Laying the Foundation for Agentic AI with Agentic AI

Boomi’s Mani Gill offers commentary on laying the foundation for agentic AI with agentic AI. This article originally appeared in Insight Jam, an enterprise IT community that enables human conversation on AI.

AI agents are approaching ubiquity in the business world. In fact, according to the International Data Corporation, agentic AI spending is set to reach $1.3 trillion by 2029. As businesses look to lay the foundation for this transformative technology, some organizations are finding that their current IT infrastructure is prohibitive to agentic AI success.

AI agents need data that is clean, connected, and accessible, thereby allowing agentic systems to complete tasks with the proper context. Most enterprises are still wrestling with data that’s messy, fragmented, or simply unreliable. To make data truly ready for Agentic AI, organizations need systems that can seamlessly handle structured, semi-structured, and unstructured data—while eliminating silos, aligning teams and tools, and infusing agentic AI throughout the entire data lifecycle.

 Identifying Problematic Data and Silos

Organizations often have the data they need to make agentic AI useful. However, that data is often unstructured and, therefore, more challenging for AI to process. Data also frequently ends up in silos, largely because the teams and tools that manage it operate in isolation from one another.

For example, the sales team, finance team, and operations team may not share data. Simultaneously, the organizations may use different apps for each team. Historically, organizations have also created silos between data and application teams, which can also present hurdles to agentic AI.

For years, data teams and application teams worked in parallel with no crossover. Data teams are responsible for data pipelines and workloads, while app integration teams are responsible for integrating applications to streamline business processes and workflows.

For Agentic AI to drive the autonomous processes that will free up IT teams to focus on more strategic initiatives, it must be able to access and understand both reliable data and business processes. When data teams and application integration teams are separated, it takes longer for AI agents to acquire the data necessary to complete tasks and understand how to execute those tasks. Organizations with siloed data and app integration teams may also risk agentic AI that attempts to complete tasks without the full context.

Setting the stage for Agentic AI

To pave the way for successful agentic AI systems, organizations must first begin by identifying where AI agents will be most useful. Sit with your team and do a time audit, figuring out where they are losing the most time on manual tasks. Then, identify which data in your organization is most necessary to complete these tasks. To make AI agents most useful, it will be important to turn unstructured data into structured data. Organizations can use AI agents to create the structured data necessary to drive further agentic implementation. For example, agents can take all the sales from a particular month and put them into a graph, making it easier for agents to reason over.

Businesses also need to create frameworks in which both teams — and the tools they use — are unified, making it easier for AI agents to operate with the proper context. To do this, organizations must identify where they plan to use agentic AI systems. This will make it easier to determine where teams need to be unified and to which applications/data AI agents must have access. For example, if an organization wants to implement an agent that can summarize budgets, the agent may need to access the data, applications, and workflows from both the finance and sales departments.

It’s also important to note that, in any agentic AI implementation, the data and application integration teams should work together. This includes, from a technology perspective, through platforms that can integrate both teams’ workflows, and a personnel perspective.

From a technology perspective, AI agents need access to both reliable and context-rich data (which is usually the remit of the data team) as well as a deep understanding of business processes (which is usually the prerogative of the application integration team). When organizations combine these efforts, it removes silos that may result in outputs based on fragmented data. In other words, businesses need to leverage platforms that can integrate with both teams’ workflows and create unified, visible data pipelines to each AI agent.

From a personnel standpoint, a combined data and application integration team will require the grouping of roles that haven’t historically worked together. This includes data engineers, integration architects, automation specialists, and AI/ML engineers. When these teams come together, they can oversee the aforementioned unified pipelines, ensure agent outputs remain relevant and valuable, and intervene in agent workflows if necessary.

Once AI agents are up and running, it’s necessary to measure their effectiveness through adoption rate rather than capability. In other words, it’s much more important for an agent to be simple and remove 90% of manual tasks than be sophisticated and only remove 10% of manual tasks. The ultimate goal is for agents to completely remove manual tasks and give valuable time back to your teams.

Preparing for the Future with Agentic AI

AI agents will soon become a business imperative for almost any enterprise. Those who inefficiently leverage this technology will find themselves bogged down in manual tasks, spending less time on innovation, and lagging behind their competition. However, when businesses can clean their data, break down silos, and successfully integrate their teams and tools, they pave the way for agentic AI systems to impact their company at scale. As a result, they’ll be prepared to not only rise above their competitors but also deliver services that transform the lives of their clients.