# AI & machine learning use cases

> Give agents identities and connect AI outputs to inputs, actions, and human review.

Canonical page: https://www.truthlocks.com/industries/ai-ml
Category: Technology & communications

AI work becomes easier to examine when you know which agent acted and which evidence was captured. Truthlocks provides identity and action records around the AI applications you connect.

## Agent platforms: Distinguish agents and their delegated work

Several agents can share tools, making it difficult to identify responsibility for an action.

### What to record

Register separate identities and record assigned tasks, delegated permissions, tool references, and signed action receipts.

### What to review

Follow a task’s recorded actions and compare them with the access your connected tools are configured to enforce.

### The outcome

An inspectable identity and action history for each participating agent.

## Model development: Connect an evaluation report to its inputs

A model score can circulate without the dataset or evaluation configuration that produced it.

### What to record

Record model and dataset references, configuration fingerprints, the evaluation output, and the responsible reviewer’s recorded approval.

### What to review

Check the report against its captured context and independently assess whether the evaluation is appropriate for deployment.

### The outcome

A reproducible reference point for discussing a model evaluation result.

## AI-enabled products: Show where people reviewed AI-generated work

An output may move into a business process without its review history following it.

### What to record

Capture the agent identity, output fingerprint, source references, and the human acceptance, correction, or rejection event.

### What to review

Inspect the captured evidence and review decision before using the output in a consequential workflow.

### The outcome

A clear record of the transition from generated output to reviewed work.

## Before you begin

Define agent permissions, event capture, and review gates. Traceability depends on connected events and does not establish that a model’s answer is correct.

## Relevant capabilities

- [Machine identity](https://www.truthlocks.com/product/machine-identity)
- [AI workflows](https://www.truthlocks.com/product/ai-orchestration)
- [Agent workforce](https://www.truthlocks.com/product/agent-workforce)

These are practical workflow examples, not certification claims or a record of customer deployments.