AI Agent Handbook

This handbook provides use cases and example prompts for the Ataccama ONE AI Agent.

For details about how the AI Agent works, its interface, and how to start conversations, see AI Agent.

What the AI Agent can do

The AI Agent covers the following areas:

Catalog discovery and search: Find catalog items, business terms, data quality (DQ) rules, and reference data tables. Also find assets by the business term assigned to them, and by how well they fit a stated purpose.
Metadata inspection: Retrieve schemas, profiling statistics, anomaly reports, and relationship mappings, along with data protection classifications and upstream sources.
Data quality analysis: Run profiling and evaluation jobs on demand, analyze the results and evaluation history, identify the weakest attributes, and propose rules that close the gaps.
Data quality rule management: Create evaluation rules from business logic, assign them to attributes, audit the library for duplicates, and delete rules that are no longer needed.
Business glossary management: Search and compare glossary terms, create new terms, and link them to the data they describe.
Documentation and governance: Generate descriptions, assign stewardship, and flag critical data elements (CDEs), all in one request if you want. Together these improve an asset’s Data Trust Index.
SQL and data exploration: Answer questions about the data itself on compatible items, and save queries as SQL catalog items.
Data transformation: Create transformation rules and transformation catalog items, and explain what existing rules and plans do.
Reference data management: Find, create, compare, and edit reference data tables, records, and schemas, and publish pending changes.
Product help: Answer questions about ONE itself and point you to the relevant documentation.

Who the AI Agent is for

Data stewards, data quality analysts, data engineers, and governance teams all benefit from the AI Agent. It eliminates repetitive navigation, reduces context-switching between screens, and compresses multi-step workflows into single conversational requests.

Making a catalog item trustworthy

The AI Agent’s individual capabilities are most useful chained together. This section covers the full path from an undocumented table to a governed, evaluated one. You can run it a step at a time, or hand over the whole sequence.

Ask what governance is missing

If you don’t already know what an asset lacks, ask for an assessment before you start.

Example prompt

Govern this table. Check what’s missing and suggest improvements

The AI Agent reports what is absent, such as no steward, undocumented attributes, no business terms, or no DQ rules, and suggests what to do about each. You can follow its suggestions, or work through the sequence yourself.

Work through the governance sequence

This is the order to work in, whether or not you started with an assessment.

Step	Example prompt
1. Ownership	Assign a stewardship group to 'CUSTOMER' based on its content
2. Documentation	Generate descriptions for 'CUSTOMER' and its attributes and assign them
3. Business meaning	Assign business terms to 'CUSTOMER' and its attributes
4. Criticality	Suggest Critical Data Elements for 'CUSTOMER', each with a category and rationale
5. Profiling	Run profiling on 'CUSTOMER'
6. Rule selection	Review the profiling results, suggest which existing DQ rules to apply, and create and apply new rules for any uncovered gaps
7. Assignment	Apply the suggested rules to the relevant attributes in 'CUSTOMER'
8. Evaluation	Run the DQ evaluation on 'CUSTOMER'
9. Outcome	What is the Data Trust Index for 'CUSTOMER'?

Step

Example prompt

1. Ownership

Assign a stewardship group to 'CUSTOMER' based on its content

2. Documentation

Generate descriptions for 'CUSTOMER' and its attributes and assign them

3. Business meaning

Assign business terms to 'CUSTOMER' and its attributes

4. Criticality

Suggest Critical Data Elements for 'CUSTOMER', each with a category and rationale

5. Profiling

Run profiling on 'CUSTOMER'

6. Rule selection

Review the profiling results, suggest which existing DQ rules to apply, and create and apply new rules for any uncovered gaps

7. Assignment

Apply the suggested rules to the relevant attributes in 'CUSTOMER'

8. Evaluation

Run the DQ evaluation on 'CUSTOMER'

9. Outcome

What is the Data Trust Index for 'CUSTOMER'?

Hand over the whole sequence in one request

You can also give the AI Agent the entire sequence in one request.

Example prompt

Assign a stewardship group, add descriptions for the asset and each attribute, assign relevant glossary terms, run a CDE check and set criticality, profile the dataset, suggest and apply DQ rules to the critical attributes, run the evaluation, and give me the Data Trust Index

The AI Agent works through the steps in order and shows you each one as it goes. Some actions pause and ask for your approval before they run.

Once the sequence finishes, Review changes lists everything the AI Agent did, so you can check its work and revert anything you’re not happy with. See Review and approve changes.

Long sequences can take several minutes to complete.

For an asset that is already partly governed, a shorter request is often enough.

Example prompt

Help me improve the Data Trust Index of this catalog item by suggesting a steward, adding a description, applying rules, and applying terms

Cataloging use cases

Cataloging use cases help you discover, explore, document, and govern assets in the Ataccama ONE catalog. The AI Agent acts as an intelligent search and documentation assistant that can traverse the catalog, inspect metadata, and update descriptions and term assignments without requiring you to navigate through multiple screens.

Catalog search and discovery

The AI Agent can search across all catalog entity types including catalog items, business terms, DQ rules, and reference data tables.

Describe what you are looking for in plain language, and the AI Agent translates the request into targeted searches, filters irrelevant results, and presents the most relevant matches.

Example prompts
Use case	Example prompt
Find catalog items	Find all catalog items related to 'customers'
Search business terms	Search for business terms related to 'money'
Find a DQ rule	What rule should I use to validate addresses? Ignore attributes, I only need the name of the rule that fits best
Find duplicate rules	Review DQ rules relevant to 'email validation' and identify duplicates
Search documentation	Search in documentation: how is data quality calculated in ONE?
Find fit-for-purpose data	Find me the highest-quality data I can use for a sales dashboard
Find data by business term	Find all catalog items tagged with the 'Customer' term
Find attributes across the catalog	Show me all attributes related to 'loan' across the catalog
Filter by owner	Show me all catalog items owned by the Finance team

Catalog item exploration

The AI Agent can summarize what a dataset contains, compare it with another, and explain how it relates to the rest of your catalog.

Example prompts
Use case	Example prompt
Get an overview of a dataset	Describe what the 'CUSTOMER' table contains
Check joinability	Can I join 'CUSTOMER' with 'INVOICE'?
Compare two datasets	Compare the differences between 'CUSTOMER' and 'CUSTOMER_MASTER'
Find ungoverned attributes	Which attributes in 'CUSTOMER' have no business term assigned?
Check applied rules	Which DQ rules are assigned to 'CUSTOMER'?
Get a full context summary	Give me a full overview of 'CUSTOMER' including its data quality and business context
Look up data classifications	What data protection classifications have been applied to 'CUSTOMER'?
Trace upstream sources	Where does the data in 'CURATED_FUND_HOLDINGS' come from?

For details about classifications, see Data Protection Classification.

Business glossary and terms

The AI Agent can search the glossary, explain the difference between terms, create new terms, and link terms to the data they describe.

Example prompts
Use case	Example prompt
Search the glossary	Search for glossary terms related to personal data
Describe a term	Generate a description for the term 'customer_id'
Link terms to attributes	Assign business terms to 'CUSTOMER' and its attributes
Create a new term	Create a new business term for 'Customer Lifetime Value' and apply it to the CLV attribute of 'CUSTOMER'
Get term suggestions before assigning	What business terms make sense to assign to 'CUSTOMER' and its attributes?
Compare two terms	What’s the difference between the 'Personal Data' and 'PII' glossary terms?
Tag a single item	Add the 'SSN' term to 'CUSTOMER'
Find data by term	Find all catalog items tagged with the 'Revenue' term

Governance and stewardship

The AI Agent can review what is missing from a catalog item’s governance, flag critical data elements, assign stewardship, and carry out a full governance pass in a single request.

Example prompts
Use case	Example prompt
Assign stewardship	Assign a stewardship group to 'CUSTOMER' based on its content
Review governance completeness	Govern this table. Check what’s missing and suggest improvements
Flag critical data elements	Suggest Critical Data Elements for 'CUSTOMER', each with a category and rationale
Document and flag criticality together	Go through the columns of 'CUSTOMER', add business-friendly descriptions, assign glossary terms, and flag the attributes that should be considered critical
Bulk stewardship and documentation	Assign stewardship and generate descriptions for 'CUSTOMER', 'INVOICE', and 'PAYMENT'
Improve the Data Trust Index	Help me improve the Data Trust Index of this catalog item by suggesting a steward, adding a description, applying rules, and applying terms
Onboard an asset end to end	Assign a stewardship group, add descriptions for the asset and each attribute, assign relevant glossary terms, set criticality, then suggest and apply DQ rules to the critical attributes

For details about critical data elements, see Define Critical Data Elements.

Descriptions and metadata

The AI Agent can write and refine descriptions for catalog items, attributes, and business terms.

Example prompts
Use case	Example prompt
Describe a catalog item	Generate a description for 'CUSTOMER'
Describe a dataset and all its attributes	Generate descriptions for 'CUSTOMER' and its attributes and assign them
Describe a business term	Generate a description for the term 'customer_id'
Improve an existing description	Update the description of 'CUSTOMER' to make it more concise

Query data values

The AI Agent can answer questions about the data itself, not only its metadata, on catalog items where querying is enabled.

Example prompts
Use case	Example prompt
Filter records	Find customers in 'CUSTOMER' whose country is not the United States
Explore a column’s values	What payment methods are used in 'PAYMENT'?
Aggregate across tables	List each customer alongside their invoice count

SQL catalog items

The AI Agent can write a query, test it against sample data, and save the result as a new catalog item.

Example prompts
Use case	Example prompt
Select a subset of columns	Create an SQL catalog item that selects the finance and customer attributes from 'CUSTOMER'
Create a filtered view	Create an SQL catalog item of non-US attendees from 'ATTENDEE'
Create an aggregation	Create an SQL query that aggregates total sales per CustomerId and save it as a new catalog item

Data quality evaluation use cases

Data quality evaluation use cases help you understand the current state of a dataset and decide what to check. The AI Agent reads profiling statistics, anomalies, and past evaluation results, explains what it finds, and proposes the rules that would close the gaps.

Profiling and evaluation jobs

The AI Agent can run profiling and data quality evaluation on demand, then use the results to suggest rules in the same conversation. You don’t have to start the job and come back to it later.

Example prompts
Use case	Example prompt
Run profiling and evaluation	Run profiling and DQ evaluation on 'CUSTOMER'
Profile, then get rules for the lowest-scoring columns	Profile 'CUSTOMER', identify its biggest DQ problems, and suggest rules for the worst performing columns

Data quality analysis

The AI Agent can read profiling results, anomalies, and evaluation history to explain the state of a dataset and where its problems are.

It works through the following steps:

Locates the target catalog item and retrieves its attribute schema.
Fetches profiling statistics, such as null counts, distinct values, patterns, and minimum and maximum ranges, for each attribute.
Retrieves profiling anomalies such as unexpected value distributions or format violations.
Pulls the latest DQ job results and Data Trust Index score.
Analyzes all findings and proposes targeted DQ rules to address the issues it identified.

Example prompts
Use case	Example prompt
Assess a dataset	What is the data quality of 'CUSTOMER'? Are there any issues in the profiling results?
Retrieve a score	What is the data quality score for 'CUSTOMER'?
Identify the lowest-scoring columns	Which columns in 'CUSTOMER' have the most quality problems?
Retrieve the last evaluation	What were the results of the last data quality run on 'CUSTOMER'?
Survey rule coverage	How many DQ rules exist in total, and which dimensions do they cover?
Find coverage gaps	Which catalog items have no DQ rules assigned?

Data quality monitoring use cases

Data quality monitoring use cases help you turn business logic into executable rules and keep the rule library healthy. The AI Agent translates plain-language requirements into rules, assigns them to the right attributes, and checks for existing rules before creating anything new.

Create and assign DQ rules

The AI Agent can find an existing rule, create new ones from business logic, assign them to attributes, and keep the rule library free of duplicates. It always searches for an existing rule before creating a new one.

When deciding what to create, it applies the following logic:

Completeness rules: Creates one generic rule per data type, for example 'String Not Empty', and applies it to all relevant attributes rather than creating duplicates.
Validity rules: Creates attribute-specific rules, for example email format or phone number format, because each attribute has unique format requirements.
Deduplication: Searches existing rules before creating anything. If a suitable rule already exists, it reuses it.

Example prompts
Use case	Example prompt
Find the right existing rule	What DQ rule should I use to validate addresses?
Suggest rules for a dataset	Suggest DQ rules for 'CUSTOMER'
Create a completeness rule	Create a DQ rule that checks for null values in the EMAIL column of 'CUSTOMER'
Create a conditional rule	Create a DQ rule that requires company_name is not empty when customer_type is 'Business'
Create several rules at once and assign them	Create the following DQ rules and assign them to 'CUSTOMER': email format, phone format, non-empty last name, valid country code, and no future birth dates
Assign an existing rule	Apply the 'Email validity' rule to relevant attributes in 'CUSTOMER'
Audit for duplicates	Review DQ rules relevant to 'email validation' and identify duplicates
Improve a dataset end to end	Review the profiling results for 'CUSTOMER', suggest which existing DQ rules to apply, and create and apply new rules for any uncovered gaps
Create a format rule	Create a DQ rule to validate that PHONE contains valid US phone numbers
Create a cross-column rule	Create a DQ rule that checks END_DATE is not earlier than START_DATE
Create a date validity rule	Create a DQ rule that checks BIRTH_DATE values are not in the future
Generate rules from a definitions table	Create DQ rules from the definitions in 'Rule Definitions' and assign them to 'CUSTOMER'
Create and apply in one step	Create a DQ rule that checks phone numbers match E.164 and apply it to 'CUSTOMER'
Delete a rule	Delete the DQ rule called 'Legacy phone check'

Data remediation use cases

Data remediation use cases help you fix and standardize data once you know where the problems are. The AI Agent can build transformation rules from a plain-language description, and manage the reference data that standardization depends on.

Create transformation rules and catalog items

The AI Agent can create transformation rules from a plain-language description, apply them to a catalog item to produce a transformation catalog item, and explain what existing rules and plans do.

Example prompts
Use case	Example prompt
Create a transformation rule	Create a transformation rule that standardizes PHONE to E.164 format
Create the rule and the catalog item together	Create a transformation rule that standardizes PHONE to E.164 format, then create a transformation catalog item from it on 'CUSTOMER'
Normalize inconsistent formats	Create a transformation rule that normalizes ORDER_DATE from DD/MM/YYYY or MM-DD-YYYY to ISO 8601, and returns NULL when the format is unrecognized
Map values to a standard set	Create a transformation rule that maps RISK_CATEGORY values L and low to LOW, M and med to MEDIUM, H and high to HIGH
Join and transform across tables	Create a transformation rule that joins 'CUSTOMER' and 'INVOICE' on CustomerId and lowercases Country
Explain an existing rule	What does the 'Standardize address' transformation rule do?
Apply an existing rule	Apply the 'Standardize address' rule to 'CUSTOMER' and create a transformation catalog item
Find transformation plans	Find all transformation plans preparing addresses
Explain a plan	Explain what the 'Address preparation' plan does
Check plan run status	Show me transformation plans that failed in their last run

For details about transformation rules and transformation catalog items, see Transformation Rules and Prepare Data with Transformation Rules.

Manage reference data tables and records

The AI Agent can find reference data tables, create them, compare their structures, onboard reference data from the catalog, and edit records and schemas.

Example prompts
Use case	Example prompt
Compare two tables	Compare the structure of 'Country Codes' and 'Region Codes'
Find reference data tables	Search for reference data tables
Create a table from scratch	Create a new reference data table called 'Currency Codes' with the attributes code, name, and symbol
Onboard reference data from the catalog	Import reference data from 'COUNTRY_LIST' into a new reference data table
Suggest types and detect formats	Suggest data types and detect format patterns for 'Currency Codes'
Edit records	Update the record for code 'USD' in 'Currency Codes'
Bulk-replace values	In 'Currency Codes', replace all empty symbol values with 'N/A'
Edit the schema	Add an attribute called is_active to 'Currency Codes' and rename name to currency_name
Apply a DQ rule to reference data	Apply a completeness DQ rule to 'Currency Codes'
Publish pending changes	Publish all pending changes to 'Currency Codes'
Describe tables in bulk	Generate descriptions for all reference data tables

Structural and record changes are made in draft, so you can review them before they are published.

For details about reference data tables and records, see Set Up Access and Governance and Work with Reference Data Records. For the review and publishing workflow, see Publish and Approve Reference Data.

Ask questions about ONE

The AI Agent can answer questions about ONE itself and point you to the relevant documentation.

Example prompts
Use case	Example prompt
Look up a concept	Search in documentation: how is data quality calculated in ONE?
Ask what the AI Agent can do	What can you do?
Ask how to do something	How do I profile a catalog item in Ataccama? Point me to the documentation

Was this page useful?