Sanas has introduced Supervised AI, a customer experience platform in which AI agents handle customer calls while a human supervisor monitors and can intervene in up to four conversations simultaneously. The system is designed to combine AI-driven call handling with human judgment, allowing supervisors to guide an agent, take over a conversation and return control without forcing customers to restart their interactions.

Sanas Proposes a Hybrid Model for AI-Powered Contact Centers

Sanas is positioning its new Supervised AI platform as an alternative to two established approaches to AI in customer service: using AI to assist a human agent during a call, or allowing an AI agent to manage the interaction autonomously until escalation becomes necessary.

Under the model described by Sanas, AI agents conduct customer conversations while a human supervisor oversees several calls from a single console. The supervisor can listen in, privately guide the AI, take control of a conversation or return it to the AI when appropriate.

The company says this approach is intended to avoid a common weakness in conventional escalation workflows: a customer who must repeat information after being transferred to a human agent who has not followed the interaction from the beginning.

Sanas says the supervisor remains present throughout the conversation, with alerts directing attention to calls that may need human judgment. These alerts can be triggered by falling customer sentiment, declining AI confidence or an approaching compliance step.

The distinction is operationally significant. Rather than treating human involvement as a fallback that begins only after an AI agent fails, Supervised AI makes human oversight part of the call-handling model. The approach could be relevant in industries where customer interactions involve sensitive information, complex decisions or regulatory requirements.

However, the platform’s effectiveness will depend on how accurately it identifies situations requiring intervention, how quickly supervisors can respond and whether one person can realistically oversee multiple complex conversations without missing important signals.

Human Intervention Without a Conventional Handoff

Sanas says each call begins with disclosure that the customer is speaking with AI and that a human supervisor is in command. Supervisors can then move between listening, guiding the AI and taking direct control.

The company’s stated design aims to keep the customer within the same conversation during these transitions. That could reduce friction compared with workflows in which an automated system transfers a caller to a separate queue.

The platform is designed for one supervisor to oversee up to four simultaneous conversations. This creates a potential productivity benefit, but it also introduces a capacity question: the number of calls a supervisor can effectively govern will vary with their complexity, customer needs and the frequency of intervention.

The announcement does not provide independent benchmarks for resolution rates, average handling time, customer satisfaction or supervisor workload. Enterprises will need operational testing to determine whether the model improves service quality as well as agent productivity.

Auditability and Deployment Flexibility

Sanas says every recommendation made to the supervisor cites its source document, while AI actions and human interventions are recorded in a complete audit trail. These capabilities could help organizations review decisions, investigate service problems and document how AI handled a particular interaction.

The platform is designed to run in a customer-controlled environment, including private cloud or on-premises deployments, using Sanas Fabric. The company says enterprises can change the underlying speech and language models. In the current release, the AI uses a disclosed Sanas voice.

This architecture may appeal to organizations with strict data-residency, security or technology-control requirements. Nevertheless, buyers should validate the specific deployment options, model-switching limitations, retention policies and security controls available in their intended configuration.

Enterprise Adoption Will Depend on Governance

Sanas says the platform builds on five years of work in speech AI and is intended for enterprise contact centers, including healthcare, banking, financial services and telecommunications.

The company also cites interest from UnitedHealthcare and TELUS Digital. UnitedHealthcare executive Marius Maree described the model as a way to combine AI scale with human accountability in sensitive member interactions. TELUS Digital’s Nick Williams said the company is evaluating the technology and identifying suitable use cases. These comments indicate interest, but they do not establish broad deployment or independently verified results.

For enterprises, the central question is whether supervised AI can balance automation with reliable human intervention. Buyers will need to evaluate disclosure practices, escalation accuracy, response times, audit records, privacy safeguards and the ability to handle vulnerable customers or high-risk decisions.

Sanas Supervised AI reflects a developing direction in customer experience technology: AI handles more of the conversation, while people retain responsibility for judgment and oversight. Its long-term value will depend on whether that arrangement improves customer outcomes without creating an unsustainable supervisory workload.

Market Landscape

Contact center technology is evolving across three broad models: conventional human-led service with AI assistance, autonomous AI agents and systems that combine AI-led conversations with ongoing human supervision.

Sanas is differentiating Supervised AI through the third approach. Its defining feature is not simply that human escalation remains possible, but that a supervisor can monitor multiple live interactions and intervene within the existing conversation.

That model addresses important enterprise concerns around accountability and continuity. However, it also creates a different staffing and operating model. Supervisors must be able to identify problems quickly, make informed decisions and manage competing demands across simultaneous calls.

For contact center buyers, relevant comparisons include voice quality, multilingual performance, integration with customer service platforms, escalation accuracy, compliance controls, deployment options and total cost per resolved interaction. Independent performance evidence will be important in determining whether supervised AI outperforms existing agent-assist or autonomous systems in particular use cases.

Strategic Outlook

Sanas Supervised AI could be relevant for organizations that want to automate routine customer conversations without removing human oversight from sensitive interactions. Healthcare support, financial services and telecommunications are potential evaluation areas, although suitability will depend on the specific use case and applicable requirements.

Before deployment, enterprises should establish clear intervention thresholds, define which decisions AI can make, test disclosure language and determine how supervisors handle simultaneous alerts. They should also measure customer satisfaction, first-contact resolution, escalation rates, compliance incidents and supervisor workload against an existing baseline.

The key test is whether the platform can deliver the efficiency of AI while preserving meaningful human accountability. The announcement outlines the operating model, but independent evidence of customer outcomes and productivity gains remains necessary.

Top Insights

• Human-supervised AI: Sanas lets AI agents handle customer calls while supervisors oversee up to four conversations, combining automation with ongoing human control.

• Continuous intervention: Supervisors can guide the AI, take over mid-conversation and return control without requiring customers to restart, according to the company.

• Auditability: Source-linked recommendations and recorded AI actions and human interventions are designed to support traceability and enterprise review.

• Deployment flexibility: Private-cloud and on-premises options may suit organizations with specific infrastructure and data-control requirements, subject to configuration and security validation.

• Performance remains unproven: Enterprises should test resolution rates, customer experience, intervention speed and supervisor workload before assuming the model improves contact center efficiency.

   

  Get in touch with our MarTech Experts.