AIM in Practice: Putting AI Agents on Top of Your Existing Systems
Agents decide, Intelligence remembers, MCP acts. We put the three layers on top of what you already run.
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.
Plugging AI agents straight into existing systems is how costs run away and mistakes slip through. AIM gives agents a structure: a layer that reasons, a layer that remembers and retrieves, and a layer of tools they are allowed to act through, with every decision traceable from request to action.
From the research
“AIM is not a framework or a RAG system; it is the architecture those tools plug into.”
What we take from it. Nothing you already run has to go. We put the three layers in front of the services you have, so the agents gain structure and your spend stays visible.
How we apply it
Microservices were designed for callers that had their contracts written into code in advance. An agent has to find services at run time, understand them, be stopped at the boundary and stay within a budget. We do not throw the services away; we front them with the three AIM layers and keep each layer to one job: no reasoning in the tool layer, no memory inside the agents, no actions that bypass the tool layer.
On AWS this is an orchestrator and sub-agents on Amazon Bedrock, a Knowledge Store for the Intelligence layer, and Lambda-backed MCP tools with Bedrock Guardrails at the edge, all in your account. The same shape runs on Azure and Google Cloud. It is the architecture behind our LLM platform on AWS Bedrock and AI engineering services, and the first thing we check in an architecture review of an agent programme that has stalled.
What an engagement looks like
A typical start is a two-week diagnostic: where your current estate breaks under an agent caller, which of the four requirements (discoverable, reasoning-safe, guarded at the edge, cost-bounded) it fails, and a target design in AIM. Delivery follows capability by capability, with cost and traceability built in from the first one.
The research behind it
This page is Aitina Tech’s applied account of Microservices Break Under Agents, and Why AIM Fits, 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.
Watch the episodeSaaS to SaaC Model | AIM Pattern | The AI Era ShiftThe Software Lens on YouTube →