The Automation Design Problem
Most content automation projects start with a tool, not a design. A team identifies a capability — AI can draft content, a workflow platform can route approvals, an integration can sync assets between systems — and builds an automation around it. The result is a collection of automations that each solve a local problem but do not add up to a coherent system.
Automation designed from capability produces fragmented efficiency gains. Automation designed from process produces compounding returns — each automated step enabling the next, the whole system becoming more capable than the sum of its parts.
The Four-Phase Automation Sequence
Phase 1 — Standardise inputs: Automate the enforcement of structured inputs before automating any production step. Brief templates, metadata requirements, and content type specifications must be standardised before AI can reliably generate from them.
Phase 2 — Automate classification and routing: Once inputs are standardised, automate the categorisation and routing logic. AI can classify content by type, risk level, and required governance path — routing each piece to the correct review chain without manual triage.
Phase 3 — Automate quality gates: Pre-publication quality checks — brand compliance, metadata completeness, structural validity — can be automated once the quality criteria are defined. This reduces the review burden on human approvers.
Phase 4 — Automate optimisation: Performance-based content optimisation — identifying underperforming content, triggering update cycles, personalising variants — is the highest-value automation tier. It requires Phases 1–3 to be stable before it can function reliably.
Key Takeaways
1. Automation designed from capability produces fragmented efficiency gains — automation designed from process produces compounding returns.
2. The four-phase sequence — standardise inputs, automate classification, automate quality gates, automate optimisation — must be followed in order. Skipping phases creates the chaos automation was supposed to prevent.
3. The question is not what you can automate — it is what you should automate, and when. Not every step that can be automated should be automated.