AI agents, the much-touted next phase of generative AI, have commanded enterprises’ attention. Right now, 61% of business leaders are actively adopting AI agents, according to a recent survey by my organization – with ambitious plans to scale them organization-wide.
The fixation is justified: agents can work autonomously, navigate complex workflows, learn from experience, and leverage other software as tools. They are a step change from AI that talks with you, like chatbots, to AI that works for you. The result is major productivity: Gartner estimates that by 2028, agents will automate 15% of day-to-day business decisions.
But business leaders shouldn’t mistake agents’ sophistication with omnipotence. Agents can fall into the same traps as older, less sophisticated software – including dreaded IT silos.
Vice President, watsonx Orchestrate Agent Domains, IBM.
Battling silos
For decades, IT professionals have battled silos: applications, databases, and other systems that aren’t interoperable. In the 1980s and 1990s, enterprises struggled to connect disparate applications into a single ERP solution.
Accounting, procurement, and sales workstreams were stubbornly separate – squandering coveted cross-company insights. More recently, enterprises have struggled to unify crucial customer data across disparate CRMs, and have also labored to integrate data spread across on-premise locations and multiple cloud environments.
No matter the decade or technology, the result of silos is always the same: wasted time, wasted resources, and wasted potential. When agents become trapped in silos, the outcome is no different. Their return on investment plummets, too.
We are already seeing agentic silos take shape. Enterprises are using agents with rigid divisions – one agent for sales activities, another for procurement tasks, a third for CRMs – with little connective tissue between them. What if those agents need to work together to troubleshoot a complex problem, like a sudden and unexpected shift in product demand?
If they’re siloed, they cannot pool their abilities and function as a whole greater than the sum of their parts. Not orchestrating agents is like hiring several subcontractors to build a house but restricting their tools and communication. The result is a poorly built house – or jumble of agents with poor performance.
Agents and silos
Agents can also be siloed from the technology that enterprises already have in place. Imagine an HR Agent tasked with orchestrating employee PTO – but unable to access certain calendar applications and documents.
Imagine an IT Agent tasked with troubleshooting software problems – but unable to access troves of past incident reports and help desk tickets. These agents would fail to complete their fundamental tasks, and the time and resources that went into building them would be wasted.
There is something deeply ironic about siloed agents. Agents’ value lies in their very ability to traverse the full enterprise stack, bridging tools and processes that require human time and talent. When agents get stuck, they are a victim of the very problem they are trying to solve. Businesses are investing in the problem, not the solution.
Siloed agents have an additional pitfall: they need to be governed and secured piecemeal. Relying on an ad hoc, patchwork approach to governance and security means an agentic use case is likely to fall through the cracks. If this occurs, agents’ most valuable asset – their autonomy – can quickly turn into a liability. Issues like bias, drift, and security vulnerabilities are amplified by agents’ access and independence.
Reaching potential
For agents to reach their full potential, business leaders must first fix the fragmentation underneath. Enterprises need a single data fabric that can unify the structured and unstructured data that powers agents. While many enterprises haven’t achieved this yet, a growing number understand the value: 72% of leaders view their organization’s proprietary data as key to unlocking the value of generative AI, according to my company’s most recent CEO Study.
Enterprises also need a hybrid control plane automating the sprawling landscape agents work across, unifying APIs, apps, events, files, and mainframe data. And enterprises should invest in a central nervous system for their agents. The future is multi-agent: It will be teams of agents, rather than a single agent, that tackle complex tasks. Enterprises need a single hub to supervise and route those agents. In other words, enterprises need a general contractor for all those subcontractors.
The need for orchestration
Better integrated and orchestrated agents also boost observability. Rather than governing and securing agents piecemeal, enterprises can apply comprehensive rules and oversight from a single point. This also allows AI security teams and AI governance teams to collaborate: if a shadow agent deployment is spotted by security tools, it can quickly and automatically steer the agent into the proper governance workstream.
Enterprises are rightfully investing in agents. But if they want that investment to translate into impact, they should be making equal commitments to agent integration and orchestration. Otherwise, they will end up with a whole that is less than the sum of its parts. In 2025 and beyond, it won’t just be the businesses with the best agents that win. It will be the businesses with the most flexible ones.
This article was produced as part of TechRadarPro’s Expert Insights channel where we feature the best and brightest minds in the technology industry today. The views expressed here are those of the author and are not necessarily those of TechRadarPro or Future plc. If you are interested in contributing find out more here: https://www.techradar.com/news/submit-your-story-to-techradar-pro
Services Marketplace – Listings, Bookings & Reviews