Selectra Is Glad That It Switched to Gladia

Selectra is a utility comparison and sales company that operates across 17 European countries to help customers lower their bills and simplify the management of them. The Paris-based company also operates across multiple industry verticals, including banking, insurance, energy, and telecom, so its large sales team manages thousands of customer calls each week. In France alone, it has more than 200 sales agents handling calls daily. Much of the work involves explaining options to customers, according to Alex Bouju, Selectra’s chief technology officer.

To ensure that salespeople were properly explaining different options, Selectra’s quality control department would listen in on calls, but the sheer volume meant that only a very small portion of the calls could be reviewed. “Our quality development costs were increasing a lot because we were always needing more quality department agents to listen to more calls. We had to think about doing this a different way,” Bouju recalls.

So Selectra sought a transcription and audio intelligence solution to monitor more calls.

“We looked for a solution that would help us accurately transcribe those calls so that [agents] could do their jobs better and faster,” Bouju adds. “We wanted our quality department to bring more value than just listening to the call and pointing out the points that weren’t OK.”

Selectra wanted agents to follow the lead of the best agents, who were taking the time to provide customers with good advice as well as to improve their own in-call sales skills.

Key criteria for any solution was that it offered high-quality transcription, support for a number of languages, and freedom from hallucinations to ensure the best possible input for its large language models, according to Bouju.

After reviewing its options, Selectra piloted Gladia’s speech-to-text engine to improve quality monitoring and extract in-depth insights from every call. The solution met the company’s needs during the trial, so a year ago, Selectra signed a contract for full implementation.

Going live with the Gladia technology was relatively simple, according to Bouju. “Their APIs are very well-documented. One of the great strengths of Gladia is that they were already integrated with us thanks the great period of testing, which was very positive.”

The technology’s API is optimized to provide next-best-action recommendations to customer support and sales agents while on calls.

With Gladia’s solution in place, Selectra can automate much of the quality control work that used to be done by hand, Bouju says. No time is wasted writing up call summaries, as had been done previously, because Gladia automates this task. “This lets us empower the [quality control] agents to do more because that part of the job is already done. They can spend more time helping the agent speak better or sell better.”

LLM technology enables Selectra to uncover deep insights about calls for improved training. Quality control agents now have more time to work with agents on conversation nuances like tone, empathy, and word choice, which received much less attention before.

Call reviews are completed in a third of the time that they took previously, according to Bouju.

The automation also enables Selectra to quickly verify subscription terms with customers, ensuring that contracts are accurate before sending them to the provider and that customers fully understand the terms and are satisfied.

“We earn more money depending on how long a customer stays with a provider,” Bouju says. “By verifying everything before we send it to the provider, we are sending only quality leads and qualified customers, so [the providers] are paying us better.”

Selectra hopes to improve its work with consumers even more in the future by adding live call transcriptions, providing more immediate feedback and benefits than the post-call transcriptions used today. 

The Payoff

Since implementing Gladia’s speech-to-text engine for quality monitoring and call insights, Selectra has seen the following results:

• progressed from very limited call monitoring to monitoring virtually all calls;
• saved three minutes per call, validated by a quality monitoring agent;
• spent 2.4x less time on quality monitoring with AI; and
• cut call reviews time by a third.