Speech Recognition Makes its Claim at FirstAssist

In an industry notorious for mountains of paperwork, cutting time and reducing costs is seen as a necessity in the insurance business. It is a challenge that FirstAssist knows all too well. Handling thousands of insurance claims each year and resolving hundreds of technical queries each day, FirstAssist is one of the UK's leading providers of legal expenses insurance.

For the legal expenses insurance division of FirstAssist, improving response times to customers with complex claims has always been a priority. The department has weekly performance indicators, which the employees have to meet, and although the claims handlers complete their part of the job quickly, they have traditionally been heavily reliant on a typing pool to transcribe reports before they could be sent to customers.

Julia Abercromby, legal expenses project manager for First Assist, says her claims officers required a solution to speed up this lengthy process and reduce document turnaround times. They had previously used analogue dictation machines to record their work, before providing the tapes to a typist to transcribe and then send an electronic copy back to the claims handler for review.

"Unless a report was urgent, it would go into a pile within the typing pool, which was sometimes stacked with tapes awaiting transcription. The claims controllers would dictate say, 10 letters on a Monday and they might not go out until Thursday, which became very frustrating," Julia says. "There was nothing worse than doing a whole afternoon's work then discovering a few days later that it hadn't gone anywhere.
"Unexpected sickness and absenteeism within the typing pool would exacerbate the problem. While we brought in temporary staff to cover the workload, they were unfamiliar with the systems that we had in place and it would therefore take them longer to transcribe documents and the backlog would build."

The Right Time for Speech Recognition
Speech recognition was a solution that had been considered for several years before, but after a presentation from speech technology experts SRC, Julia and her team decided the time was right to invest in March 2002.

FirstAssist turned to SRC for advice on the right solution for their needs and a pilot implementation soon followed. During this pilot, 6 claims controllers were trained by SRC to use a desktop speech recognition solution combining Dragon NaturallySpeaking Professional V5 speech recognition software, plus a dedicated corporate law vocabulary developed by SRC, which allows documents to be dictated through microphone headsets directly into users' computers.

A roll-out of the solution to a further 24 claims controllers followed in July 2002 using Dragon NaturallySpeaking Professional V5, with SRC delivering a comprehensive train-the-trainer course to enable FirstAssist's own team to deliver user training as required.
"SRC's experience in developing vocabularies and successfully deploying speech solutions was key to ensuring that FirstAssist maximised their return on investment and that aggressive targets for document turnaround times were met," SRC commercial director, Colin Howman, says.

Speaking the FirstAssist Language
"SRC's corporate law vocabulary for FirstAssist included 4,500 specialist legal words and phrases to ensure that even the most complicated 'legalese' would be recognised by the system.

"A dedicated vocabulary to match the needs of users is absolutely vital to achieve maximum usability. It ensures there are no lengthy user "learning" sessions that give poor initial accuracy and make for a steep learning curve - instead, users at FirstAssist were able to use the system effectively from day one and see immediate benefits."
"Buying the specialist legal vocabulary and having SRC on hand to update the vocabulary with specific phrases and company names we use regularly, proved to be worthwhile. We also encourage our employees to add to their own words, like client company names, to their own system's vocabulary, if required, especially if they are handling claims which may go on for many months, to lift accuracy even further," Julia says.

Achieving Clear Benefits
The division began using speech recognition last autumn and has already seen clear benefits. "By using speech recognition, claims handlers can now produce their own documents in minimal time and it has reduced the chance of them getting lost or delayed within the typing pool. Our employees are very happy with the system as they find they can complete routine everyday tasks in half the time it used to take," Julia says.

"The software gives our division the capability to generate reports instantly and the ability to send them out the same day. Our claims handlers now have more control when documents go out to customers, which is a huge benefit.”

"Current users do not use speech recognition for short work such as standard paragraphs for letter writing as this can be done through macros and typing. Employees tend to use the technology when writing long, non-standard letters and claims responses, which can be dictated much faster than typing. We no longer have to factor typing pool backlogs into our service standards and customers are getting their responses faster as document production times have dramatically improved throughout the division."

Several other divisions of FirstAssist have now decided to pilot speech recognition to help improve efficiency. These include the advice and counselling division, which has several counsellors who need to keep a lot of handwritten notes and it is hoped speech recognition will accelerate turnaround times. The travel and medical team, which provides medical aid for travel insurance, is also piloting the system.

No longer fighting mountains of paperwork in the legal expenses division, Julia says the technology benefits are well proven. "We are already seeing a return on the investment within the legal expenses division, she says. "Productivity on a day-by-day basis has dramatically increased, and we expect to claim the costs back after the first year. We would definitely recommend speech recognition."