Integration for AI Agents, Part 2: The Architecture We Build To
The first ninety days decide whether an agent programme pays off. We bring the predictable mistakes and the target architecture.
About this page. Aitina Tech’s R&D wing applies the research of Imtiaz Hassan, published as The Software Lens, in client work and in our own product. This page is the applied view: what the pattern is, how we build it, and what an engagement looks like. The full research is on The Software Lens.
Most AI projects that connect to existing systems go wrong in the same few ways in the first three months: systems wrapped without being described, guardrails left in the client, no cost model, evidence rebuilt from logs after the fact. Avoiding them is cheaper than learning them.
From the research
“The old way is not switched off; it is replaced one capability at a time.”
What we take from it. The ninety-day mistakes are avoidable if the architecture is set before the first system is wrapped. That is where an engagement with us starts.
How we apply it
We build to five layers an integration architect already knows, with identity and authority running across all of them: the systems of record stay where they are, each is wrapped as a described capability, a catalogue and a business vocabulary sit above them, agents compose capabilities at run time, and every call leaves a signed receipt. The standards stay in place: in higher education the agent layer sits on top of LTI and OneRoster; in payments on top of ISO 20022.
The two sectors where getting this wrong costs most are the ones we work in most: a student's funding depends on eight systems agreeing about one person, and a regulator's finding depends on one customer being one customer across the estate. The old way is not switched off; it is replaced one capability at a time, which is how our software integration engagements are run.
What an engagement looks like
We usually start with one record that drifts between systems (a student, a customer, an employee), wrap the systems that hold it as capabilities, and put a consistency agent in front of them with receipts. The measurable question, whether it is cheaper to change than what you have, is answered on that first slice before anything else is committed.
The research behind it
This page is Aitina Tech’s applied account of Enterprise Integration Patterns in the Agentic Era, Part 2, by Imtiaz Hassan, published on The Software Lens. The full argument, the diagrams and the worked examples are there; this page is what we do with it for clients.