A data-readiness gate: the criteria to pass before scaling AI
Before a pilot becomes production, it should pass an explicit gate. Here are the criteria, and why a vague definition of ready is how pilots die on the way to scale.
A data-readiness gate: the criteria to pass before scaling AI
Most pilots do not have a gate between “the demo worked” and “let us scale this.” The decision gets made in a meeting on the strength of a good demo, and the data questions get waved through. A readiness gate is the explicit checkpoint that stops that. Pass it and you scale. Fail it and you know exactly what to fix first.
Define done, or the gate is theater
A leadership strategist and Lean Six Sigma black belt we spoke with kept returning to one discipline: define what done actually means, with real acceptance criteria, so teams stop over-promising against a vague target. A readiness gate is that discipline applied to data. If “the data is ready” has no written criteria, the gate is just a feeling, and a feeling always caves under the pressure of a leadership team that already wants to ship. Write the criteria down before the pilot, so passing is a fact and not a negotiation.
The five criteria
A useful gate checks the same five dimensions readiness spans: quality, governance, architecture, discoverability, and compliance. Quality: does the production data clear the accuracy bar the use case needs, measured on real records, not the curated sample. Governance: is there a named owner and an explicit permission for using this data this way. Architecture: can the production pipeline deliver the data inside the latency and volume the feature requires. Discoverability: can the system reach every relevant record, not just the ones that happened to be in the pilot. Compliance: has every regulated field been accounted for in the design. A pilot that passes four and fails one is not ready, because the one it failed is what breaks at scale.
Commit to a framework instead of improvising
A field CISO who advises mid-market companies gave advice that transfers directly. Pick a control framework, NIST or CIS or whatever fits, and commit to it, because it tells you where you are excelling and where you are falling down, and you will need that structure sooner or later anyway. A readiness gate works the same way. A named, repeatable set of criteria beats improvising the bar every time, and it makes the go or no-go decision deliberate rather than accidental.
The clock on compliance is real
One criterion is getting sharper by the month. EU AI Act enforcement lands in August 2026, and at that point an incomplete inventory of what data your AI touches stops being a hygiene issue and becomes a compliance violation. If your gate does not include “we know exactly what regulated data this feature uses and we are allowed to use it,” you are scaling into a deadline that will not move for you.
Make failing the gate cheap
The point of a gate is not to block, it is to make the block cheap and early. A pilot that fails on discoverability in a two week check costs you two weeks and a fix list. The same failure discovered three months into a production rollout costs a rollback and the credibility of the whole program. A gate turns a slow, expensive, public failure into a fast, cheap, private one.
How we approach it at Density Labs
The AI Readiness Assessment is that gate, run as a fixed two week engagement priced at $2,500. We score your use case against the five criteria, tie each to a written pass or fail, and flag the compliance exposure before the enforcement deadline reaches it. You leave with a clear verdict and a prioritized list, which is what a leadership team actually needs to decide whether to scale.
Scaling without a gate is not confidence. It is just skipping the check that would have saved you.