by Ed Macosky, Chief Product and Technology Officer, Boomi
Q1: Everyone is investing in AI so why are so many organizations still struggling to see real results?
2026 is the year organizations need to move from AI experimentation to activation at scale. And yet, so many are still stuck in pilot purgatory, not because their AI tools are not capable, but because the foundation underneath them is not ready. The challenge is not building agents. It’s giving them the data, memory, and governance they need to operate in real enterprise environments.
What I see consistently is that organizations have layered agents onto a fragmented landscape with disconnected tools, siloed data, and no unified governance. And when you do that, you don’t get results. You get magnified governance gaps and operational risk. AI amplifies the chaos underneath it rather than creating order.
The enterprises breaking through are the ones that recognized this is a moment of inflection. Every enterprise transformation has one. For agentic AI, that moment is now, and the organizations that are winning are building on an active data foundation, not bolting AI onto infrastructure that was never designed for it.
Q2: Do you think integration is the missing piece in making enterprise AI actually work at scale? Why?
Integration is more than the missing piece; it’s the operating system for the agentic enterprise. However, I’d push the conversation further than traditional integration. What enterprises need today is an active data foundation: a platform that doesn’t just connect systems, but connects data, orchestrates workflows, and governs AI for both people and agents.
That’s fundamentally different from what integration meant five years ago. We’ve evolved well beyond iPaaS. What we’ve built at Boomi is a governed data-to-agent platform, because that’s what this moment requires. Agents need real-time data access, context, and intelligent orchestration across your entire enterprise environment. Without that active foundation, agents can be built endlessly and outcomes will still not be scalable.
The enterprises that treat integration as an afterthought are the ones who call us after their AI projects fail to move beyond the demo stage. The ones building on the right foundation are the ones turning AI into real operational impact.
Q3: How important is unified data when it comes to building effective and reliable AI systems?
Only 7% of enterprise data is AI-ready today. Think about what that means. You have 30 years of accumulated enterprise data, and the overwhelming majority of it can’t reliably power the AI systems you’re trying to build. Those projects will be abandoned unless you get your data activated and ready.
Unified data isn’t just a nice-to-have. It’s what determines whether your AI agents make decisions you can trust or decisions that create risk. An agent grounded in fragmented, inconsistent data is going to produce fragmented, inconsistent outcomes at machine speed, at scale. That’s a fundamentally different kind of problem than a human making a bad call.
The answer is building data readiness into your architecture from the start. That means ingesting, synchronizing, and contextualizing real-time data so that AI and operations run on consistent, reliable information. It means treating data activation as infrastructure, not a side project.
Q4: What does “data activation” really mean, and why should businesses care about it right now?
Data activation is what happens when data stops being something you store and starts being something that drives action. It’s the difference between data as an asset on a balance sheet and data as fuel powering your AI agents, your analytics, and your operations in real time.
At Boomi, we’ve repositioned ourselves as the data activation company for AI, and that’s intentional. Enterprises have spent decades collecting data. The problem is that most of it is trapped in legacy systems, in departmental silos, behind APIs that don’t talk to each other. Data activation is about liberating that data, contextualizing it, and putting it in motion so the right agent or system can act on it at the right moment.
And why does this matter right now? Because AI is becoming the primary interface for work. The enterprise is going headless, meaning the real action is increasingly happening agent-to-agent, not human-to-UI. In that world, data that isn’t activated isn’t just underutilized, it’s invisible. And invisible data in an agentic enterprise means your AI is operating blind.
Q5: How can companies break down data silos across hybrid and multi-cloud environments?
The first thing I’d say is to stop trying to move all your data into one place. That model doesn’t work in a hybrid, multi-cloud world. It’s too slow, too costly, and it creates its own governance problems.
Activating data is about getting the right data to the right place at the right time. You don’t need to store all your data in one location, but you do need a foundation that allows you to access any data, anywhere, when it’s needed. Precision, context, and timeliness are what matter in the age of AI. In many cases, an overemphasis on centralization actually works against those principles.
But breaking down silos is also a trust problem, not just an access problem. Federated access to inconsistent records still leaves agents and applications working from conflicting information. That’s why organizations need both reach and reliability: the ability to access data where it lives, while also synchronizing and resolving data across systems so there is a trusted, authoritative view of key business entities. Those capabilities are complementary, not competing approaches.
The model that works is an active data foundation that spans your entire environment (on-prem, public cloud, SaaS applications, wherever your data lives) and gives agents and applications governed access to data where it sits. Combined with data synchronization, entity resolution, and trusted reference points, organizations can deliver the context AI needs without forcing everything into a single repository.
Model Context Protocol (MCP) is a big part of how this becomes practical. It’s emerging as the standard for agent-level data access, and it changes the equation significantly. Instead of building bespoke integrations for every data source, you can publish agent-ready tools with enterprise governance baked in. That’s how you actually dissolve silos in a heterogeneous environment, not by consolidating everything into a single lake, but by making data accessible in a governed, context-aware way wherever it lives.
Q6: Many enterprises still rely on legacy systems — what’s the biggest risk of not modernizing in the age of AI?
The biggest risk is that they are building their agentic future on a foundation that can’t support it. And the collision that creates is happening right now. The shift toward a headless, agentic enterprise is inevitable. AI is becoming the primary interface for work, and your systems need to be able to talk to agents, not just users. Legacy systems weren’t built for that.
But I want to be precise about what “not modernizing” means, because it doesn’t have to mean rip-and-replace. What it means is that your legacy systems have to become part of your connectivity fabric. They have to be able to expose their data and functionality in a way that agents can consume, securely, with governance, in real time. The risk isn’t legacy systems per se. The risk is trapped data and disconnected systems in an era where those things are existential to your AI strategy.
Layering agents onto fragmented legacy infrastructure doesn’t solve the problem. It magnifies it. You end up with governance gaps, operational risks, and skyrocketing costs with nothing to show for it. The path forward is using integration as the bridge, bringing legacy into the active data foundation incrementally, without the disruption of a full rebuild.
Q7: What does an “AI-first” enterprise tech stack look like in 2026?
It’s three parts: headless, agentic, and governed.
Headless means the enterprise is no longer organized around user interfaces. SaaS applications still matter, but they’re transforming into data and functionality providers for agents. The emphasis shifts from what a human sees on a screen to how data flows between agents, systems, and workflows. By the end of 2026, most enterprise software vendors will have a headless strategy. That shift is happening across the industry.
Agentic means you have the infrastructure to build, orchestrate, and manage AI agents at scale, with full lifecycle management, deep telemetry, and a centralized registry so you actually know what’s running and what decisions it’s making. It also means your integration and automation layer has evolved to support orchestrated agentic workflows, not just point-to-point connections.
And governed is the part that determines whether the first two are sustainable. The active data foundation at the center of an AI-first stack has to provide policy controls, real-time observability, audit trails, and the ability to enforce compliance boundaries, including regional data sovereignty requirements as you scale globally. You can’t deploy AI at enterprise scale without governance built in from the ground up. It’s not optional infrastructure. It’s what makes everything else trustworthy.
Q8: How can integration-led strategies help businesses become more agile and innovation-driven?
An active data foundation removes friction from every layer of the business. When your data flows in real time, when your systems are connected with governance, when your teams can build and deploy agentic workflows without waiting for months of custom integration work, that’s when the speed of innovation fundamentally changes.
What we’re focused on at Boomi is simplifying the development of agents and helping enterprise customers unlock the workloads they want to run. That’s the core of what integration-led agility looks like in 2026. You’re not just connecting systems. You’re enabling humans and agents to work together across every layer of the business (IT, developers, knowledge workers) on a shared, trusted foundation.
The organizations that are becoming genuinely innovation-driven are the ones treating their integration and automation layer as a strategic asset. When you can compose new workflows, connect new tools, and deploy new agents in hours rather than months, iteration speed compounds. That’s a structural competitive advantage, and it comes directly from getting the foundation right.
Q9: With the rise of AI agents, how should organizations rethink governance and control?
Governance has to be infrastructure, not policy. That’s the fundamental rethink required.
We’ve been in the age of API sprawl, and now we’re entering agent sprawl. Organizations are building agents in every corner of the business without a coherent picture of what’s running, what data it’s touching, or what decisions it’s making at machine speed. Success in the next phase of enterprise AI won’t be defined by how many agents you deploy. It’ll be defined by how well they are connected, governed, and grounded in trusted data.
That means building governance into the architectural layer, not adding it as an afterthought when something goes wrong. It means a centralized agent registry with full lifecycle management. It means real-time policy enforcement, observability at every step, audit trails, and the ability to enforce compliance boundaries, including regional data sovereignty as you scale globally. It means having a control tower that gives you a single pane of glass across every agent, every workflow, every data access event.
The organizations that employ governance from day one are going to be able to scale responsibly and fast.
Q10: What role does integration play in ensuring security, compliance, and visibility in AI-driven ecosystems?
In an agentic enterprise, data is in constant motion, between agents, between systems, across cloud environments, across geographic and regulatory boundaries. Without a platform that enforces security policies, manages access, and provides end-to-end visibility at that layer, you have no reliable picture of what’s happening across your environment. And what you can’t see, you can’t govern.
We built Boomi Agentstudio and Agent Control Tower specifically to address this. Enterprise AI is only as good as the data that feeds it and the governance that controls it. The platform has to be the place where policy is enforced, where compliance boundaries are respected, and where every data access event by every agent is auditable. The agent itself can’t make those decisions. The foundation has to.
From a security standpoint, MCP with enterprise governance is how we are enabling agents to get instant access to more than 1,000 enterprise applications without creating uncontrolled exposure. Every tool published through the platform carries governance with it. That’s the model for security in an agentic ecosystem, enabling governed access that’s secure, scalable, and compliant wherever your business operates.
About Ed Macosky:
Ed Macosky is the Chief Product and Technology Officer at Boomi. He has more than 20 years of experience building high performing agile teams, designing and launching new software and service products, solution delivery and customer retention.
Ed works to establish and execute Boomi’s product vision and roadmap – delivering an intelligent, flexible and scalable integration platform that accelerates business outcomes by making information, interactions and innovations flow faster.
Prior to joining Boomi in early 2019, Ed spent almost five years at Oracle, where he served as Vice President of Middleware/PaaS DevOps & Release Engineering. Prior to Oracle, he spent some time at Dell. In addition, starting in 2003, Ed spent more than seven years at Boomi during the formative years of the company.
Named one of the Top 25 Software Product Executives of 2022, Ed studied Information Technology at Penn State and holds six U.S. patents for integrating software systems.
