Sphinx has launched Sphinx 2.0, a new version of its enterprise AI product designed to reduce errors in AI-generated outputs by capturing and applying an organization’s internal knowledge.

The company says the platform sits between AI tools and a business’s data systems, acting as a control layer that checks queries, applies business logic and traces results back to source data. Sphinx describes the product as an institutional knowledge layer intended to make AI more reliable for teams that depend on data for decisions.

The launch comes as many companies continue to struggle with trust in AI systems. Sphinx says businesses often spend as much time checking AI results as they do using them, especially when different dashboards, metric definitions and business rules do not align. The company argues that this creates problems not only for accuracy, but also for accountability, since teams may not be able to explain how an AI output was produced or why it was wrong.

According to Sphinx, the new platform is built around a centralized, continuously updated knowledge base. It is designed to capture definitions, schemas, metrics and company-specific logic, then keep that information current as data environments change. The company says this approach helps AI systems use the right methodology instead of relying on inconsistent or incomplete context.

Sphinx 2.0 also emphasizes governance and traceability. The company says users can set role-based and domain-based controls for functions such as finance, human resources and operations. It also says organizations can decide how much autonomy the system has and where human review is required.

In practical terms, Sphinx says the workflow starts by connecting to existing data warehouses, databases and analytics tools. After that, the system learns a business’s definitions and rules. When an AI tool queries the data, Sphinx validates the request before outputs are delivered. It then monitors for drift, inconsistencies and hallucinations, while making results traceable back to the underlying data.

The company is positioning the product for data and analytics teams, data owners, platform leads and business executives. For technical teams, Sphinx says the goal is to reduce manual quality assurance and maintenance work. For leadership, it says the platform provides output lineage, reproducible reasoning and audit logs that can support compliance needs.

Sphinx says the product is built for enterprise environments and includes features such as zero data retention, SOC 2 Type 2 compliance, encryption, single sign-on and access tracking. It also says it supports integrations with systems including SQL databases, Google Cloud, Amazon Web Services, Snowflake and Databricks.

The company says customers are already using the platform to speed up analysis and catch issues earlier in the workflow, though the launch materials rely heavily on product claims and customer testimonials rather than detailed public performance benchmarks.

With Sphinx 2.0, the company is betting that enterprise buyers want more than faster AI. Its pitch is that AI systems need a dependable layer of organizational memory, policy and validation before their outputs can be trusted in business settings.