AI Agent Handbook

This handbook provides use cases, example prompts, and time-savings estimates for the Ataccama ONE AI Agent.

For details on the AI Agent interface and how to start conversations, see AI Agent.

How AI Agent works

The AI Agent understands natural language instructions and translates them into structured action plans. By chaining together over 50 tools, it executes complex, multi-step tasks that would otherwise require significant manual effort across multiple platform screens.

Each task is carried out with built-in deduplication checks, contextual awareness, and confirmation steps to ensure accuracy.

Core capabilities

Catalog discovery and search: Find catalog items, business terms, data quality (DQ) rules, reference data tables, and transformation artifacts by keyword or natural language description.
Metadata inspection: Retrieve attribute schemas, profiling statistics, anomaly reports, DQ scores, Data Trust Index values, and relationship mappings for any catalog item.
Data quality analysis: Analyze profiling results, identify anomalies, propose rules to address issues, and fetch DQ job scores.
Data quality rule management: Search for existing rules, create new evaluation rules, assign rules to catalog item or reference data attributes, and deduplicate before creation.
Reference data management: Create, rename, describe, and restructure reference data tables.

Add, edit, or delete records. Import data from catalog items.
Documentation and governance: Generate or improve descriptions for catalog items, terms, rules, and attributes.

Assign stewardship. Link glossary terms to items and attributes.
SQL and data exploration: Run SQL queries, create SQL catalog items from queries, sample data, and find specific records matching criteria.
Transformation management: Search, describe, generate, publish, run, schedule, and manage the lifecycle of transformation plans and rules.

Who benefits

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.

Cataloging use cases

Cataloging use cases cover the discovery, exploration, documentation, and governance of assets within 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 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?

Catalog item exploration

The AI Agent retrieves detailed metadata about catalog items including their attributes, assigned rules, linked terms, relationships to other items, profiling statistics, and Data Trust Index scores. This allows you to quickly understand the structure and governance state of any data asset.

Example prompts
Use case	Example prompt
Describe a table	Describe what the <Catalog item> table contains
Analyze relationships	Can I join <Catalog item> with <Catalog item>?
Compare catalog items	Compare the differences between <Catalog item> and <Catalog item>
Find specific data	Find customers in <Catalog item> whose country is not the United States
List ungoverned attributes	Give me a list of attributes from <Catalog item>, <Catalog item>, and <Catalog item> that don’t have any DQ rules applied
Check attribute rules	What DQ rules are applied to the 'email' attribute in <Catalog item>?

Documentation and governance

The Agent can generate AI-powered descriptions for catalog items, business terms, rules, and attributes. It can also assign stewardship to user groups and link glossary terms to catalog items and their attributes, streamlining the governance process.

Example prompts
Use case	Example prompt
Generate description	Generate a description for the term 'customer_id'
Assign stewardship	Assign an appropriate steward to <Rule>
Link terms	Assign the term 'Customer Identifier' to the customer_id attribute in <Catalog item>

SQL catalog item creation

When you need a filtered or transformed view of an existing catalog item, the Agent can construct SQL queries, test them against sample data, and create new SQL catalog items. This is particularly useful for creating filtered subsets or joining data from multiple sources.

Example prompts
Use case	Example prompt
Create filtered view	Create a SQL catalog item of non-US based attendees of the conference from <Catalog item>
Select specific columns	Create a SQL catalog item that selects the finance and customers attributes from the <Catalog item> table

Potential time savings
Task	Manual time	Agent time	Savings
Find and review a catalog item	5-10 min	30 sec	~90%
Search and compare DQ rules	15-20 min	1-2 min	~85%
Generate descriptions for 10 items	30-60 min	3-5 min	~90%
Create SQL catalog item with testing	15-25 min	2-3 min	~85%
Assign stewardship across items	10-15 min	1-2 min	~85%
Identify ungoverned attributes across three tables	20-30 min	1-2 min	~93%

Data quality evaluation use cases

Data quality evaluation use cases focus on assessing the current state of data quality for catalog items.

The Agent inspects profiling statistics, anomaly reports, and existing DQ job results to surface issues and recommend improvements. This is the analytical side of data quality management, helping teams understand where problems exist before taking corrective action.

Data quality analysis

The Agent performs a comprehensive data quality analysis by fetching attribute profiling statistics, identifying anomalies in record patterns, and retrieving DQ scores from the latest evaluation jobs. It synthesizes these findings into an actionable assessment, highlighting attributes with completeness issues, format inconsistencies, or unusual distributions.

How it works:

The Agent locates the target catalog item and retrieves its attribute schema.
It fetches profiling statistics (null counts, distinct values, patterns, min/max ranges) for each attribute.
It retrieves profiling anomalies such as unexpected value distributions or format violations.
It pulls the latest DQ job results and Data Trust Index score.
It analyzes all findings and proposes targeted DQ rules to address identified issues.

Example prompts
Use case	Example prompt
Full DQ analysis	Analyze the data quality of the catalog item 'CUSTOMER'
Improve data quality	Improve the data quality of <Catalog item> by suggesting existing rules or creating new ones
Review DQ state	What is the current data quality score for <Catalog item>?

Rule suggestion and evaluation

The AI Agent can suggest appropriate data quality rules for catalog item attributes based on profiling data and attribute types. It distinguishes between generic completeness checks (which should be reusable across attributes of the same type) and validity checks (which are attribute-specific based on format patterns).

Before suggesting any new rules, it always searches for existing rules to avoid duplication.

The Agent follows this decision logic:

Completeness rules: The Agent 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: The Agent creates attribute-specific rules (for example, email format, phone number format) because each attribute has unique format requirements.
Deduplication: Before creating any rule, the Agent searches existing rules. If a suitable rule already exists, it reuses it.

Example prompts
Use case	Example prompt
Suggest rules for all attributes	Suggest data quality rules for all attributes in the catalog item 'CUSTOMER'
Suggest rules with profiling	Please look at this table and then suggest or attach any relevant DQ rules to the attributes that do not have a rule assigned
Create completeness checks	Create completeness checks for five string attributes in <Catalog item>
Create validation rules	Create validation rules for the email and phone_number attributes in <Catalog item>

Potential time savings
Task	Manual time	Agent time	Savings
Full DQ analysis of a table	30-60 min	3-5 min	~90%
Suggest rules for 10 attributes	20-40 min	2-4 min	~88%
Create and assign completeness checks	15-25 min	1-2 min	~90%
Identify anomalies in profiling data	15-20 min	1-2 min	~90%

Data quality monitoring use cases

Data quality monitoring use cases focus on the ongoing creation, assignment, and management of DQ rules that continuously monitor data assets. The Agent helps teams build and maintain a comprehensive rule library, assign rules to the right attributes, and ensure consistent monitoring coverage across the catalog.

Data quality rule creation

The Agent creates new data quality evaluation rules based on natural language descriptions of the business logic. It translates requirements like 'company_name must not be empty when customer_type is Business' into executable rule definitions.

Before creating any rule, it performs a mandatory deduplication search to check for existing rules that already satisfy the requirement.

Example prompts
Use case	Example prompt
Create a single rule	Create a DQ rule that requires company_name attribute is not empty when customer_type attribute is 'Business'
Create a validation rule	Create a rule that checks if an email is valid
Create multiple rules	Create 2 DQ rules for attributes of <Catalog item>. The first ensures that when customer_type is 'Business', company_name is not empty. The second ensures that when employment_status is 'Terminated', the termination_date is present.
Create from data sample	Take a look at the <Catalog item> table and please create and apply data quality rules to each attribute

DQ rule assignment

Once rules exist, the Agent assigns them to the appropriate catalog item attributes or reference data table attributes. It can process bulk assignments efficiently, applying a single reusable rule to many attributes in one workflow.

Example prompts
Use case	Example prompt
Apply existing rules	Take a look at the <Catalog item> table and apply appropriate existing data quality rules to each attribute, if you can’t find any appropriate rules for an attribute then create and apply new ones
Apply rule to RDM	Apply <DQ rule> to the 'code' attribute in <Reference data table>
Apply rule to catalog items	Apply <DQ rule> to the email attribute in <Catalog item>

Rule library maintenance

The Agent helps maintain a clean, deduplicated rule library by identifying duplicate or overlapping rules, renaming rules for clarity, and generating descriptions that explain what each rule does. This ensures the rule library remains organized and reusable across teams.

Example prompts
Use case	Example prompt
Find duplicates	Review DQ rules relevant to 'email validation' and identify duplicates
Find a best-fit rule	What rule should I use to validate addresses? Ignore attributes, I only need the name of the rule that fits best
Create rules from catalog item	Create rules that are defined in rows of a catalog item 'Data Quality Rules'

Potential time savings
Task	Manual time	Agent time	Savings
Create a conditional DQ rule	10-20 min	1-2 min	~88%
Assign rules to 10 attributes	15-25 min	2-3 min	~85%
Audit rule library for duplicates	30-60 min	3-5 min	~90%
Create rules from a rule definition table	60-120 min	5-10 min	~90%
Full rule creation and assignment for a table	30-45 min	3-5 min	~88%

Data remediation use cases

Data remediation use cases cover the correction, transformation, and management of data within reference data tables as well as the creation of transformation artifacts. The Agent provides a conversational interface for performing data edits, managing table structures, importing data, and orchestrating transformation workflows.

Reference data table management

The Agent can create new reference data tables, add or remove attributes, rename tables and attributes, change data types, create foreign key connections between tables, and generate AI-powered descriptions.

All structural changes are made in draft mode, giving you the opportunity to review before publishing.

Example prompts
Use case	Example prompt
Create a new reference data table	Create a new reference data table called 'Currency Codes' with attributes for code, name, and symbol
Add an attribute	Add a new attribute called 'effective_date' with type Date to <Reference data table>
Rename a table	Rename <Reference data table> to 'Country Master Data'
Generate description	Generate a description for <Reference data table> based on its structure and data
Compare structures	Compare the structure of <Reference data table> and <Reference data table>

Reference data editing

The Agent can add, update, and delete records within reference data tables using natural language instructions.

It translates your requests into SQL-based updates or direct record operations, applies them in draft mode, and surfaces the changes for review. This is especially powerful for bulk edits that would be tedious to manually perform row by row.

Example prompts
Use case	Example prompt
Bulk edit data	In the reference data table <Reference data table> and its Code attribute, replace empty values with 'N/A'
Capitalize values	Edit the reference data table 'CUSTOMER', change the First_Name attribute to always be capitalized
Delete records	Delete all records from <Reference data table> where status is false
Add records	Add the following records to <Reference data table>: US - United States, CA - Canada, MX - Mexico

Data import

The Agent can import data from catalog items into reference data tables, either by creating a new reference data table or importing into an existing one.

The auto-import capability creates a transformation plan that handles the data movement, making it easy to seed reference data tables from authoritative catalog sources.

Example prompts
Use case	Example prompt
Import into new table	Create new reference data table named 'Table - Import' and import <Catalog item> into it
Import into existing table	Import <Catalog item> into <Reference data table>

Transformation management

The Agent can search for, describe, generate, publish, run, and schedule transformation plans and rules. It provides a conversational interface for managing the lifecycle of transformation artifacts, from initial creation through to scheduled execution.

The following capabilities are available:

Search and describe: Find existing transformation plans and rules, generate AI-powered descriptions of what they do.
Generate: Create new transformation rules from a natural language prompt describing the desired logic.
Lifecycle management: Publish, discard, delete, or confirm deletion of transformation plans and rules.
Execution: Trigger one-time runs of published standalone plans.
Scheduling: Create, modify, pause, resume, or delete schedules for standalone plans.

Potential time savings
Task	Manual time	Agent time	Savings
Create RDM table with five attributes	10-15 min	1-2 min	~87%
Bulk edit 100+ records	20-40 min	1-2 min	~95%
Import catalog item into RDM	10-20 min	1-2 min	~88%
Add records manually (five rows)	5-10 min	30 sec	~90%
Compare two RDM table structures	10-15 min	1 min	~90%
Generate and publish a transformation rule	15-30 min	2-3 min	~85%
Configure a plan schedule	5-10 min	30 sec	~90%

Was this page useful?