How Shrinithi scaled policy booking beyond 2,000 documents a day with AI

A year ago, increasing policy volumes created a straightforward but expensive challenge for Shrinithi Insurance Broking: every new policy document required more manual data entry.

External data-entry support typically cost between ₹7 and ₹12 per policy document. One employee could enter approximately 70 to 90 policies in a working day. When volumes increased, the firm offered its team additional earning opportunities to continue the work from home after office hours.

The arrangement helped clear immediate backlogs, but it was difficult to scale. Higher business volumes meant more manual work, more overtime and a greater risk of policy booking being delayed.

This was the operational problem that led SIBRO to introduce AI-assisted policy data entry.

The constraint was not employee effort

The data-entry team was already working hard. The limitation came from the nature of the process.

For every policy, someone had to open the document, read its contents and manually enter information into the appropriate fields. Even an experienced employee could process only a limited number of documents each day.

At 70 to 90 policies per employee, a sudden increase of several hundred policies created an immediate capacity problem. The firm could add more people or ask the existing team to work longer—but both options made every additional policy more expensive to process.

The real question was not “How can the team type faster?” It was “How much of this work actually requires a person?”

Introducing AI-assisted policy booking

SIBRO introduced AI-based optical character recognition and data extraction primarily to address Shrinithi’s challenge.

When a policy document is uploaded, the system reads the document and extracts the information it can identify. Instead of manually typing every available field, the data-entry team reviews the extracted information and handles the details that still require human judgement.

Today, AI performs approximately 90% of the policy entry. The team focuses on the remaining exceptions, including:

• Salesperson or business-owner information that is not present in the policy document
• Internal information that cannot be inferred from the document
• Values that do not match an available master or dropdown option
• Unclear, missing or unusually formatted information
• Records that require verification before the policy can be completed

This changes the role of the employee. The person is no longer responsible for reproducing the entire document inside the system. Instead, the employee checks the result, completes the missing context and resolves exceptions.

From 70–90 policies per person to days exceeding 2,000

Shrinithi now has days on which it books more than 2,000 policies. The total data-entry team consists of four people.

Under the earlier fully manual process, four employees working at the reported rate of 70 to 90 policies each would represent a daily capacity of roughly 280 to 360 policies. Processing 2,000 policies would have required substantially more people, longer working hours or a backlog extending across several days.

With AI handling most document-based entry, the same team can support approximately five to six times more overall volume. There is no longer a recurring need to offer after-hours data-entry work simply to keep pace with normal business growth.

Understanding the cost correctly

Previously, data-entry support cost approximately ₹7 to ₹12 for each policy document. SIBRO’s AI usage charge is ₹1 per document.

That does not mean the broker’s complete processing cost is only ₹1. The four-person team remains essential for reviewing the AI output, adding internal information and resolving exceptions.

The more meaningful comparison is this:

• The repetitive document-reading and typing component now costs ₹1 per policy for AI usage.
• Approximately 90% of the entry is handled by AI.
• The existing team concentrates on the smaller portion that requires human involvement.
• The same team supports five to six times more business volume.
• Additional volume no longer requires a proportional increase in data-entry staffing or overtime.

The benefit comes from combining inexpensive automated extraction with focused human review—not from attempting to remove people from the process entirely.

Why exception-based work scales better

Traditional automation is often presented as replacing a complete job. In practice, insurance documents contain too many variations for that to be the most useful objective.

Some information is consistently available and can be extracted automatically. Other information belongs to the broker’s internal process. Certain documents will always contain unusual formats, incomplete values or details that cannot be matched confidently.

A more practical operating model is:

1. Let AI complete the predictable, document-based work.
2. Identify information that is missing or uncertain.
3. Present those exceptions to an employee.
4. Let the employee apply business context and judgement.
5. Use the completed record in the rest of the policy, accounts, servicing and renewal workflow.

This approach gives the organisation the speed of automation without losing human control over the final policy record.

What Shrinithi’s experience demonstrates

The most important outcome is not simply that policy documents are processed faster. It is that policy volume and data-entry effort no longer have to increase at the same rate.

Shrinithi can now handle days exceeding 2,000 policy bookings with a four-person data-entry team. AI completes most of the repetitive work, while employees deal with the information and exceptions that genuinely require their attention.

For Shrinithi, AI was not introduced as a general technology experiment. It was introduced to solve a specific operating constraint—and its value is visible in the volume the same team can now support.

See how AI-assisted policy booking could work for your team →