Tray.ai today extended the reach of its integrated platform-as-a-service IPaaS) environment to make it simpler to build, integrate and manage artificial intelligence (AI) agents.
The Tray Merlin Agent Builder extends the scope of the low-code tools Tray.ai provides to enable IT teams to compose asynchronous tasks that are completed by multiple AI agents, says Tray.ai CEO Rich Waldron.
In addition, Tray.ai is making available Tray Agent Accelerators, a set of templates for using retrieval-augmented generation (RAG) techniques to build specific classes of agents capable of processing, for example, customer service or help desk requests. Those additions to the Tray.ai extend RAG, governance and connectivity to large language models (LLMs) that the company previously added to its iPaaS).
That approach enables organizations to create any number of AI agents without having to lock themselves into any one specific application environment, says Waldron. Most workflows require access to multiple data sources to be completed so providing that capability is essential, he added.
In fact, a survey of 1,045 enterprise technology professionals at organizations with 1,000 or more employees in the U.S. published today by Tray.ai finds 42% of respondents need access to eight or more data sources to deploy AI agents successfully.
An equal percentage of respondents work for organizations that plan to build over 100 AI agent prototypes. More than two thirds (68%)are budgeting $500,000 or more annually on AI agent initiatives and nearly 90% of enterprises said they consider integration with organizational data systems essential for AI agent success.
That’s crucial because most AI agent workflows are going to be designed by business leaders that deeply understand how processes need to be completed, says Waldron. “The business has to be front and center of the change management process,” he says.
However, the survey also finds 86% of respondents work for organizations that will require upgrades to their existing IT stack in order to deploy AI agents.
It’s not clear to what degree organizations might replace their existing integration frameworks because of the rise of AI agents, but the survey notes nearly half of respondents (48%) reporting their existing iPaaS frameworks are only “somewhat ready” for AI.
Overall, the survey finds organizations will be relying on a mix of build and buy (41%), single-purpose SaaS app agents (28%) or custom development (24%) to create AI workflows.
Top use cases reported were IT service desk automation (61%), data processing/analytics (40%) and code development/testing (36%). Additionally, nearly half of respondents (49%) said their organization is prioritizing increased customer satisfaction as a key success metric.
The survey identifies cost reduction (64%) as a top priority, with 52% aiming to increase process automation rates. Less than a quarter (24%) at this stage have identified a positive revenue impact as an important measure of AI agent success.
Regardless of approach, just under two thirds for respondents (64%) want to be able to deploy AI agents in three weeks.
The one thing that is certain, however, is that no matter how long it takes to build and deploy AI agents, most organizations in a few short years might find themselves managing thousands of them.
