By admin 
After a whirlwind of two years characterized by dazzling AI demonstrations, hastily developed agent prototypes, and overly optimistic predictions, enterprise technology leaders in 2026 are adopting a decidedly more pragmatic approach. The cutting edge of artificial intelligence within large organizations is no longer solely about showcasing novel capabilities; instead, it’s increasingly focused on the foundational, yet critical, aspects of governance, orchestration, and iterative development, alongside the intricate process of integrating AI agents into the complex, decades-old systems that form the backbone of modern enterprises. This shift in focus was a central theme during a recent webinar hosted by OutSystems, where a distinguished panel of software executives and seasoned enterprise practitioners converged to articulate this evolving landscape.
The prevailing sentiment among enterprise leaders today is a strong emphasis on fundamental operational efficiencies and tangible business outcomes. The primary objective is to leverage new AI technologies not for abstract innovation, but as potent accelerators for productivity gains, enhancements in software delivery pipelines, and the generation of measurable, bottom-line business results. This pragmatic orientation is shaped by three interconnected elements: the imperative to manage and mitigate the inherent risks of AI, the necessity of robust orchestration to harness the collective power of AI agents, and the economic realities of deploying and scaling AI solutions within existing enterprise infrastructures. Against this backdrop, the panel delved into the intricacies of governance frameworks, the nuanced economics of enterprise AI investment, and the inherent limitations of large language models (LLMs) when deployed without sophisticated orchestration. Ultimately, the conversation converged on the innovative ways leading organizations are architecting and implementing multi-agent systems, critically grounding these systems in the rich tapestry of existing enterprise data and established workflows.
Agents in the Real World: From Standalone Assistants to Coordinated Teams
The successful deployment of AI agents in production environments across the enterprise is significantly streamlined by the adoption of a unified platform capable of managing the entire lifecycle from development and iteration to deployment. This is precisely where the value of capabilities like the Agent Workbench within the OutSystems platform becomes paramount, according to Rajkiran Vajreshwari, Senior Manager of App Development at Thermo Fisher Scientific. He explained that such a platform provides the essential infrastructure for learning, iterating, and governing AI agents at scale, a critical factor for large-scale deployments.
At Thermo Fisher Scientific, Vajreshwari’s team has transitioned away from the simpler model of single-task AI assistants, particularly in customer service, towards the development of a coordinated ecosystem of specialized agents. This sophisticated approach, facilitated by the Agent Workbench, allows for a dynamic and intelligent handling of support cases. Upon the arrival of a new support request, a dedicated triage assistant meticulously classifies the inquiry. This classification then triggers a dynamic routing mechanism, directing the case to the most appropriate specialist agent. These specialist agents are designed for specific functions, such as an intent and priority agent that further refines the issue, a product context agent that gathers relevant product information, a troubleshooting agent equipped to diagnose and resolve technical problems, or a compliance agent that ensures adherence to regulatory requirements.
"We don’t have to think about what will work and how. It’s all pre-built," Vajreshwari elaborated, highlighting the efficiency gains. "Each agent has a narrow role and clear guardrails. They stay accurate and auditable," he added, underscoring the importance of reliability and transparency in enterprise AI deployments. This structured approach minimizes the potential for errors and ensures that each AI agent operates within well-defined boundaries, contributing to a more predictable and controllable AI ecosystem.
Governing the Risks of Shadow AI: Establishing Guardrails for Enterprise-Scale AI
A significant and emerging category of risk arises when AI technologies empower individuals within an organization to generate production-level code without the traditional oversight of IT departments. This phenomenon, often termed "shadow AI," presents a fertile ground for a host of potential problems. These homegrown AI solutions are particularly susceptible to issues such as hallucinations (generating incorrect or fabricated information), data leakage, violations of corporate policies, model drift (where the AI’s performance degrades over time), and agents executing actions that have never been formally approved or understood by the organization.
To proactively address and mitigate these risks, leading organizations are implementing a three-pronged strategy, as outlined by Luis Blando, CPTO of OutSystems. "Give users guardrails. They’re going to use AI whether you like it or not," Blando asserted, emphasizing the inevitability of AI adoption within enterprises. "Companies that seem to be getting ahead are using AI to govern AI across their full portfolio," he continued, suggesting a meta-level approach to AI management. "That is the difference between shadow AI chaos and enterprise-grade scale," he concluded, drawing a stark contrast between uncontrolled AI adoption and a strategically managed, scalable AI implementation.
Eric Kavanagh, CEO of The Bloor Group, further elaborated on the concept of governance, describing it as a layered discipline. This includes robust data security measures, continuous monitoring of AI models for performance degradation or "drift," and making deliberate, strategic choices about where and how AI integrates with existing business processes. Kavanagh also pointed out that the burden of creating these controls does not have to fall entirely on manual efforts. "Companies don’t have to be manually creating these controls," he stated. "A lot of those guardrails and levers are baked into platforms like OutSystems," he added, suggesting that integrated platforms can provide inherent governance capabilities, simplifying the compliance and security overhead for organizations.
Why the Real Orchestration Challenge is Models vs. Platforms
Much of the initial enthusiasm surrounding enterprise AI was centered on the selection of the most advanced large language model. However, the more complex, and ultimately more enduring, challenge and source of value lies in orchestration. This encompasses the sophisticated coordination of tasks, the seamless routing of workflows, the meticulous governance of AI execution, and the seamless integration of AI capabilities into existing enterprise systems.
Scott Finkle, VP of Development at McConkey Auction Group, provided a crucial perspective, noting that LLMs, while impressive in their own right, are merely components within larger, intricate workflows, not complete solutions. He stressed the importance of building systems that are flexible enough to "hot-swap" between different LLMs – such as Gemini, ChatGPT, Claude, and future emergent models – without necessitating a complete rebuild of the agentic system. This adaptability is crucial in a rapidly evolving AI landscape.
A platform equipped with robust orchestration capabilities is essential for achieving this flexibility. Such platforms manage the AI agent lifecycle, provide critical visibility into operations, and ensure that processes execute reliably, even as AI handles the complex reasoning layer that sits atop these workflows. "The AI and the models change, the workflows can change, but the orchestration remains the same," Finkle emphasized, highlighting the stability and longevity of a well-designed orchestration layer. "That’s how we’re going to extract value out of AI," he concluded, pointing to orchestration as the key to sustained AI-driven business value.
The Economics of Enterprise AI Investing: Focusing on Incremental Wins
In 2026, investment in enterprise AI will increasingly prioritize security, compliance, governance, and platform-level AI capabilities, particularly as AI permeates core business functions such as finance and supply chain management. The prevailing economic philosophy favors an approach of achieving incremental wins rather than expecting massive, immediate returns. This strategy acknowledges the complexity of enterprise systems and the iterative nature of AI integration.
"We’re focusing on base hits," Finkle stated, employing a baseball analogy to illustrate the preferred approach. "The way it counts is by getting something into production and having it make an impact. Big investments in pilot projects that don’t make it into production don’t save any money," he explained, underscoring the importance of deployment and tangible results. "It’s not going to happen overnight, but over time I think we’ll see tremendous savings," he added, expressing confidence in the long-term economic benefits of a pragmatic AI investment strategy.
There remains a discernible divergence in how enterprises are approaching AI transformation. Some organizations are opting for a complete reimagining of every process, starting from scratch. Others, particularly those with substantial investments in existing, depreciating infrastructure, are keen on integrating AI directly with their current systems. Their objective is to leverage agentic systems that can reuse existing data, APIs, and proven processes, thereby accelerating delivery timelines. The agent platform approach, as exemplified by OutSystems, is well-suited to serve both these camps, but it particularly resonates with the latter. This allows organizations to strategically deploy agents where they offer clear, immediate value, all while preserving the integrity and reliability of their established, deterministic workflows.
The Rise of the Enterprise Architect and the Generalist Developer
As AI technologies accelerate the pace of code generation, traditional bottlenecks in software delivery are gradually dissolving. In their place, there is a burgeoning premium on systems thinking. This refers to the crucial ability to comprehend the broader enterprise architecture, to skillfully decompose complex business problems into manageable components, and to reason effectively about how AI solutions integrate with and enhance existing infrastructure. Kavanagh specifically identified enterprise architects as the professionals uniquely positioned to capitalize on this transformative era.
"We’re entering a very interesting age of the generalist," Kavanagh explained. "The better you know your enterprise architecture and your business architecture and how those things align, the better off you’re going to be," he elaborated, emphasizing the value of holistic understanding. This broad perspective enables them to identify strategic opportunities for AI integration that can drive significant business value.
The ultimate outcome of this shift, as Kavanaugh noted, is a demonstrable improvement in the speed and quality of software delivery. "The result is faster delivery with fewer interruptions and fewer bugs," Kavanaugh stated. "You can focus on the non-repetitive tasks. It’s a benefit to the developer, to the business, and to the whole IT organization," he concluded, highlighting the widespread positive impact of embracing AI-driven development and a systems-thinking approach.
For those interested in exploring these critical aspects of enterprise AI further, the entire webinar, "AI Predictions: The Agentic Enterprise," is available for viewing.
Catch the entire webinar here.
Sponsored articles are content produced by a company that is either paying for the post or has a business relationship with VentureBeat, and they’re always clearly marked. For more information, contact [email protected].