Day Zero.
The day before the chaos started.
Most AI engagements arrive at the consultant’s door at Day 90 of chaos. Three vendors have been hired. Two prototypes are running. A roadmap exists but doesn’t match the reality, and the team is debating which fire to put out first. Day Zero is the day before any of that started — and the framework for finding the smallest stable foundation from which the project can be rebuilt.
The four moves
Walk these in order. Each one is a decision, not a project.
-
1
Name the user burden, not the feature.
If you can describe the AI you want, you’re too far ahead. Walk it back to the human burden it’s supposed to dissolve. “Managers spend 4 hours assembling status from 6 systems” is a Day Zero statement. “We want an AI dashboard” is a Day 90 statement.
-
2
Identify the locked-in decisions.
List every API contract, vendor commitment, schema, and UI binding made in the last 30 days. Mark which ones foreclose options that the user-burden statement requires. That subset is the rewrite list, whether you’re honest about it yet or not.
-
3
Install the orchestration boundary.
Between the user-facing surface and the underlying systems, install a thin layer that accepts intent and returns outcomes. Today it can be a pass-through. Tomorrow it becomes the orchestrator. Without this boundary, every AI roadmap item becomes a rewrite. With it, they become extensions.
-
4
Ship one orchestration outcome end-to-end.
Pick one user burden from step 1. Route it through the boundary from step 3. Ship the outcome — not the dashboard, not the form, the outcome. That single end-to-end is your proof the architecture works. Everything after is replication.
Clippy-shape projects.
The 1997 Microsoft Office assistant became the canonical failure case: AI overlaid on a surface that still demanded the user maintain context. The reason Clippy failed wasn’t intelligence — it was architecture. If your AI watches the user click and offers suggestions, you’ve built Clippy. If your AI receives intent and returns outcomes, you’ve built what Clippy was pretending to be.
Day Zero is the day before you Clippy your project. Most engagements arrive here too late. We try to be useful anyway.
If your team is mid-engagement and this rings true,
a 20-minute prescreen with Reid is the cheapest way to find out which moves apply to you.
Open Reid →