Profound Brings GenAI Tech To IBM i Apps with Profound AI

June 24, 2024 Alex Woodie

For more than 20 years, Profound Logic has been at the cutting edge of bringing the latest Web, mobile, and API technology to IBM i shops. With the launch of Profound AI, it’s now helping IBM i shops adopt the latest in user interface tech: natural language and generative AI.

GenAI has taken the world by storm since OpenAI launched ChatGPT in late 2022. Since then, nearly every large enterprise has explored the potential for using large language models (LLMs) to develop a range of GenAI products, including chatbots, autonomous agents, question-and-answering systems, and knowledge management apps.

Despite the potential for GenAI to create up to $4.4 trillion in annual global economic growth, according to McKinsey, the technology is proving hard to master. For instance, a survey at the end of 2023 by cnvrg.io (a subsidiary of Intel) found just 10 percent of companies had launched GenAI apps into production in 2023, and other surveys have found similarly dismal numbers.

While there are a variety of reasons for the low deployment rate – including concerns over privacy, security, governance, and LLMs’ tendency to “hallucinate,” or make things up – the newness of the technology and the lack of familiarity that users have with it is providing to be a big stumbling block, too.

Profound AI enables customers to develop GenAI apps that can analyze database records and even create graphs.

Lowering the technical barrier to GenAI is the big driver behind Profound Logic’s creation of Profound AI. The Newport Beach, California company introduced Profound AI in late 2023 to help jumpstart customers’ GenAI projects, and has been in beta tests with early adopters ever since. Those tests are wrapping up, and the product is expected to become generally available soon, perhaps before July, according to Profound Logic CEO Alex Roytman.

“The idea behind the solution is it helps you integrate AI into your applications, to make it very easy to do, without having to have a lot of knowledge about underlying models,” Roytman told IT Jungle at the recent COMMON POWERUp conference in Fort Worth, Texas.

Profound Logic has done the work of building an abstraction layer in front of a handful of the most popular commercial and open source LLMs, including those from Open AI, Google, Anthropic, Meta, and others. Profound AI functions basically like a framework that makes it easier for customers to build GenAI interfaces atop the LLMs, integrate them into their existing applications, and feed the LLMs with information stored in their databases, including Db2 for i.

“We support pretty much anything that’s out there,” Roytman said of the LLM providers. “That’s part of the big work that we did. Every model, every vendor, has their own API unfortunately. There’s no standardization, at least not yet, so we’ve figured out how to standardize it all, and we just kind of plug it into the solution.”

Customers can use Profound AI to create automated agents that follow workflows.

Accessed from Profound’s integrated development environment (IDE), Profound AI guides users through the process of building a variety of GenAI apps. Customers choose which LLM they want to use from a dropdown menu, define their data sources, and then instruct the product what to build, such as a chatbot, an agent, or a knowledge management system.

GenAI apps developed with Profound AI can handle a range of tasks. For instance, they can answer questions about information housed in the database, like what is the revenue for region Y for the past 14 months. You can even ask the agent to create a graph based on the data that’s returned from the SQL query, and it will give you something that looks like what you could create in Excel.

Profound AI can even handle more complex analysis tasks, Profound AI said. “Imagine an AI agent capable of creating insightful charts and graphs or dissecting data with precision just from natural language instructions,” the company said in its online documentation for Profound AI.

The software can also create agents that can handle basic tasks, such as order processing and data retrieval activities, all from a natural language interface. Profound AI also includes a low-code interface that supports the development of agents for more complex software engineering tasks, such as calling RPG or COBOL applications, using service programs, or accessing Web services, the company said.

There is also a Knowledge Documents feature that helps customers develop GenAI apps designed to return specific data based on information stored in detailed product specifications, historical data, or procedural manuals. This can be helpful for customer support and internal operations, the company said.

While a pre-trained LLM has the capability to generate human-like answers in response to inputs or questions, the magic of Profound AI is how the software understands a customers’ data and crafts the prompt that is then sent to the LLM as the input, Roytman said.

“The models we’re working with are smart enough to run an SQL statement,” he said. “So when a customer asks a question, we behind the scenes engineer a prompt that says, OK we have this database layout and this question. What SQL statement would retrieve the data that you need to answer the question?”

Security is baked into Profound AI, Roytman added. That means that not only does the GenAI agent need to be granted authorization to access data, but it’s smart enough to identify so-called “prompt injection” attacks designed to trick LLMs into giving up sensitive information.

“We would block that,” he said. “By default, we actually can’t see any data until you start saying ‘This agent can access this data.’”

Profound AI was developed in Node.js and can run on IBM i or on any other system that can run Node.js, whether on prem or in the cloud. Customers have their choice of LLMs, including commercial offerings in the cloud like OpenAI’s GPT-4, currently considered the gold standard for fast and accurate generation of natural English language.

But customers can also run their own open source LLM, such as Meta’s Llama-3 70B, on prem if they like. Profound AI works with pre-trained LLMs or, if customers have the time, inclination, and money to fine-tune their own model, the software can work with that, too, Roytman said.

The possibilities of GenAI are wide and getting wider by the day. Organizations of all sizes are exploring GenAI to see how it can help them run better. Luckily for IBM i shops, the technology just needs access to good, clean data to work effectively, which is something that IBM i shops have as much of as companies that run on other computer systems.

For more information on Profound AI, see Profound Logic’s documentation for the product.

RELATED STORIES

Profound Logic Explores AI Paths for IBM i

Generative AI Is Part Of Application Modernization Now

Profound Bolsters Security, Authentication in Node.js Dev Tools