Digitalising paper in a fully automated process.

How can a single piece of improvement in a workflow make a total difference to the overall operations of a business.

A long-established medium-size insurance agency based in Europe which, to safeguard its privacy, I will call “ITB” mainly issues single trip insurance policies for all sorts of travellers, students and backpackers etc. Their Administration Director explained, “we offer a standard policy, our own wording with fully delegated authority, backed by a small panel of insurers. It’s a coupon-based application form appearing in selected magazines, and these days also online blogs and Instagram.”

It is a high-volume low-ticket business with a surge of applications in the summer and another peak over Christmas. “Whether online or paper-based, ITB customers seem to leave their applications to the last minute “ the director said, “although we need only basic data to issue a policy and the process is straightforward, we never have much time to deal with these peaks comfortably by issuing the policies in time.”

The decommissioned online application legacy system used by ITB lacked Instagram API integration capability. For paper-based applications, they tried several approaches over the years, attempting to deal with the volumes flooding into their administrative centre particularly Monday and Friday. These measures included overtime, outsourcing and temporary staff. “None of these solutions were entirely satisfactory” the director confessed, “they were expensive, given the low margin of each policy, difficult to control and there were too many data errors entering the information from the paper applications.”

No wonder that we spent far too much of time managing volumes, matching resource to capacity and dealing with bottlenecks. “If we ever fell seriously behind with processing” stated, “we would then be hit by emails, social media comments and enquiries diverting our resources from issuing policies, a bit of a workload delay vicious circle. It’s very difficult to manage a business if you are spending so much time fire-fighting”.

ITB & Vesuvio Labs swap business cards in a Brokers’ Association, a meeting was scheduled a few weeks after to explore further ways of collaboration. After giving the green light for development, within about four weeks they had a new and enhanced online system up and running for testing, that streamlined all the former stages of the old process and quickly and accurately produced the policy ready for the traveller. By the end of the sixth week, the system was virtually bug-free. That however, is another case study for another article!

Returning to the hundreds of paper applications that ITB received weekly, and in pick times thousands, Vesuvio labs realised that this process was currently costly and time-consuming, because:

• The quality of handwriting and ink colour on the application coupon varied considerably.
  • The application format differed by size, appearance and question sequence depending on the publication in which they appeared.
  • Some applications were creased and stained.
  • Although the policy cover was standard, there were some variables and the duration and areas of travel varied, which produced different premiums.
  • Not all premiums had been correctly calculated by the applicants.
  • Data entry staff were subject to constant repetitive fatigue, leading to input errors.
  • All applications had to be reviewed first to identify the minority that required further attention.

After manual data entry, the outputs ITB still required were:

• Identifying and resolving applications containing queries and questions preventing processing (circa 15%).
• Premium calculation.
• Policy issue (e.g. production and distribution).
• Allocation of and bordereaux to, the insurer (on a variable percentage treaty basis).
• ITB policy records for administration and claims handling.
• Accounting and Financial controls.
• Other reports and controls (administrative, sales, underwriting suites).

The current workflow was failing and there was room for improvement. The director was however concerned that the variety and inconsistency of the applications themselves would make automation difficult to achieve. The temptation was to bane the coupon form accounting for 35–40% and to submit all the applications to online/phone; but instead of disrupting the business, we figured out:

The ideal future objective was to scan the incoming application forms and extract all the data required for the above outputs automatically.

We are witnessing adoption of platforms driven by intelligent and API-ready workflows. Despite OCR tech has been around for quite a long time, Vesuvio Labs together with its data scientists felt able to allay director’s reservations about accuracy and overcome all the problems then associated with ITB paper-based applications. Vesuvio Labs’ Advanced Optical Character Recognition (OCR) AI platform was identified as the potential cornerstone of a solution for ITB.

Essentially, for ITB, the procedures were agreed as:

• Working together to define a new cloud-based storage system.
• To establish the nature of documents to be scanned with the layout and structure of data they contained.
• Scanning application coupons for a quality enhanced image.
• Data extraction from the processed image.
• Allocation of this data from the specific fields identified for premium calculation, output processing, indexing and downloading as required.
• Production of policy records, bordereaux, policy documents, reports and analytics.Vesuvio Labs streamlines automatically expected cases while applications requiring specific attention are flagged, manually handled only by exception. The Munin OCR system achieves this amongst other things by advanced filtering and image quality enhancements. Handwriting presented no challenge when trained for coupon scale and differently formatted documents whether by form design or varying data sequences, were accommodated.

After thorough and careful testing, we were all happy to see that Vesuvio Labs could indeed successfully implement these procedures and that text and data could be accurately extracted line by line to rapidly process the applications, allocating resulting information to the specific functions required. More than that, extracted data could be posted to various locations and to a variety of useful file formats such as Excel, JSON or APIs. Also, the OCR model learned through AI the detailed layout and structure of the various documents scanned and increasing accuracy over time by recognising patterns.

“Policy documents are now streamlined after initial change; detailed bordereaux are created so effortlessly that each insurer knows the business being written faster than ever. Instead of being swamped by data we now have structured information”.

One piece of improvement in the supply chain makes the difference and adds up to the overall efficiency “the Vesuvio Labs system does not need breaks or time off. It carries-on hour after hour, extracting essential data in a seamless process, saving ITB the considerable expense of temps, outsourcing and overtime shift work.”

The various underwriting and sales reports, as well as administrative controls, that flowed naturally from this information in time to be of effective use. “What is more”, the director explained, “time spent fire-fighting and dealing with customer progress chasing enquiries has considerably diminished, so I am far more able to developing our business and to manage our people”.

Vesuvio Labs would like to thank “ITB” for their cooperation and permission to produce this illustrative case study on Munin OCR capabilities.

We would like to thank adviser Graham Ripley for his important contribution to this article.

Vesuvio Labs is a Venture builder based in London helping entrepreneurs by accelerating the process of turning ideas into a scalable business.

Follow Vesuvio Labs on LinkedIn | Listen to the Digital Rocket Fuel podcast on Spotify | Sign up for our newsletter | #vesuviolabs