How One Fintech Turned Thousands of PIN Reset Questions into an Automated Flow

About the business: The company asked to remain anonymous. It is a fintech startup that provides installment financing for gadgets and electronics through hundreds of offline retail partners across Java and Bali. With around 80,000 active customers, its customer service volume was far above average for a company of its size.

We spoke with its Head of Customer Experience, who had used the platform for eight months.

---

More Volume Than a Team Could Handle Manually

"Our first problem was not the quality of customer service. We had a strong, well-trained team with deep product knowledge. The problem was volume."

At the height of its growth, the customer service team received about 1,200-1,500 WhatsApp messages a day. With 12 people working across two shifts, each agent had to manage 50-60 conversations per shift. Consistent response quality became impossible.

"What surprised us was that almost 45% of questions covered only three topics: resetting a PIN, uploading application documents, and paying a bill. The answers were already on our website, but customers still asked on WhatsApp because it was more convenient."

---

A Phased Rollout Focused on the Biggest Problem

The company did not try to automate everything at once. It started with its highest-volume problem: PIN resets.

"We received 150-200 PIN reset questions every day. The process was simple—verify an ID card, confirm the bank account number, and send a reset link—but every case took an agent 5-10 minutes."

The customer service and technical teams built a webhook connected to the internal system. An Orqestra flow identified a PIN reset request, verified the customer's identity through a series of questions, called an internal API to create the request, and sent a link to the verified number.

"During the first week, the team's workload dropped significantly. Agents who had spent almost half their time on PIN resets could focus on more complex cases."

After PIN resets, the company automated document uploads and payment guidance. Within three months, it had automated most first-tier questions.

---

Unexpected Challenges

The rollout still required adjustment.

Some customers did not want a bot. Certain customers, especially older users, immediately typed words such as "agent" or "human" when the bot responded. The team added a shortcut to every flow that detected these keywords and escalated directly to an agent.

Customers used unexpected language. The first flows relied on formal keywords. In practice, customers described the same issue in many ways: "reset PIN," "forgot PIN," "PIN keeps failing," or "can't access the app." The team continuously updated trigger phrases based on conversations the automation missed.

"That work is still ongoing. Every week, we review conversations that an agent opened manually but a flow should have handled. Then we add keywords or improve the flow."

Customer expectations needed attention. At first, some customers were confused by replies that felt too fast and standardized. The team adjusted the bot's tone and varied the phrasing so the conversation felt more natural.

---

The Impact on the Team

Eight months later, the customer service team still had 12 people while supporting much greater volume as the business continued to grow.

"Without automation, we would probably need 20-25 people for the same volume. More importantly, the quality of our team's work improved. Turnover fell because agents were no longer burning out on the same questions. They now work on cases that are more challenging and meaningful."

The company shared these results:

• Average agent handling time fell from 12 minutes to 7 minutes per conversation
• 61% of conversations were resolved without agent escalation
• Customer Satisfaction (CSAT) rose from 3.8 to 4.3 out of 5
• First Contact Resolution increased from 68% to 84%

---

What Other Teams Can Learn

The company emphasized two lessons for businesses starting a similar journey:

Start with the problem you understand best. Do not automate everything at once. Choose one or two frequent questions with straightforward answers and build from there. An early result gives the team confidence to expand gradually.

Involve agents from the beginning. Agents understand how customers actually speak, which words they use, and which scenarios create the most friction. A flow designed by the technical team without agent input is often rigid and ineffective in real conversations.

"Great customer service is not about how sophisticated your bot is. It is about giving customers the answer they need in a way that feels human—whether it comes from an agent or thoughtfully designed automation."