# AI data readiness worksheet

Assess data for a specific use case. A large archive is not automatically usable: you need representative examples, clear meanings and appropriate access to the relevant information.

Revision: 2026-09-24

## When to use it

Before estimating an AI project or approving a trial involving documents, CRM records, support tickets or internal knowledge.

## Expected output

A source inventory with observed defects, remediation actions and an approved sample for the trial.

## 1. Identify the sources

Start with information needed for the intended decision or action. Exclude irrelevant archives and record who understands the meaning of each source.

### Source, format and owner

Record the system, operational owner, available format and stable record identifier.

________________

### Access and intended use

Describe who may access the data, for which activity and who authorises its use in this project.

________________

### Updates and retention

Record update frequency, obsolete information and internal removal rules that need confirmation.

________________

> Illustrative example, replace with your own: Illustrative example: technical procedures in a shared folder, reviewed quarterly by the support owner. Personal copies and unapproved drafts remain outside the trial.

## 2. Inspect a sample

Select common and difficult cases. Record the selection method: a sample containing only clean files hides the conditions encountered in actual work.

### Sample and coverage

Describe period, size and variety: languages, formats, teams, rare cases and incomplete documents.

________________

### Observed defects

Count missing fields, duplicates, conflicting versions and records that cannot be matched.

________________

### Meaning and correct reference

Identify who can explain fields and establish the expected answer when information is ambiguous.

________________

> Illustrative example, replace with your own: Illustrative example: 40 documents cover digital PDFs, scans and tables. Five have conflicting versions: identify the authoritative version before constructing the evaluation set.

## 3. Plan the preparation

Separate issues that prevent a trial from manageable limitations. Assign concrete actions without waiting to clean every record in the business.

### Remediation, owner and date

For each gap state what changes, who acts and how completion will be checked.

________________

### Excluded or restricted data

Define categories, permissions and cases that must stay outside the initial trial.

________________

### Entry criteria

Specify which sources, labelled examples and authorisations must exist before the trial starts.

________________

> Illustrative example, replace with your own: Illustrative example: begin with one approved document family. Unreadable scans go to manual processing and are counted separately when results are reported.

## Pitfalls to avoid

- Confusing technical access to a file with permission to use it.

- Assessing quality only on easy cases.

- Launching an assistant with conflicting versions and no authoritative source.

Stolen Orbit — https://stolenorbit.com/en/templates/ai-data-readiness/