The 10-80-10 Rule for Creative Teams using AI [+Workflows]

Pixar has a standing process called the Braintrust — a small group of trusted creatives who review every film at the start and again near the end. In between, the production team builds freely for months.

Ed Catmull, Pixar’s co-founder, wrote in Creativity, Inc. that the Braintrust was the most important creative process at the company. The 10% at each end determines everything that happens in the 80% in between.

This is the 10-80-10 rule. And Steve Jobs used it as a delegation model while developing the iPhone.

10-8-10 is the most useful framework for any creative team today, especially ones automating enterprise design with AI.

The framework is a proprietary methodology. But it cuts directly against the way most teams approach AI creative tools, as a button you press and hope for the best.

The teams seeing transformational results from AI are the ones who understand exactly where human judgment belongs in the production process, and where it does not.

What is the 10-80-10 rule?

The framework traces its roots to management thinking that emerged around the Jobs-era Apple product philosophy, the idea that the critical human decisions cluster at the beginning and the end of any production process, with a well-defined execution phase in the middle.

Applied to AI creative production:

• First 10%: Human direction

This is the brief building stage. Selecting references, devising a creative strategy and setting up a design workflow. The ceiling of everything that follows sets here.

• 80%: AI execution

This is where the design happens: asset generation, format adaptation, variation creation, voiceover production, video generation, upscaling, relighting. The entire production run happens here, at a speed and volume no human team can match manually.

• Final 10%: Human review and refinement

This is where brand judgment, strategic fit, cultural sensitivity, legal compliance, and creative taste are applied. The output of the 80% passes through this gate before anything ships.

10-80-10 Rule: a detailed breakdown for enterprise design workflows

The first 10%: creative briefing

This is the most underinvested part of AI creative work.

The critical insight that most teams miss: the 10% on each end is not a small contribution. It is the ceiling setter and the quality gate. A weak first 10% produces mediocre outputs at scale, which is worse than producing fewer mediocre outputs manually, because volume amplifies the problem.

The common failure mode is to spend three minutes writing a prompt, hit generate, be disappointed with the results, and conclude that “AI isn’t good enough yet.” The problem is never the model. It is the brief.

A real first 10% on ImagineArt is a structured mental design investment in the production system that will run the 80%:

The Prompt Node

A production-grade brief in the Prompt node is built in layers: subject definition (what exactly is in frame), style specification (aesthetic reference, color treatment, compositional rules), lighting (direction, temperature, mood), output constraints (aspect ratio, resolution, what must not appear), and emotional register (how should this feel to the viewer). Each layer narrows the generation space in the right direction. Skip a layer and the 80% wanders.

AI Copilot

ImagineArt’s AI Copilot node accepts text, image, and video inputs. Before committing the brief to the full workflow, use AI Copilot to analyze reference images and generate prompt variations. Feed it the brand’s existing approved assets. Ask it to identify the visual characteristics that make them consistent. Use those characteristics to improve the Prompt node before the workflow runs.

Style References in Edit Image

Feed approved brand references directly into Edit Image nodes before any generation runs. The visual DNA of the brief becomes structural. This is how you get consistency across a hundred variations instead of fifty different interpretations of the same prompt.

Workflow Setup

Which nodes, which models, in which sequence — these decisions are the first 10% at the system level. Does this campaign need Kling 3.0 for high-motion video or Veo 3.1 for cinematic quality? Does the image pass through Relight before or after Edit Image? Is Image Iterator the right node for catalog processing or should this run as individual Generate Image calls? The architecture of the workflow determines the ceiling of everything the 80% can produce.

The 80% — watch ImagineArt design

This is the engine. What used to take a production team weeks — shot by shot, format by format, channel by channel — now takes a workflow run measured in minutes.

What AI executes in the 80%:

• Asset generation at volume: Hundreds of variations from one brief — Generate Image with Flux 2 or ImagineArt 1.5, generating the full range of options the brief can support
• Format adaptation without manual work: AI Resize adapts every asset to every required channel dimension, 9:16 for Reels, 1:1 for feed, 4:5 for paid, 16:9 for YouTube — in one node pass
• Catalog processing: Image Iterator runs every SKU through the same production workflow — same style, same lighting, same quality — without re-briefing for each product
• Video from stills: Generate Video with Veo 3.1, Seedance 2, or Kling 3.0 turns hero images into campaign video without a production shoot
• Audio production: Generate Audio produces voiceover and music, sound effects
• Relighting, reframing, extending: Relight adjusts the lighting pass without reshooting; Multiple Camera Angles generates additional perspectives; Extend Video adds duration without new footage
• Quality pass: Upscale Image and Upscale Video bring resolution to delivery standard

A full campaign production pipeline on ImagineArt

Prompt → AI Copilot → Generate Image → Edit Image → Relight → Image Iterator → Generate Video → Generate Audio → Combine Video + Audio → AI Resize → Upscale Video

Real Applications by Segment

Fashion brand

First 10% = art direction session, garment reference uploads, model brief and demographic specification.

80% = ImagineArt generates the full lookbook across all SKUs, all camera angles (Multiple Camera Angles), all channel formats (AI Resize).

Final 10% = art director reviews the full batch, selects hero images, approves the campaign set.

Marketing team

First 10% = campaign brief with performance goals, visual references, copy inputs.

80% = 30+ creative variants across all paid and organic channels, video versions generated from hero stills, voiceover produced in three regional voices.

Final 10% = team selects the A/B test set, approves creative for spend.

Financial services firm

First 10% = compliance team reviews master template, legal approves the brief structure, brand lead sets visual parameters.

80% = regional campaign variants generated across 20 markets from the same workflow.

Final 10% = regional compliance sign-off on the batch output, one market at a time.

The final 10%: non-negotiable human review

This is where brand identity, strategic judgment, and creative taste live. The final 10% is a substantive evaluation and the evaluation criteria are organizational, strategic, and cultural in ways that require human context.

What the final 10% actually does:

• Brand accuracy check: Does this look like us? Does it feel like our brand, not just technically compliant with the brief?
• Strategic fit: Is this the right message for this moment, this market, this campaign objective?
• Cultural sensitivity: Are there things in this batch that read differently in Market X than at HQ?
• Compliance and legal: For regulated industries, this is the formal sign-off pass — the audit trail, the approval record

Tip: ImagineArt Enterprise users get built-in collaboration features. Depending on role-based access, team users can comment, approve and overview activity.

Why this framework matters at enterprise scale

A single creator with good instincts can approximate the 10-80-10 process intuitively. The problem is that intuition does not scale. When five users are generating assets for the same brand across three markets, informal 10-80-10 produces five different versions of the first 10%, five different interpretations of what “brand-appropriate” means in the 80%, and five different quality bars in the final 10%. The result looks like five different brands.

With 10-80-10 formalized at the organizational level:

• Standardized first 10%: Master workflow templates encode the brief structure. Every campaign starts from the same approved architecture.
• Governed 80%: The workflow canvas is the documented production process. It is auditable, reproducible, and scalable.
• Accountable final 10%: Clear owner, clear criteria, clear audit trail — the approval record that regulated industries require and any serious brand should want.

The App Builder

ImagineArt App Builder makes the 10-80-10 structure more architectural and bureaucratic. The design team encodes the entire first 10% into the app — the approved Prompt node, for example, the locked mockup variations, style references, the Relight configuration, the model selection. Marketing runs the 80% through a clean interface that shows only what they need to operate. The brand-critical settings are untouchable. The final 10% reviews the batch output against the standard the first 10% set.

The compounding effect starts right here: each approved workflow becomes a template.

• The first 10% gets faster with each cycle because the template already exists.
• The final 10% gets faster because the outputs become more predictable.

A team that runs 10 campaigns on the same approved workflow template is not doing ten times the work of the first campaign. They are doing approximately one campaign’s worth of first-10% work, multiplied across ten executions. The denominator grows. The numerator stays largely fixed. That is what compounding looks like in creative production.

How to implement 10-80-10 rule

Step 1: Pick one design/content operation.

recurring content type your team produces manually — not a one-off, but something that repeats, for example, social campaign assets, product photography, email hero images. Pick the thing with the highest repetition rate.

Step 2: Create the first 10% brief.

Write a structured first 10% brief: proper, complete, guided. Add Reference images (3–5 approved brand assets), brand inputs (colour palette, prohibited visual elements, character demographics), channel specifications (every format you need to deliver).

Step 3: Build the workflow in ImagineArt.

Start with the nodes that match your content type. Image campaign: Prompt → Generate Image → Edit Image → Relight → AI Resize → Upscale Image. Don’t over-engineer on the first run — build the minimum viable pipeline for the content type you chose.

Step 4: Run a pilot batch.

Generate 20–30 assets from the workflow. Do not curate as you go. Let the 80% run completely before the final 10% begins.

Step 5: Check the final 10% as a team.

Review the batch together. Select the heroes. Document specifically what you would change in the brief to produce better outputs — not “this looks off” but “the lighting is too warm, the Relight node should be set to 5000K instead of 4000K.” That specificity is the input to the next first 10%. Use team collaboration features in the workflows and improve the brief based on comments. Implement the changes from step 5 in the workflow. Save it.

Step 6: Create an app from the workflow - optional

By now, you have an approved design workflow for specific types of campaigns — considering making it an app to distribute to marketing team as well. This is now your approved production template for this content type — a repeatable system.

Step 7: Scale.

Next campaign of this type runs the same workflow. Adjust only the variables that actually change: product, copy, seasonal color, market. Everything else flows down from the approved template.

Ready to implement 10-80-10 rule to your workflows?

The teams getting transformational results from AI are the ones treating it as a production system. 10-80-10 is the operating model that makes AI output worth using. It is the structure that lets creative ambition run at scale without degrading into inconsistency.