In a recent conversation with Digitalisation World, I challenged one of the most common assumptions in enterprise AI right now. That agents are plug-and-play. When businesses rush to deploy AI agents without first aligning the definitions, decision logic, and context that sit underneath them, results in agents fighting against the very thing they’re supposed to be addressing by masquerading as efficiency boons. Unfortunately, they only accelerate inconsistency. Every system that was already pulling in a different direction now pulls faster.
In this interview, I unpack what that misalignment looks like in practice, why it's so widespread, and what it takes to fix it. Building the shared layer of meaning that makes agents trustworthy in the first place, as opposed to a better agent.
That's the work we do at Valliance. Helping organisations structure their data, define what things mean across the business, and create the conditions where AI can operate reliably at scale.
Watch the full conversation below.
FAQs
Can you deploy AI agents without an ontology?
It depends on what you are asking the agent to do, because an agent automating a defined task inside one system can ship without an ontology, and an agent expected to reason across the business cannot, since it has no reliable way to know what your entities are or how they relate without one. Workflow automation is the clear case. An agent that files a claim or routes a ticket inside one application needs the right access and a defined task, and nothing more. The picture changes as soon as the question spans systems. Dom Selvon's rule for an enterprise intelligence platform is that every layer can ship as a thin version of itself and none can be skipped, and the ontology is one of those layers. For a first agent that can mean a one-page artefact naming the entities it touches. Tarek Nseir makes the case for building that context layer first in Why enterprise AI agents stall, and what to build first.
What breaks first when you move AI agents from pilot to production at enterprise scale?
Trust, and usually before anything technical. A pilot tolerates output that is roughly right. Production does not, because the people relying on it can see when a recommendation misreads how decisions are actually made in their part of the business. The underlying cause is almost always missing context rather than model capability, which is why more capable models rarely rescue a stalled deployment.
What is an enterprise ontology, and do we need one before scaling AI across systems?
An enterprise ontology is a formal, machine-readable map of the concepts, entities and relationships a business runs on. It defines what terms mean, how business units interact and how data connects across different software systems. You do not need a complete one before your first agent, but you do need one for the domain that agent works in, otherwise it has nothing reliable to reason over. The work compounds across every use case that follows, which is why starting it late is more expensive than starting it small.
