Sovra provides an end-to-end procurement platform connecting government agencies with vendors across the United States and Canada. With over one million vendors on its network, the company processes thousands of customer interactions monthly.
Recently acquired by KKR, Sovra is onboarding over 10,000 users per month while operating under a mandate to scale lean.
Before Voxjar
At 10,000 to 15,000 monthly interactions and climbing, Sovra's quality assurance function was stretched to the breaking point. A single QA professional was manually reviewing roughly 3% of calls - a shotgun approach that left the organization flying blind across its most important customer touchpoint.
The Bottlenecks
- Manual process introduced bias and routinely skipped longer calls.
- Evaluator found to be too lenient on critical behaviors like empathy and brand language.
- Zero capacity left for coaching or building training documentation.
- Thousands of monthly conversations containing strategic intelligence were completely inaccessible.
"Calls were always a black hole of information for us. So now that we have visibility, there's so much information we didn't realize we were missing. It's fantastic."
The Solution
Louis evaluated roughly a dozen QA tools. What set Voxjar apart was accessibility. Other vendors required lengthy sales processes and live demos, but Voxjar's self-service model let Louis validate fit before committing.
Self-service Onboarding
Sign up, prep AI questions, and get up and running in a matter of minutes.
Cost Control
Flexibility to set strict budget limits on spend, functioning similarly to PPC limits.
"I signed up for an account and started prepping AI questions. In a matter of minutes I was up and running. I knew exactly what the product was. Best onboarding experience I've ever had."
The Results
Voxjar enables Sovra to analyze approximately 80% of its monthly call volume, up from 3%, with the same single QA professional. That person now spends their time on coaching and documentation instead of simply listening to recordings.
Key Breakthroughs
- 27x more coverage with zero new headcount.
- Baseline agent scores improved dramatically to 80s and 90s.
- The manual scoring of calls is largely eliminated.
Calibration that drove deeper understanding
Using Voxjar's global prompts and per-question instructions, the team fine-tuned how the AI evaluated everything from script compliance to empathy. The calibration process itself proved unexpectedly valuable - when the AI scored empathy lower than the human evaluator, the team initially questioned it, then realized the AI was applying a more rigorous standard.
"Figuring out how the AI worked to score empathy actually helped improve all of us and directed our research on how to teach our agents to be more effective at demonstrating empathy."
The rollout to agents was managed carefully. Louis brought an agent ambassador into early testing to champion the tool, ensuring agents understood Voxjar was a development tool - not a punishment mechanism.
Agent scores that speak for themselves
With consistent, comprehensive feedback reaching every agent, performance improvements were dramatic. Agents who had scored in the 50s and 60s began routinely hitting the 80s and 90s. Then something happened that had never occurred in the company's history.
"The first time somebody got a hundred on the scorecard, we were shocked. Nobody gets 100. Nobody. We went back and looked, and the AI was accurate. His call was perfect in every respect. Now the entire team regularly reach 100, a significant change."
Customer calls became a strategic asset
What was once a black hole of information became an intelligence pipeline serving multiple departments. Voxjar now surfaces competitive intelligence, product complaints, and customer sentiment from every analyzed call - insights that reach the CEO, sales leadership, product, and retention teams.
"Being able to gain additional insights in customer pain points at scale to provide the best experience possible has been invaluable in improving the customer experience."
The company also tracks sales transfers - moments when support agents identify upsell opportunities and hand off to sales - monitoring both script compliance and transfer volume over time.
Learnings for successful implementation
"Don't expect it to work perfectly right out of the gate. AI isn't meant for that. But once you tweak it according to your business need, your gains will be tremendous."