Traceability

In enterprise environments, AI cannot operate as a “black box”. Users and administrators must be able to verify exactly how an agent arrived at a specific conclusion and what data it used to generate its response.

DataQI supports strong traceability across agent interactions, allowing actions, decisions, and data retrievals to be audited.

Understanding where answers come from

DataQI focuses on data provenance. When an agent answers a question or generates a report based on enterprise data, it provides explicit Context.

Rather than trying to guess after the fact, the platform transparently informs you exactly what information the AI looked at to generate the answer. This context includes the specific documents sent, the connections used, and the agents and tools involved in providing that conversation turn response. If an agent cannot find relevant context, it is instructed to explicitly state that it does not know the answer, rather than attempting to guess a response.

Auditing and event history

To support security and compliance, DataQI securely records the context of interactions in a dedicated audit database.

flowchart LR
    Interaction([Agent Interaction]) --> Context{Context Captured}
    Context -->|Who, What, Which| Audit[(Audit Database)]
    Context --> AgentFeed[Agent Feed]

This creates a verifiable trail of:

• Who asked the question or triggered the workflow.
• What exactly was asked.
• Which specific documents, files, or MCP tools the AI accessed to formulate the response.

Agent feed traceability

While audit logs capture the high-level “who and what”, the Agent feed captures the granular “how” for background operations.

The Agent feed allows for a chronological view of autonomous agent actions in two ways:

• By workflow run: Seeing exactly what happened in relation to a specific triggered event.
• By environment timing: Seeing what happened across the environment at a specific point in time.

Within these views, the feed provides deep, step-by-step visibility into the agent’s internal decision-making process, including:

• Internal thought processes and troubleshooting ideas.
• Which tools it used and what data it found.
• When it asked other agents for help and how it broke down tasks.
• All artefacts generated during the workflow run.

This observability demystifies the AI’s operations, allowing administrators to troubleshoot failed workflows and build trust in the platform’s autonomous capabilities.