David MartinCreative Professional
The idea

Why directing matters.

AI has changed the economics of production. Cheaper production does not produce better work.

Things that once required a team can now be produced by one person working with a collection of models, tools, agents, and automated workflows. But when generation becomes easy, direction becomes more important.

The problem

Without direction, AI produces more of everything.

More versions. More content. More ideas. More noise. More inconsistency. The bottleneck moves from making things to deciding what is worth making.

The answer

A defined role for the human.

AI changes production. It does not eliminate responsibility. Set the direction. Build the system. Brief the crew. Protect the standard. Approve the work.

A director is not the person who does everything. A director is the person responsible for making sure everything is moving toward the right result.

  1. Q1What are we trying to accomplish?
  2. Q2Who is this for?
  3. Q3What does good look like?
  4. Q4What information does the system need?
  5. Q5What should happen automatically?
  6. Q6What requires human judgment?
  7. Q7What can never ship without approval?
The model

From tool user to director.

The old model

Human → Tool → Output

You ask. The model responds. You accept, reject, or try again. The human is essentially operating a very sophisticated interface.

The director model

Objective → Context → Crew → Workflow → Review → Approval

The human establishes the system. AI becomes part of the crew. The shift is from prompting to directing.

What are we trying to make? The role comes before the prompt.

If you are directing an AI-generated video, you are responsible for the creative intent. If you are deploying an AI agent, you are responsible for what the agent is allowed to do. AI can contribute to the work. It cannot own the objective. The first prompt is not the beginning of the production. The decision to direct is.

The principles

Eight creative principles.

My creative philosophy, not an AI methodology. The technology serves these; it never replaces them.

  1. 01Start with the idea.

    Technology is not the idea. A model is not the idea. A visual style is not the idea. Start with what you are trying to make someone think, feel or do. Everything downstream serves that decision.

  2. 02Direct, don’t decorate.

    Every visual decision should serve the idea. Good creative direction is not adding more. It is knowing what belongs and what doesn’t. Restraint is a creative decision.

  3. 03Give the work a world.

    Strong creative work has an internal logic. Colour, typography, composition, movement, sound, characters, environments and language should feel like they belong to the same world. When the world is coherent, the audience stops noticing the craft and starts feeling the idea.

  4. 04Reference everything.

    AI makes references more powerful, not less. The right references communicate things words cannot. A reference is a decision made visible.

  5. 05Separate exploration from execution.

    The first job is to discover what might be possible. The second is to make the strongest possibility real. AI makes exploration cheap; direction makes execution decisive.

  6. 06Make the impossible testable.

    Creative technology is most interesting when it turns an idea that would have been too expensive, too slow or simply impossible into something you can actually prototype. Prototype first, commit second.

  7. 07Protect the human point of view.

    AI can generate infinite possibilities. Infinite possibilities are not a creative strategy. Someone still has to choose. Taste is the strategy.

  8. 08Ship the work.

    Creative direction ultimately ends in a decision. Not another prompt. Not another variation. The work gets finished. Done is a creative decision too.

Taste

The scarcest skill.

AI can generate ten thousand options. Taste is knowing which one is worth pursuing.

Creative direction is partly the ability to recognize a strong idea before it is fully formed. It is knowing when something feels too obvious, too polished, too safe, too familiar, too much, not enough, or exactly right. AI can help generate possibilities. It cannot outsource the responsibility of choosing among them.

The more abundant creation becomes, the more valuable taste becomes.

Process

From idea to execution.

A creative idea is only the beginning. My process usually moves through three stages. The objective is not to make technology visible. The objective is to make the idea visible.

  1. Stage 01Concept.

    What is the idea? What is the audience supposed to feel, understand or remember?

  2. Stage 02Treatment.

    What does the idea look like? What is the visual language? What is the tone? What references define the world? What does the audience experience?

  3. Stage 03Production.

    How do we actually make it? What needs to be filmed, generated, designed, animated, edited or built? Where can emerging technology expand the possibilities? Where does human craft matter most?

Craft

Visual language.

Every project needs its own visual grammar. These decisions create something more important than style. They create identity.

  1. 01Composition.

    What does the viewer see first?

  2. 02Colour.

    What emotional world does the palette create?

  3. 03Typography.

    What personality does the type carry?

  4. 04Motion.

    How does the work move?

  5. 05Photography / imagery.

    What makes the visual world feel specific?

  6. 06Sound.

    What does the audience hear before they consciously notice it?

  7. 07Pacing.

    When does the work breathe?

  8. 08Contrast.

    What makes one moment more important than another?

Practice

How I work with AI.

I don’t use AI as a button that makes finished creative. I use it throughout the creative process.

  1. 01Explore.

    Generate possibilities quickly.

  2. 02Research.

    Build context and understand the problem.

  3. 03Develop.

    Explore visual directions, treatments, scripts and concepts.

  4. 04Prototype.

    Make an idea tangible before committing to full production.

  5. 05Produce.

    Use AI-assisted workflows to create images, video, sound, motion and content.

  6. 06Iterate.

    Generate variations and test directions quickly.

  7. 07Refine.

    Bring the strongest material together through human editing and judgment.

  8. 08Finish.

    Make the final creative decisions.

AI increases the number of doors I can open. Creative direction decides which door I walk through.

Technology

Creative technology.

I am interested in technology for one reason: what does it make possible that wasn’t possible before? I don’t approach these as tools to collect. I approach them as creative materials.

Generative video · AI image systems · Synthetic characters · Node-based workflows · AI-assisted editing · Procedural animation · Generative sound · Interactive experiences · AI agents · New production pipelines

The question isn’t “Which AI tool should we use?” The question is “What could we make if the constraints of the old production model disappeared?”

Underneath

The systems advantage.

My interest is where creative direction and systems thinking meet. Great creative is not only about individual talent. It is also about creating a system where good work can happen repeatedly: briefs, references, moodboards, creative standards, templates, production workflows, approval processes, asset libraries, knowledge, automation, AI. These are not separate from creativity. They are the infrastructure that allows creativity to scale.

The stack

Direction + Context + Workflow + Tools + Evaluation + Judgment.

  1. 01Direction.

    The objective, audience, constraints and definition of success.

  2. 02Context.

    References, source material, brand standards, examples and constraints, designed as production infrastructure.

  3. 03Crew.

    People, models, agents, software: separated responsibilities working toward one outcome.

  4. 04Workflow.

    Research → draft → generate → evaluate → revise → approve.

  5. 05Memory.

    Characters. Brand. Facts. Voice. Decisions. References. Consistency designed into the production.

  6. 06Quality control.

    Standards, checklists, evaluation criteria and human review.

  7. 07Approval.

    Who decides what ships? The final gate remains human. Nothing ships simply because a model said it was finished.

The loop

Every production improves the next.

Define → Contextualize → Direct → Produce → Evaluate → Refine → Approve → Learn

The final step matters. A good production does not simply create an output. It improves the system that creates the next output.

Maturity

Know what level the problem needs.

  1. 01Experiment.

    “Let’s see what this can do.”

  2. 02Repeat.

    “We know how to get a decent result.”

  3. 03Systemize.

    “We can produce this consistently.”

  4. 04Orchestrate.

    “The system knows what happens next.”

  5. 05Direct.

    “We have built a production capability.”

Field notes

Notes from the chair.

A growing record of essays, experiments, and observations on creative direction in the AI era.

Note 01Why AI projects fail before the AI doesMost failures are direction failures: no objective, no context, no standard. The model was never the problem. Note 02The difference between a prompt and a production systemA prompt is a request. A production system is direction, context, workflow, tools, evaluation and judgment. Note 03Context is the new interfaceThe most important design work in AI systems is not the prompt. It is the environment the prompt runs in. Note 04Why consistency becomes harder when generation becomes cheaperFive hundred outputs need architecture, not inspection. Consistency is a systems problem. Note 05What humans should never delegate to an AI agentIntent, taste, accountability, and the final approval. The chair stays occupied. Note 06Why every AI workflow needs an approval gateAutomation should remove repetitive work. It should not remove judgment. Note 07Building a company knowledge base that AI can actually useTrustworthy, structured, accessible: the reference layer every serious AI system needs. Note 08When an AI agent is actually worth buildingNot every workflow needs an agent. Know the level the problem actually requires. Note 09The hidden cost of unstructured organizational knowledgeInformation across documents, emails and people’s heads is a production tax on every AI project. Note 10Why AI adoption is an operating-model problemTools do not change organizations. Direction, workflow and approval do. Note 11The end of the one-prompt workflowAs AI moves into agents and automated workflows, prompting becomes one layer of a larger system. Note 12What happens when the bottleneck moves from production to judgmentWhen making things is cheap, deciding what is worth making becomes the scarce skill.
Manifesto

The Director’s Manifesto.

AI is not the director. AI is the crew.

The director defines the destination. The crew helps build the road.
The director decides what matters. The crew explores the possibilities.
The director establishes the standard. The crew produces against it.
The director protects the story. The crew creates the takes.
The director reviews the work. The crew learns from the decision.

And when the work is ready: the director decides what ships.

That is the point of the chair. Not control for its own sake. Responsibility for the result.

See the method in practice.

Case studies from productions that ran on it.

Selected work →