Informatica Inc. kicked off its Informatica World conference in Las Vegas today with a promise to infuse generative artificial intelligence into every facet of its product line and to enable users to build natural language features into their applications.

Citing its research that found that 58% of chief data officers need five or more data management tools to manage their data estates, the company said it’s perfectly positioned to solve the data integration problems that hold back AI development.

“Our vision is to make all the new innovations in generative AI available to our data management users, whether they are data engineers, data stewards, data reliability engineers and people in the business,” said Gaurav Pathak, Informatica’s vice president of product management for AI and metadata. “Our goal is to understand the organization’s data assets, including technical data assets, business context, operational metadata and usage metadata and bring all of those together.”

Generative AI everywhere

A new version of the company’s Cloud-centric AI-backed Real-time Engine engine for data management embeds generative AI into every layer of Informatica’s Intelligent Data Management Cloud. The company said Claire GPT aggregates an organization’s metadata system of record to allow users to talk to their data.  The software can be used to find, access and interact with data without SQL, create automated procedures, support self-service data exploration and provide multistep reasoning to break down complex data management operations into more digestible tasks.

“Claire GPT is already built on the knowledge a particular organization has in Snowflake, Salesforce, SAP and other important data sources,” Pathak said. “We know the business context of those data assets, so a user who wants to do customer churn analysis can ask what data sets to use and get the answers quickly.”

The software also understands organizational variables like quarterly closing dates, so users don’t have to supply details specific to the company. “Making it organization-specific is very important to being able to answer questions,” Pathak said.

Multi-step reasoning

Informatica designed Claire GPT to incorporate multi-step reasoning, a process that breaks down complex problems or tasks into smaller, manageable steps. “If I want to create a data product for doing customer churn analysis, it can find the right sources, check the data quality, integrate and create the right data pipelines, break down that complex prompt into pieces and create that project for the end-user,” Pathak said.

The Claire copilot will also provide natural language support for data integration, application integration, data quality and data governance.

Informatica is also extending generative AI to application development with a low-code/no-code interface allowing non-technical users to create applications that incorporate generative AI quickly and without extensive programming knowledge.

The copilot supports many large language models and vector databases and includes built-in security and governance. It can chunk, embed and ingest data into vector databases to ground LLMs on an organization’s data.

Just add chat

“It’s a no-code way for customers to build gen AI apps using the same drag-and-drop tool they have been using for 20 years with added gen AI components,” said Sumeet Agrawal, vice president of product management. “If they already know Informatica technology, they don’t have to learn anything new. The same technology can be used to build gen AI apps.”

For example, an existing loan-processing application can be enhanced with a natural language interface without extensive coding, he said. “They can build these apps in minutes instead of hours or days,” Agrawal said. “Because it’s built on IDMC, all the governance is already baked in.”

Informatica said it will work with any LLM and provide an assortment of connectors for popular models. The toolset will also come with customizable templates, pre-built recipes and tools to simplify prototyping, testing and deployment of new or existing applications. It will also work with unstructured data when that support is added to IDMC.

Photo: Informatica