Authors
How organisations can extract value from existing technology by shifting where intelligence operates.
Most organisations still assume that to do AI properly, they need to first modernise their core systems, replace ERP, rebuild data platforms, or procure new digital stacks.
An agentic orchestration layer offers a fundamentally different path. Legacy systems hold significant trapped value; the barrier has always been the cost and disruption of extracting it through large scale modernisation. Agentic mesh architecture narrows that gap by deploying intelligent, autonomous agents that wrap around, bridge between, and orchestrate across legacy systems capturing value without requiring a full reset.
Maintaining agents on top of legacy systems is not a substitute for modernisation, it's a parallel strategy. Agents let you unlock value from existing infrastructure now, while you work a longer-term roadmap to address the real debt underneath: fragile integrations, data quality gaps, and aging platforms. Used well, the mesh buys you time and optionality. Used poorly, it becomes another layer of complexity to maintain. The discipline is knowing which systems deserve an agent wrapper and which deserve to be rebuilt.
The real implication of agentic mesh is strategic, not technical. It changes the sequencing of transformation. Instead of waiting for systems to be modernised before deploying Al, organisations can deploy Al to accelerate modernisation itself.
From a business perspective, this is far more powerful. Leaders do not need to ask whether two systems are integrated; they need to ask whether an agent can operate across them. The focus shifts from integration to execution capability. This also does not mean the architecture becomes simpler. It becomes more abstract. Systems remain heterogeneous, but that heterogeneity is hidden behind an intelligent layer that standardizes interaction at the level of intent rather than implementation if domain boundaries and business intent models are clearly understood.
Absorbing complexity into an agent layer does not make it disappear; it relocates it. Two categories of risk come with this approach, and both must be addressed upfront.
The first is organisational. Agents operating on top of legacy systems must be tightly scoped, governed, and monitored. Access controls become more critical, not less, because agents can traverse multiple systems. In this situation, the challenge becomes more than technical implementation; it becomes organisational. Who decides what an agent is allowed to do? Who owns its outcomes when it spans three departments and two business units? Who has the authority to expand its scope, and who signs off on the risk? Governance in an agentic environment is less about policy documents and more about decision rights, accountability, and a culture that can adapt as agents take on more consequential work.
The second is technical. When agents operate across multiple systems, errors can propagate faster and at greater scale. Poorly constrained agents can introduce inconsistency rather than reduce it. Security boundaries become more complex as agents gain access to multiple domains. But these are not reasons to avoid the model; these are design challenges that must be addressed upfront. In many ways, they are the same challenges organisations already face, just amplified by speed and autonomy. The difference is that the upside is equally amplified.
The bottom line is simple. Run your modernisation program in the background. But don't wait for it to finish before you capture value. Agentic mesh lets you move now, extracting intelligence from the systems you already have while your longer-term roadmap matures underneath; however, it requires continuous tuning to prevent drift in agent behaviour. This tuning must be intentional, with explicit monitoring, ownership, and control designed into the operating model, rather than applied reactively. Over time, as cleaner systems come online, agent logic migrates into them, and the mesh evolves with your stack rather than against it. The advantage is not having better systems tomorrow; it is in using the systems you have today in a fundamentally better way while you build toward what comes next.
