Index
PRODUCT RESEARCH + SYSTEMS THINKING
Rethinking commercial identity in the AI era
Rethinking commercial identity in the AI era
Rethinking commercial identity in the AI era
Most AI image tools optimize for abundance. Pluribus started from a different observation: production teams do not need more images as much as they need images they can trust, reuse and keep consistent over time.
Most AI image tools optimize for abundance. Pluribus started from a different observation: production teams do not need more images as much as they need images they can trust, reuse and keep consistent over time.
Most AI image tools optimize for abundance. Pluribus started from a different observation: production teams do not need more images as much as they need images they can trust, reuse and keep consistent over time.
The case study is the thinking behind that product: workflow-first architecture, identity persistence, approval as a product layer and memory as the system’s compound advantage.
The case study is the thinking behind that product: workflow-first architecture, identity persistence, approval as a product layer and memory as the system’s compound advantage.
System visual: the gap between generating 100 images and using 100 images consistently
01 / WHY CONSISTENCY FAILED
The real constraints were not creative. They were operational.
The real constraints were not creative. They were operational.
The real constraints were not creative. They were operational.
Enhanced needed high quality campaign imagery for its athletes. Traditional photoshoots were expensive, slow to organise, and difficult to scale.
I had been experimenting with AI-generated images and believed the latest models had become good enough to replace traditional shoots. I built an internal workflow to test the idea.
The images were surprisingly good. The workflow wasn't. AI had made creating images almost instant, but we still lacked the infrastructure to manage consent, approvals, reference assets, and brand safe reuse at scale.
02 / THE THESIS
Objectives
Objectives
Objectives
Reduce production costs
Shorten campaign turnaround
Preserve athlete identity
Maintain creative quality
Standardise approvals
Enable marketing teams to self-serve
Build a workflow that improved over time
System diagram: approved outputs become reusable production memory
02 / THE THESIS
AI generation is abundant. Taste is scarce.
AI generation is abundant. Taste is scarce.
AI generation is abundant. Taste is scarce.
The product should turn taste into repeatable infrastructure. Approved outputs should not become static files; they should become memory. Rules like identity locks and do-not-change constraints are not limitations. They are how production teams preserve intent when generation becomes cheap.
The product should turn taste into repeatable infrastructure. Approved outputs should not become static files; they should become memory. Rules like identity locks and do-not-change constraints are not limitations. They are how production teams preserve intent when generation becomes cheap.
The product should turn taste into repeatable infrastructure. Approved outputs should not become static files; they should become memory. Rules like identity locks and do-not-change constraints are not limitations. They are how production teams preserve intent when generation becomes cheap.
System diagram: approved outputs become reusable production memory
03 / RESEARCH FINDINGS
What production teams actually needed.
What production teams actually needed.
What production teams actually needed.
The bottleneck was approval and memory, not generation. Consistency mattered more than novelty because campaign work is judged as a set. Identity was sacred: teams cared less about impressive images than whether the subject remained recognizably themselves. Workflow mattered more than any individual output.
The bottleneck was approval and memory, not generation. Consistency mattered more than novelty because campaign work is judged as a set. Identity was sacred: teams cared less about impressive images than whether the subject remained recognizably themselves. Workflow mattered more than any individual output.
The bottleneck was approval and memory, not generation. Consistency mattered more than novelty because campaign work is judged as a set. Identity was sacred: teams cared less about impressive images than whether the subject remained recognizably themselves. Workflow mattered more than any individual output.
Approval is not overhead. It is the taste layer that improves future work.
Approval is not overhead. It is the taste layer that improves future work.
Consistency beats novelty when assets need to survive a campaign, not just impress once.
Consistency beats novelty when assets need to survive a campaign, not just impress once.
Identity accuracy matters more than visual surprise because the subject is the brand asset.
Identity accuracy matters more than visual surprise because the subject is the brand asset.
Workflow organization is as valuable as generation quality when teams need repeatability.
Workflow organization is as valuable as generation quality when teams need repeatability.
Memory is what turns isolated outputs into a production system.
Memory is what turns isolated outputs into a production system.
04 / SYSTEM ARCHITECTURE
Workflow-first, not prompt-first.
Workflow-first, not prompt-first.
Workflow-first, not prompt-first.
Pluribus revolves around subjects, campaigns, recipes, approvals and memory — not a blank prompt box. A subject is not a folder; it is a profile with persistent identity. A campaign is not a gallery; it is the context that gives outputs meaning. Approval is not an afterthought; it is the mechanism that decides what the system should remember.
Pluribus revolves around subjects, campaigns, recipes, approvals and memory — not a blank prompt box. A subject is not a folder; it is a profile with persistent identity. A campaign is not a gallery; it is the context that gives outputs meaning. Approval is not an afterthought; it is the mechanism that decides what the system should remember.
Pluribus revolves around subjects, campaigns, recipes, approvals and memory — not a blank prompt box. A subject is not a folder; it is a profile with persistent identity. A campaign is not a gallery; it is the context that gives outputs meaning. Approval is not an afterthought; it is the mechanism that decides what the system should remember.
Architecture diagram: subjects, recipes, campaigns, approvals and memory interconnect
05 / CREATIVE CONSTITUTION
Encoding taste to prevent AI chaos.
Encoding taste to prevent AI chaos.
Encoding taste to prevent AI chaos.
The Creative Constitution defined what good production output meant. The subject must remain recognizably themselves. The image must feel art-directed, not random. Assets should work together as a campaign, not as isolated wins. The final output has to be usable in a real workflow.
The Creative Constitution defined what good production output meant. The subject must remain recognizably themselves. The image must feel art-directed, not random. Assets should work together as a campaign, not as isolated wins. The final output has to be usable in a real workflow.
The Creative Constitution defined what good production output meant. The subject must remain recognizably themselves. The image must feel art-directed, not random. Assets should work together as a campaign, not as isolated wins. The final output has to be usable in a real workflow.
Quality dimensions: Identity → Anatomy → Photography → Styling → Campaign Fit
07 / PRODUCTION LOOP
Approved work becomes the next generation’s context.
Approved work becomes the next generation’s context.
Approved work becomes the next generation’s context.
The core workflow is a loop: identity informs recipes, recipes shape campaigns, campaigns generate options, review turns options into decisions, and approved decisions become memory. Feedback is not just a comment. It is a learning signal for future production.
The core workflow is a loop: identity informs recipes, recipes shape campaigns, campaigns generate options, review turns options into decisions, and approved decisions become memory. Feedback is not just a comment. It is a learning signal for future production.
The core workflow is a loop: identity informs recipes, recipes shape campaigns, campaigns generate options, review turns options into decisions, and approved decisions become memory. Feedback is not just a comment. It is a learning signal for future production.
Loop diagram: Identity → Recipe → Campaign → Generate → Review → Approve → Memory → Reuse
08 / RESULTS
The working system proved that reliability compounds.
The working system proved that reliability compounds.
The working system proved that reliability compounds.
Teams cared more about consistency than novelty. Approval memory became more valuable over time because every decision reduced future ambiguity. Workflow organization was as important as generation quality. Human judgment was not something to automate away; it was the layer that made production reliable.
Teams cared more about consistency than novelty. Approval memory became more valuable over time because every decision reduced future ambiguity. Workflow organization was as important as generation quality. Human judgment was not something to automate away; it was the layer that made production reliable.
Teams cared more about consistency than novelty. Approval memory became more valuable over time because every decision reduced future ambiguity. Workflow organization was as important as generation quality. Human judgment was not something to automate away; it was the layer that made production reliable.
Consistency over novelty ✓
Consistency over novelty ✓
Approval memory compounds over time ✓
Approval memory compounds over time ✓
Workflow organization is as valuable as generation ✓
Workflow organization is as valuable as generation ✓
Human judgment is the taste layer ✓
Human judgment is the taste layer ✓
09 / REFLECTIONS
The moat is not generation. It is memory and approval.
The moat is not generation. It is memory and approval.
The moat is not generation. It is memory and approval.
Building Pluribus clarified that the hardest problem in AI products is not producing more output. It is designing systems people trust enough to build their work around. Human judgment is not a bottleneck in that system. It is the creative leverage point.
Building Pluribus clarified that the hardest problem in AI products is not producing more output. It is designing systems people trust enough to build their work around. Human judgment is not a bottleneck in that system. It is the creative leverage point.
Building Pluribus clarified that the hardest problem in AI products is not producing more output. It is designing systems people trust enough to build their work around. Human judgment is not a bottleneck in that system. It is the creative leverage point.
10 / INTERFACE EVIDENCE
Screens should illustrate the system, not decorate the story.
Screens should illustrate the system, not decorate the story.
Screens should illustrate the system, not decorate the story.
The final interface images should show identity profiles, recipes, approval queues and campaign memory. Each image should explain how the system makes production more reliable.
The final interface images should show identity profiles, recipes, approval queues and campaign memory. Each image should explain how the system makes production more reliable.
The final interface images should show identity profiles, recipes, approval queues and campaign memory. Each image should explain how the system makes production more reliable.
Identity profile: persistent subject memory
Recipe system: repeatable creative direction
Approval queue: taste as a product layer
Campaign memory: decisions compound over time