Costed Tests

A Practical Go, Wait, or Pass Framework for Production Teams

A Practical Go, Wait, or Pass Framework for Production Teams
This article introduces a practical Go, Wait, or Pass framework for creative production teams evaluating AI pilots. Instead of relying on unstructured optimism, teams evaluate projects based on three critical questions: revision survival, fully loaded costs, and transferability. This evidence-based decision model prevents perpetual testing and premature adoption, ensuring that only verified processes enter live workflows.

At the end of a pilot the team still has to decide what to do next. The options are not “adopt” or “abandon.” They are more precise than that, and the language I use with production teams is simple: Go, Wait, or Pass.

Go means the process is ready to be used on a real job under defined conditions.
Wait means the process shows promise but still lacks evidence on one or more critical points.
Pass means the current approach is not worth further investment for this team’s work at this time.

The framework exists because most pilots end in a fog of mixed signals. The pictures looked good. Some notes stayed local. The schedule still slipped. Someone is optimistic. Someone else is cautious. Without a shared decision rule, the pilot simply extends into another round of unstructured testing.

The Three Questions That Drive the Decision

Close-up of a widescreen monitor showing an AI platform interface. One approved sneaker generation is contrasted with four rejected variations, highlighting the Go/Wait decision point.

I ask the same three questions at the close of every controlled pilot. The answers determine the recommendation.

1. Did a take survive a realistic revision note without a full restart?

If the answer is no, the process is still in the demonstration phase. A strong first generation is not enough. The minimum bar for Go is that at least one realistic note stayed local or was resolved with a targeted, documented adjustment.

2. Is the fully loaded cost (labor + service) understood and acceptable for the kind of work this team actually does?

If the team cannot state the cost of reaching an exportable take, or if that cost only works under unrealistically perfect preparation, the honest recommendation is Wait or Pass. Optimism about future efficiency is not a cost model.

3. Can a different person on the team reproduce the result using only the written process?

If the answer depends on one individual’s undocumented judgment, the process is not yet transferable. That is a Wait condition. A Go recommendation requires that the critical steps are written and have been run successfully by someone other than the original operator.

All three questions must clear for a Go. One clear failure is usually enough for a Pass on that particular approach. Anything in between lands in Wait with a specific list of what still needs to be proven.

What Each Recommendation Looks Like in Practice

Go
The pilot produced a marked take that survived its revision test. The fully loaded hours and service cost were recorded and sit inside an acceptable range for the intended use case. A second person has run the same written process on a separate brief with comparable results. The team may now use the process on a live job, still under the documented constraints (reference standards, variation budget, scope limits, handoff note required).

Wait
The pictures are promising and at least one note stayed local, but one of the three questions is still open. Typical Wait reasons: cost is known only for ideal preparation; revision path works for one class of note but not the most common one; process still depends on a single person. The next action is a short, targeted test that closes the specific gap—not another open-ended exploration.

Pass
The approach consistently fails the revision test, the fully loaded cost is unacceptable for the work the team does, or the process cannot be made repeatable without disproportionate effort. Pass is not a judgment on the model in general. It is a judgment on this process for this team’s current pipeline. The team stops spending further test budget on this path and redirects attention elsewhere.

How the Framework Prevents Two Common Failure Modes

The first failure mode is perpetual testing. The pilot never ends because there is always one more variable to try. The framework forces a decision against explicit criteria after a defined amount of evidence.

The second failure mode is premature adoption. A single attractive cycle is treated as proof. The framework requires evidence on revision, cost, and transferability before the process is allowed near a client deadline.

Both failure modes are more expensive than a clear Wait or Pass.

A Calibration Example

A small agency team ran two controlled thirty-second product pilots with Seedance 2.5. The first cycle looked strong but the revision note forced a full restart. Fully loaded cost was recorded. The process still lived mostly in one person’s head.

Recommendation after the first pilot: Wait. Specific gap: revision path and transferability.

They spent the next session tightening the locked list, testing the most common note type, and having a second person run the preparation and selection steps. The second pilot produced a take that absorbed the note locally. The second operator’s cycle landed inside the same cost band.

Recommendation after the second pilot: Go, limited to product-hero social/pre-roll work under the written constraints.

The team did not declare a general victory for AI video. They declared a narrow, evidence-based permission to use a specific process on a specific class of job. That is the only kind of Go I trust.

The dog slept through both decision meetings. Morgan, after hearing the criteria, said it was the first version of “we’re ready” she had heard that actually defined what ready meant. She remains the better judge of these things.

How to Use the Framework at the End of Your Next Pilot

Documentary still life of a finalized, honest project budget on a clipboard. The document clearly itemizes the Go/Wait/Pass recommendation and conditions for live work.
  1. Answer the three questions with evidence, not impressions.

  2. Record the answers in one short paragraph.

  3. Issue a Go, Wait, or Pass recommendation with the specific conditions or gaps attached.

  4. If the result is Wait, define the single next test that will close the largest gap.

  5. If the result is Go, write the constraints under which the process may be used on live work.

  6. If the result is Pass, stop spending test budget on that path and document why.

The pictures will keep improving. The decision framework stays the same. A pilot that cannot answer the three questions is not finished. A process that has not cleared them is not yet a workflow.

If it cannot survive the handoff, it is not a workflow yet.

Last revised · 2026-10-12 15:08
Letters
Readers Write

No letters yet — be the first to write.

Write a letter
© 2026 Render Handoff. All rights reserved. Render Handoff