---
title: "What the architect does now"
url: https://raskell.io/notes/what-the-architect-does-now/
author: Raffael Schneider
date: 2026-10-26
description: "If agents propose changes and a gate scores them, the architect stops reviewing designs and starts setting thresholds and owning escalations."
tags: [enterprise-architecture, future-of-work, risk, augmentation, agents]
---

# What the architect does now

The enterprise architect's week, as I have known it, is review: solution designs arrive, the architect checks them against principles and the target landscape, and the queue is always longer than the week. It is the bottleneck in every large IT organisation I have seen, and the scarce thing in it is not the checking, which is mostly pattern-matching, but the judgement about the few designs that are genuinely new or genuinely dangerous.

With [a risk gate at design time](https://raskell.io/notes/design-time-risk-with-a-threshold/), the queue is handled by the gate and the architect's work becomes two things. Setting the thresholds, which means deciding, explicitly and in writing, how much blast radius, which data classes and which regulatory scopes a change may touch before a human has to see it, and reviewing those settings against what actually happened. And owning the escalations, the designs above the threshold, which are by construction the interesting ones. The ratio of judgement to pattern-matching in the job inverts.

It is the same change in kind as the one I described in [What Sixteen AI Agents Taught Me About Management](https://raskell.io/articles/what-sixteen-ai-agents-taught-me-about-management/) at the scale of one person's week: the work moves from doing to deciding what may be done, and the people who are good at that become more valuable, not less. The architects who will struggle are the ones whose authority was the queue.
