Sigma launches independent verification service for AI agents

Sigma has launched Sigma Eval, a human-supervised verification service for enterprise conversational AI systems that tests safety, user experience and performance before and after deployment. The company is targeting a growing governance gap as more organizations put AI agents into production without independent checks.

Why it matters: - Enterprise AI agents are moving into customer service, sales, claims handling and internal support faster than many organizations can verify their behavior. - Sigma Eval is designed to give companies an external, evidence-based check on how conversational AI systems actually behave with real users. - The service aims to reduce business risk tied to customer-facing failures, unsafe outputs and governance blind spots.

What happened: - Sigma announced the launch of Sigma Eval, an independent verification service for enterprise conversational AI agents. - The service evaluates systems across 12 dimensions of safety, user experience and performance. - Sigma says the service works both before deployment and after deployment. - The company said the launch is meant to address changing demands in AI governance for enterprise conversational AI. - Sigma positioned the offering as an assurance layer that sits outside the AI system being assessed.

The details: - Sigma Eval measures bias, toxicity, hallucinations, opacity, PII exposure and vulnerability to manipulation under the safety category. - The user experience category covers context misalignment, conversational inconsistency and user disengagement. - The performance category covers trajectory performance, customer effort and resolution cost. - The evaluation uses a proprietary system that combines synthetic users with synthetic conversations. - Sigma said the system can exercise an agent at scale and in any language. - Customers receive a scorecard that reports each dimension separately instead of a single aggregate score. - Sigma says the service requires no integration work, no code access and no model access. - An organization submits a public link to its conversational agent, along with contact details and context such as purpose, languages, KPIs and known edge cases. - Sigma Cognition runs the evaluation and sends a confidential report by email. - Reports are shared only with the requesting organization and are not used for public benchmarking or promotion. - Sigma is offering a free evaluation report to any organization operating a conversational agent. - The company is also making a limited number of repeat evaluations available so teams can test the impact of remediation work. - Sigma said its verification work builds on applied research in AI safety, evaluation and anonymization with academic and institutional partners. - That research base also supports Sigma Cypher, which removes personally identifiable information before analysis.

Between the lines: - The launch reflects a broader shift from internal testing toward third-party assurance for high-impact AI systems. - Sigma is arguing that the same entity building or tuning an agent should not be the only one judging its safety and performance. - The company is also betting that point-in-time testing is not enough, since prompts, models, knowledge bases and integrations can change after deployment. - Sigma’s framing suggests a market opening for independent AI audit tools as regulators, boards and customers demand more proof.

What's next: - Sigma plans to use the free report as an entry point for organizations operating conversational agents. - Repeat evaluations are intended to help teams validate remediation work and monitor drift over time. - The company is positioning Sigma Eval as part of a broader product set that includes Sigma Cypher and Sigma Verified. - Sigma says independent third-party evaluation may become a standard expectation for AI governance as enterprise adoption accelerates.

The bottom line: - Sigma is trying to turn AI agent review from an internal guess into an outside verification process that businesses can defend to boards, regulators and customers.