Case Study
Winning With Quicker Polling Results And AI
Sector: Polling Services
Client Context
A market leader in polling services found that it was unable to keep up with growing demand for its services, as slow turnaround time not only kept them from growing their business, but also put them at risk of losing existing clients to more nimble competitors.
Challenge
Losing Momentum due to Slow Transcription
Poll data is only as good as the speed with which it’s collected and analyzed. With politicians trying to calibrate their campaign tactics based on immediate polling results, it’s imperative that these results be turned around as quickly as possible. Yet for our client, their phone polls took significant time to be transcribed, and then just as long to be studied and collated into actionable insights. With growing demand for their services, our client found that they were particularly unable to match occasional spikes in that demand, leading to delays for existing clients as well as missed opportunities to expand the business. The biggest hurdle was in the inability to transcribe interviews quickly enough, which caused delays along the entirety of the process.
The Work
Getting rid of the Bottleneck
NextAccess began with a comprehensive assessment of the client’s current end-to-end-polling methodology, as well as on the subsequent data processing and analysis and the impact of delays on the timeline. It soon became clear that the transcribing process was the biggest bottleneck, followed by pulling key data from the interview transcripts. Both of these largely manual processes were at the mercy of outsourced services, where delays in turnaround were not uncommon. NextAccess saw this as a perfect opportunity for the client to modernize their system by incorporating AI/ML enablers that would automate the entire backend, thereby significantly speeding up turnaround times for synthesizing interview content.
The Solution
Using AI Logic to Automate the Process
To build out a proof of concept model, we tapped into AWS APIs to write code that automated the entire process, by capturing the audio files, creating storage space, iterating through the files, and running it all through AI logic. These files were then run through a sentiment analysis algorithm, with the resulting assessment spun into charts and word clouds. This real-time beta program clearly demonstrated how implementation of AI into this part of the approach not only provided much quicker output, but also provided a bulwark against errors introduced through human interpretation.
The Result
Getting Clients Back on their Game
By delivering an automated, AI-driven alternative, NextAccess was able to help the client trim turnaround time for poll interview results from two days to two hours, while improving transcription quality and significantly reducing costs per transcription. Focusing the migration on the most cumbersome and time-consuming parts of the process meant that the polling company could focus on the core sales and marketing aspects of the business, while still realizing quantifiable cost savings and efficiency gains—and, perhaps unfortunately, enabling its politician customers to rapidly pivot to new messaging, thrown at constituents on a 24/7 basis.
