§ 5.3 · /ai-transformation/agentic-architecture

Agents that work in production need architecture, not prompts.

Most AI pilots fail not because the model can't do the task, but because the system around the model isn't built for production work. Here's how we build ours.

The difference between a prompt and an agent.

A prompt is a single call. An agent is a system: a model, plus memory, plus tools, plus confidence gates, plus human-in-the-loop routing, plus exception handling, plus audit logging, plus an integration surface to your actual ERP. Most failed AI pilots in mid-market businesses built the first thing and called it the second.

Our architecture principles.

01 Deterministic where you can, agentic where you must.

We use classical software for the 90% of logic that's rule-based and reserve the LLM for the judgement calls. That keeps cost, latency, and unpredictability low.

02 Confidence gates, not blind autonomy.

Every agent decision has a confidence score. Above the threshold, it acts. Below it, a human sees it in a workbench with the evidence attached.

03 Resolution memory.

When a human resolves an exception, the agent learns the pattern. Next time, the threshold is lower or the rule is internalised. ROI compounds over time.

04 Full audit trail.

Every step an agent takes — every input, every reasoning step, every output — is logged. Regulated operators (healthcare, finance, aviation) need this to deploy at all.

05 Local or cloud, same architecture.

We can run the same agentic system against cloud Claude, local Ollama with Qwen 3 72B or Llama 3.3 70B, or a hybrid. The architecture doesn't change — only the inference endpoint.

Want to compare architectures?