device identity, site configuration, product documentation and troubleshooting procedures each lived in a different system, and no single one held a complete answer.
Six systems, every call
Agents moved between 5 or 6 interfaces before troubleshooting could even start, while the customer waited on hold.
Senior expertise absorbed by routine work
Experienced agents spent significant portions of the day on routine L1 inquiries, reducing their availability for complex escalations.
Cost scaled with headcount
Against roughly $20 million in annual support operating cost, volume growth translated directly into hiring pressure.
What they tried first, and where it stopped.
The natural first step is a chatbot. But a chatbot only handles the front door of support. To reduce average handle time, 4 things had to happen on every call.
Resolve routine requests independently
So skilled agents stop absorbing questions that don’t need them.
Preserve context on escalation
Keep the full conversation, device and account details intact when a case genuinely needs a human.
Deliver knowledge instantly, mid-call
Put the answer in front of the agent while the call is live, not behind a search.
Fit the existing stack
Work across the CRM, telephony and device systems already in place, without replacing any of them.
Two layers. One assistant for the customer, one for the agent.
A conversational layer resolves what it can at the front door. A real-time assistance layer works alongside the agent through everything that isn’t resolved there.
One interface across 6 support systems.
No migration. No replatforming. The assistant runs against the systems already in place.
What happens when the assistant doesn’t have the answer.
When the system can’t resolve an issue, it hands it to an agent with all the context they need.
The agent used to search. Now the agent just answers.
Before: an agent moves between 5 or 6 systems mid-call, hunting for account history, device data and troubleshooting steps, while the customer waits on hold. After: the assistant retrieves everything into a single panel, and the agent focuses on diagnosis and resolution.
What changed, by the numbers.
Per interaction, before and after.
The same model, cleared to scale.
The trial results were strong enough to change the deployment plan. Leadership moved from a trial cohort to an organization-wide rollout, with more gains expected as the knowledge base matures and retrieval coverage extends further across the product estate.
Got a use case in mind?
Platform, engineers and governance all in. We’ll map your workflow against the same 4 tests this one had to pass.
