Your team has selected a campaign route and started developing it. Then someone arrives with an appealing AI concept. Should it improve the chosen route, replace it, or become another presentation the team now has to finish?
That decision matters more than the speed of generating the concept. For an overseas brand coordinating creative work with a production partner in China, an undeclared change of direction can leave one team expecting finished storyboards while another is still exploring. Keeping an idea is a commitment to further work, not merely saving an image.
A research signal, with a narrow scope
Huang, Yu, Chen and Zhou’s October 10, 2026 preprint, Too Late to Prune, reports a timed product-design experiment with 148 three-person student teams. It compared access to one language model at different stages, with stage-specific prompts. Early exploration access produced submissions rated more highly by experts than access introduced only during selection.
This was a short, structured student task—not an advertising or sales study. It provides no universal schedule for professional teams and does not establish that AI should be switched off after concept selection. ONCE’s practical question is narrower: does a late idea receive a shared comparison before it enters the deliverable?
An alternative is not an extra shot
Imagine a fictional reusable-bottle brand supplied from China. Its campaign route follows the same bottle through the small pauses in a commuter’s day. The team is developing that ordinary, observational story when a new AI concept depicts the bottle suspended inside a dramatic sculptural water installation.
The image could be attractive. But adding it to the existing story may introduce a different visual language rather than strengthen the original one. Presenting it as an additional route may reopen a decision the client believed was complete.
Neither outcome is inherently wrong. The problem is letting it happen without deciding which outcome is intended. The brand sees a promising option; the creative team sees an opportunity; the producer sees another route that appears to require execution.
ONCE’s late-route replacement note
Keep a concise record of the routes selected for development, why they were selected and what remains unresolved. This is not a demand to archive every prompt or treat every reference image as a campaign candidate. Record the choices that shape the overall expression.
When a late route appears, bring it back into the team’s shared candidate set before inserting it into the final deck. Add a replacement note. For the hypothetical bottle campaign, that might read:
The commuter story has a clear human setting, but the product’s silhouette lacks a memorable visual moment. The installation route would replace the observational approach as the main visual idea, not add an isolated spectacle. Compare short storyboard explanations of the same product moment before deciding whether to change route.
The note identifies an unresolved problem, the existing commitment that would change and the smallest useful comparison. It does not declare the newcomer a winner or require two finished concept films.
Keep the communication task stable during that comparison. If the brief remains about everyday companionship, the spectacular route must explain how it serves that meaning. If the brand now wants to prioritize product-form recognition, agree that change explicitly. Do not let a compelling image silently rewrite the brief.
Continue developing when the central idea remains intact
A clearer character action, a more legible product moment or a better sound transition can develop the existing route. AI may still help with that work.
The distinction is not whether a new frame appears. Ask whether the audience takeaway and central story relationship have changed. Improving the promised expression and replacing the promise are different decisions.
Park a route without condemning it
A late idea may belong to a different campaign or need a comparison the team cannot presently undertake. Save the concept with its intended use, without continuing to build a full version or making it an implied backup deliverable.
If it addresses a significant weakness, formally reopen the comparison: identify the existing candidates, the decision-maker, the material needed and whether current development pauses. Adding an independent route is also an option when the extra work and resources are agreed. The principle is not a mechanical one-in, one-out quota. It is an explicit decision about what the team is committing to develop.
Correction is not optional novelty
A verified product error, a substantiated local-market misunderstanding or a changed business objective may invalidate an approved route. Those issues should not be parked merely to protect the schedule.
Establish the problem, then decide whether it affects a detail or the central concept. Correct the former; redevelop the latter. Previous effort is not a reason to preserve a false premise. Equally, a necessary correction does not automatically justify reopening every unrelated creative possibility.
Review the candidate set before delivery
Remote Sparks’ October 9, 2026 guide to creative brainstorming techniques discusses organizing generation and evaluation. The issue here comes afterward: do the routes in the final presentation match the choices the team actually reviewed? Its ten techniques are not being presented as ONCE’s method.
Before delivery, check that each final route has been shared with the team, that late additions or replacements have a stated purpose and trade-off, and that abandoned directions have not survived as implied execution promises.
For a brand buying creative development, the deliverable should reveal how the team moved from possibilities to a workable choice—not just how many alternatives it generated. Exploration and selection both require work. The low cost of producing another concept paragraph does not establish the total cost of developing it into a campaign.
When discussing a brief through ONCE’s video production services, identify whether the current assignment is to explore routes, select among them or develop a chosen one. If a late idea changes that assignment, make the change visible.
The next useful question after “we could add another route” is: which unresolved problem does it address, what would it replace, and what comparison would justify keeping it? Good ideas deserve attention. The chosen ones also deserve enough attention to become finished work.
Evidence and limits
The research is arXiv:2610.14984v1, dated October 10, 2026; its findings concern expert-rated output in the specified experiment. The Remote Sparks article is dated October 9, 2026. Both were checked on October 20. The bottle scenario and replacement note are illustrations, not ONCE client work or a tested intervention. No performance or cost savings are claimed.
