AI Agent Operational Lift for Captioncall By Sorenson in Salt Lake City, Utah

Why AI matters at this scale

CaptionCall by Sorenson is a leading provider of internet-based captioned telephone services for individuals who are deaf or hard-of-hearing. The company operates as a federally funded Telecommunications Relay Service (TRS) provider, offering specialized phones and apps that display real-time captions of a hearing party's speech. With a workforce of 5,001–10,000, it serves a massive user base, processing an enormous volume of sensitive, real-time audio data. At this scale, even marginal improvements in caption accuracy, operational efficiency, or user personalization can yield significant competitive advantages and enhance critical accessibility services for a vulnerable population.

For a company of this size in the regulated telecom/assistive technology sector, AI is not a distant future but a present-day lever for transformation. The core service—converting speech to accurate, timely text—is fundamentally an AI/ML problem. Legacy systems relying on human captionists or older speech recognition are costly and can struggle with scale, accents, or specialized vocabulary. AI enables automation of routine captioning, deep personalization for users, and intelligent analysis of call quality and user needs. This can drive down operational costs per call, improve service quality metrics crucial for regulatory compliance and funding, and create new, sticky features that improve user loyalty in a mission-driven market.

Concrete AI Opportunities with ROI Framing

1. Enhanced Automatic Speech Recognition (ASR): Implementing state-of-the-art, domain-tuned ASR models can directly reduce the number of calls requiring a live captionist. A hybrid model where AI handles clear, routine calls and humans handle complex ones can optimize labor costs. The ROI is direct: reducing the variable cost per call while maintaining or improving quality standards mandated by the FCC's TRS program.

2. Predictive User Support and Proactive Care: Machine learning algorithms can analyze call history, device data, and user interaction patterns to predict which users are likely to experience technical difficulties or which might benefit from specific feature tutorials. This enables proactive outreach, reducing inbound support call volume and improving customer satisfaction (CSAT) scores. The ROI manifests as lower customer churn and reduced support center operational expenses.

3. AI-Driven Quality Assurance and Compliance: Manually auditing calls for caption accuracy and latency is sampling-based and labor-intensive. An AI model can perform 100% automated QA, flagging outliers for human review. This ensures consistent service quality, provides auditable data for regulatory submissions, and protects against non-compliance penalties. The ROI includes labor savings in the QA department and risk mitigation against costly compliance failures.

Deployment Risks Specific to This Size Band

Deploying AI at a company with 5,001–10,000 employees introduces specific challenges. Integration Complexity: Legacy systems, likely spanning call routing, captionist workstations, and customer databases, create a complex tech stack. Integrating new AI models without disrupting 24/7 service is a major technical hurdle. Change Management: Shifting workflows for thousands of employees, including captionists and support staff, requires extensive training and clear communication to mitigate resistance and ensure smooth adoption. Data Governance at Scale: Handling petabytes of audio data—which may contain protected health information (PHI)—requires enterprise-grade data pipelines, stringent security, and ethical AI frameworks to maintain user trust and meet HIPAA/FCC requirements. Regulatory Scrutiny: As a large TRS provider, any material change to the captioning service, especially one involving automation, will face intense regulatory scrutiny from the FCC, requiring transparent testing and validation to prove non-discrimination and quality maintenance.