Case study 2 min read
A fintech scale-up
An AI support assistant that cut manual ticket work by 72%
A fintech scale-up had a support inbox growing faster than its team. Infoloop delivered an AI support assistant that prepares the routine replies, while support agents keep every decision.
- Industry
- Fintech
- Services
- AI support assistant, Ongoing support and monthly tuning
- Timeline
- 5 weeks to first version live
Introduction
The objective: routine questions answered in minutes, with an agent approving each reply
A fintech scale-up was adding customers faster than it could recruit support staff. Ticket volume had outgrown the team, and response times were slipping.
The brief was a fast deployment of an AI assistant for routine support work. Support agents would keep final approval on every customer reply.
The challenge
Most tickets were the same small set of questions, asked in different words.
- Three questions filled the queue all day: refund status, account holds and unexplained payments
- Each answer meant opening three separate applications and copying account data between them
- Agents retyped the same replies many times each week
- First response time had crept from minutes to hours, and customers had started to complain openly
- Additional hiring would only have absorbed a few months of growth
The business needed fast answers to routine questions, with judgment kept in the hands of its support agents.
Our approach
We analyzed a week of real tickets instead of relying on estimates from managers. Three request types followed consistent rules and used data the company already held. Those three categories were suitable candidates for automation. Every other request type stayed with an agent, as agreed on day one.
Connected to the tools agents already use
Linked to the support inbox, the payments platform and customer records. It pulls the account data an agent would otherwise look up by hand.
Draft replies and one-click actions
Prepared replies and one-click actions cover the three most common request types.
Agent approval on anything involving money
No action that affects a customer's balance goes ahead until an agent approves it.
Queue reporting on one screen
Shows ticket volume, response times and how much work the assistant completed on its own.
The results
The first production version went live in five weeks. Routine questions now receive answers in minutes, and agents focus on the cases that require human judgment.
-72%
manual ticket hours
<2 min
first response time
5 weeks
to first version live
Technology used
A quick look at what runs behind this build.
| Technology |
|
|---|---|
| What we did |
|
| Runs today |
|
Keep reading
More of our work
Manufacturing
Predictive maintenance and one ERP across three plants cut unplanned downtime 72% and saved a global machinery manufacturer $1.2M a year.
A global industrial machinery manufacturerRead case study
Automotive
DTC retail
Automate the routine work at your support desk
AI support assistants that draft routine replies while your agents keep final approval. A written quote before work starts, with support and monthly tuning after launch.