Company Overview

A Canadian legal transcription company with clients including court reporting companies, federal and provincial tribunals, and claims adjusters, processing over 100 hours of audio per day, with turnaround times from 1 day to 30 days.

Business Challenge

From 2015 onwards, the company had been using speech-to-text in their legal and insurance dictation workflows, with pre-trained speaker specific AI models. However, the quality of transcripts for conversational recordings for insurance interviews and legal proceedings was very poor using these AI models. And it was expensive and time consuming to train these models on specific speakers as new dictation clients were onboarded. They were awarded two major federal and provincial transcription contracts in January 2020, just before the pandemic. And as the contracts ramped up mid 2020, during the peak of the pandemic, hiring become tougher due to a security requirement that required all transcripts be produced by staff seated in the office as part of a secure network. They were at risk of financial penalties if they were not able to meet their contractual obligations under these sole source contracts.

Solution

Loom Analytics deployed the Claudio Legal Transcription Assistant to assist the company with meeting transcription demand with their current staff. Their desktop app was integrated with Claudio through its OpenAPI, and all of its templates and client specific formatting guidelines were set up as individual Profiles on Claudio. Loom Analytics’ Structura product was also integrated into their workflow to extract relevant information from client transcription order forms and source legal documents to set up the first draft transcripts. And this was complemented with the configuration and rollout of Luis, the Loom Analytics Text Expander, to complement Claudio and further improve the efficiency of the company’s transcription team.

“Claudio has become an indispensable part of our intake process. Our staff now relies on it, reducing our transcription time from 10-15 audio minutes per hour to an impressive 40 audio minutes per hour. It's a tool we can't live without.”

— Head of Operations

Team Work

Multiple members of the company’s operations and process improvement team were involved in the initial evaluation and pilot phase over a period of two (2) months. This was followed by the training of a small number of members of the transcription team for a month (1), where they provided ongoing feedback to self-train and improve the accuracy of the Claudio Legal AI model for the company.

The Loom Analytics customer support team trained the training and process improvement lead on how to train other typists on essential skills for a successful Claudio adoption, including MS Word hotkeys, and how to edit transcripts instead of typing from scratch.