technologyDo

The architecture

AITLA

Agent-Intelligent Technology Lifecycle Architecture. The roles, the lifecycle and the rules that let governed AI agents do the work of an IT department. Agent-Governed Development (AGD) is the method; AITLA is how it is implemented.

Status. AITLA is at its foundation stage: the architecture and policies are written, and the multi-agent runtime is being bootstrapped by the current system. No multi-agent system is serving customers yet.

Five lifecycle stages

AITLA follows the lifecycle an IT department already knows, from deciding what to build to learning from what was built, with agents in every stage.

Strategy & Governance
Who the agents are, what each may decide, and the rules they work under.
Agent-Driven Development
Building to Agent-Governed Development’s working order, with every change reviewed by a role that didn’t make it.
Automated Transition
Moving approved changes into service: releases, changes to data, and the checks before anything ships.
Enhanced Operations
Running what was shipped: watching it, handling incidents, and calling a person when something is beyond the agents.
Continuous Improvement
Learning from each change, so the next one goes better.

Agents in ITIL roles

Each agent holds a role an IT department would recognize from ITIL v4, with that role’s responsibilities and the authority to approve within them.

ITIL v4 roleWhat it owns
Technical LeadOwns the technical design and its effect on the rest of the system.
Database AdministratorOwns the data: how it is stored, changed and recovered.
SecurityOwns security: who can reach what, and how the system is protected.
Product OwnerOwns the why: what the work is for and whether it delivers it.
Delivery ManagerOwns the plan: what is being delivered, by when, and its paperwork.

The bootstrap: the current system builds its replacement

AITLA is being built by the system it will replace. Today a single, monolithic system does the work. It is building the specialized agents that will take that work over, one at a time, under the same policies they will inherit, so the governance is in place before the agents are.

The bootstrap: the current system builds its replacementToday one monolithic system does the work; AITLA is at its foundation stage. That system builds the specialized agents that will replace it, under the same policies. The agents, governed by their three decision tiers with a human override, take the work over.today: the current systemdoes the work(the foundation stage)it builds the agentsthat will replace it,under the same policiesthe agents take the work over,governed by three tiersand a human override
Figure 1. The bootstrap. The current system builds the agents that will replace it, under the policies they will inherit. AITLA is at its foundation stage.

Who decides

Three tiers. Operational agents do the work. Specialized agents decide within their own domain. The Governance Council decides what affects the whole system. A human can override any of them.

Three decision tiers, with a human override at every one Three tiers. The Governance Council decides what affects the whole system. The specialized agents (Architecture, Product Owner, DBA, Security) decide within their own domain. The operational agents (Developer, Tester, Auditor, Deployment, Operations) do the work. Each tier escalates to the one above. A human override enters every tier. Governance Council decides what affects the whole system Specialized agents decide within their own domain Architecture Product Owner DBA Security Operational agents do the work Developer Tester Auditor Deployment Operations escalates escalates human override
Figure 2. Three decision tiers. Each decides within its scope and escalates what it can’t. A human can override any tier.

AITLA is also the first domain of a longer idea: Artificial Institutional Intelligence.