Beyond Prompting: Building a Production Capability That Scales

AI has firmly established itself within creative production. What began as experimentation is quickly becoming part of everyday workflows, helping teams generate concepts, create artwork, adapt content and accelerate production in ways that were difficult to imagine only a few years ago.

But as organisations move from testing AI to embedding it into day-to-day delivery, the conversation is changing.

As we explored in our previous article, Prompting is Production the Work Nobody Planned For, prompting is no longer a simple instruction sitting at the start of a workflow. It has become a production layer that directly influences quality, consistency, review cycles and delivery timelines. The challenge is no longer recognising that prompting matters. The challenge is building it into the way creative teams operate.

Writing a good prompt is useful. Building an organisation that can prompt consistently, adapt as models evolve, and understand how different AI tools respond to different types of instruction is transformational.

The goal isn't to create one perfect prompt formula and expect teams to use it forever. The prompting landscape is changing quickly, and different models have different strengths, behaviours and ways of interpreting instructions. What works particularly well for one model may produce a very different result in another.

That means prompting capability needs to be built for change. Teams need shared principles and production standards, but they also need the knowledge and flexibility to understand how those principles should be applied across different models and tools.

In other words, consistency does not mean doing exactly the same thing every time. It means having a consistent approach to quality, intent, governance and production outcomes, while being able to adapt the method used to achieve them.

For Creative Operations, that distinction is important. The objective isn't to make everyone follow the same prompt template regardless of the technology. It is to give teams a common production framework that can evolve as the technology does.

The capability isn't the prompt. It's knowing how, when and why to adapt it.

The conversation needs to move beyond prompt writing

There is no shortage of advice on how to write better prompts.

Most of it focuses on wording, structure or techniques that improve individual outputs. While those are valuable skills, they only address part of the challenge.

Enterprise creative production isn't built around individual outputs.

It is built around repeatability.

A campaign doesn't consist of one image. It becomes hundreds of assets across channels, markets, languages and formats. Copy evolves into social content, email campaigns, digital advertising and web experiences. Creative concepts are adapted, reviewed, localised and refined continuously.

In that environment, the question isn't whether one person can write an effective prompt.

The question is whether an entire organisation can produce consistent results, regardless of who is prompting, which team is using AI, which model sits underneath the workflow, or how the work moves between different stages of production.

This is where node-based workflows become increasingly important. Rather than relying on a single prompt to produce a finished output, teams can build structured workflows where individual stages handle specific tasks, from input and creative direction through to generation, refinement, validation and output. Each node can have its own instructions, rules and controls, creating a more repeatable process than relying on individual users to manage every step manually.

For creative studios, this can be particularly powerful when producing large volumes of artwork or variations. A workflow might take an approved creative brief, apply a defined visual direction, generate an image, check it against specific requirements, create the required formats and then pass the outputs into the next stage of production. The exact tools and models may change, but the underlying production logic remains consistent.

This also changes how we think about prompting. The goal isn't necessarily to create one perfect prompt that does everything. It is to design a series of prompts, instructions and decision points that work together as part of a controlled workflow.

From ICP's perspective, this is an important distinction. Prompting capability is not just about what an individual asks an AI model to do. It is about designing how AI participates in the production process.

That means considering prompts, models, node-based workflows, quality controls and human review as connected parts of the same production system.

The result is greater consistency without forcing every task into the same formula. Teams can adapt the model or workflow to the job while maintaining the production standards that matter.

Good prompting starts long before anyone opens an AI tool

One of the biggest misconceptions about prompting is that quality begins with the instruction itself.

In reality, the prompt is simply the final expression of the planning that came before it.

Creative teams already know how to brief photographers, designers, production partners and agencies. They define campaign objectives, audiences, brand principles, deliverables, production constraints and approval criteria before any creative work begins. A production-ready prompt is built on more than a well-written instruction. It brings together the context, creative intent, constraints and references that a person or production partner would need to do the work properly.

• What is the objective of this asset?
• Where will it be used?
• Who is the audience?
• What must remain consistent across every version?
• What production constraints need to be respected?
• What does success look like?
• Who owns prompt quality?
• Where are approved prompts stored?
• How are successful prompts documented?
• When should prompt frameworks be reviewed?
• How do we prevent prompt duplication across teams?
• How do we measure whether prompting is reducing production effort over time?
• shared prompt libraries
• documented prompting standards
• campaign prompt templates
• version control for prompt frameworks
• governance around approved prompts
• prompt quality reviews
• training and enablement
• continuous optimisation based on production outcomes

This is also where examples become particularly valuable. Creative teams already understand the importance of showing rather than simply telling. When briefing a designer, photographer or agency, saying “make it premium” is rarely enough. You provide references, previous work, moodboards, examples of what you like and, often just as importantly, examples of what you don't.

The same principle applies when prompting AI.

A prompt describing a visual direction can provide context, but examples can remove a layer of interpretation that words alone cannot always achieve. If the desired artwork should have a particular composition, lighting style, level of realism, colour treatment or editorial feel, showing examples helps establish a much clearer creative reference point.

For example, rather than asking an AI model to create “a premium beauty campaign with soft, editorial lighting,” a studio could provide examples of previous campaign imagery alongside the instruction, explaining what should be carried forward from those references and what should change.

The important point is that examples should not simply be added as inspiration. They need to be part of the brief.

What should the model take from the reference? What should it ignore? What needs to remain consistent?

This is particularly important for creative production because the same words can produce very different interpretations. “Premium,” “minimal,” “editorial” or “cinematic” mean different things to different people, and they can mean different things to different AI models too.

At ICP, we see references as another form of production language. They help translate creative intent into something more concrete and give teams a shared point of reference, whether the work is being produced by a person, an AI model or a combination of both.

The strongest prompts combine clear instructions with clear examples. Just as they would in a traditional creative brief, the words establish the objective and constraints, while the references help establish what the intended outcome actually looks and feels like.

Consistency is the metric that matters

One successful prompt proves very little. Consistent prompting across multiple teams, campaigns and markets proves everything.

This is where many organisations experience growing friction. One designer develops prompts that consistently produce excellent campaign visuals. Another team generates noticeably different results for the same campaign. Regional adaptations begin drifting from the approved creative direction. New team members start from scratch because proven prompts exist only in someone else's chat history.

. The technology may have changed. The operating model hasn't necessarily kept pace.

Creative Operations has spent years building consistency through brand guidelines, design systems, workflow governance and production standards. Prompting should be treated no differently.

From ICP's perspective, consistency is an important measure of AI maturity, but consistency should not mean rigidity. If prompting depends entirely on individual expertise, organisations introduce unnecessary variation and risk into production. If prompting is standardised, documented and shared, AI becomes a repeatable capability rather than an unpredictable experiment. But those standards also need to be flexible enough to evolve as models, tools, workflows and creative requirements change.

The goal is not for everyone to follow the exact same prompt or workflow. It is to give teams a consistent set of principles, quality standards and production guardrails, while allowing them to pivot when the technology or the task demands it.

The strongest prompting frameworks create consistency in the outcome, not necessarily in the method used to get there.

That balance matters. Creative teams need enough structure to protect quality and brand consistency, but enough flexibility to experiment, adapt and take advantage of new capabilities as they emerge. In a rapidly changing AI environment, the ability to adapt is not a departure from operational maturity. It is part of it.

Consistency isn't restrictive. It's what allows creativity to scale with confidence.

Stop writing prompts. Start building prompt frameworks.

One of the biggest shifts organisations need to make is thinking about prompts as long-term production assets rather than one-off conversations.

During experimentation, it is perfectly reasonable to write a new prompt for every task.

That approach doesn't scale.

The studios seeing the greatest operational gains are building reusable prompt frameworks that provide consistency without becoming rigid.

These frameworks establish the principles, context and quality standards that should remain consistent, while allowing teams to adapt their approach to the model, tool, task and creative output they are working with. As models change, those frameworks can evolve with them rather than forcing teams to rebuild their production approach from scratch.

Instead of creating prompts from scratch, they develop structured assets that can be refined, reused and continuously improved.

Those frameworks often include:

Creative Direction Prompts

Defining campaign mood, visual language, brand personality and creative principles.

Production Prompts

Ensuring artwork is suitable for multiple formats, aspect ratios and delivery channels.

Adaptation Prompts

Supporting localisation, market variations and content repurposing while protecting approved creative direction.

Revision Prompts

Allowing teams to improve outputs without changing composition, tone, product representation or other approved elements.

Quality Assurance Prompts

Checking assets against agreed standards before they enter review.

Over time, these frameworks become part of the organisation's intellectual property. They reduce dependency on individuals, shorten onboarding for new team members and create greater consistency across internal teams and external partners.

Most importantly, they allow organisations to improve continuously rather than starting from zero every time a new campaign begins.

Prompting should be part of production planning

Perhaps the biggest operational shift is recognising that prompting is work.

Many project plans still assume AI generation happens almost instantly. Someone opens a tool, writes an instruction and production continues.

The reality is very different.

Building prompt frameworks, testing outputs, refining instructions, documenting successful approaches and maintaining prompt libraries all require time and ownership.

They should be planned in exactly the same way organisations already plan for briefing, concept development, production, quality assurance and stakeholder reviews.

That means asking operational questions such as:

These aren't AI questions, they are Creative Operations questions, and they deserve the same governance as every other production discipline.

Building prompting into Creative Operations

As prompting becomes embedded within creative delivery, ownership becomes increasingly important.

Without defined standards, every designer, creative and marketer develops their own approach. Some prompts produce exceptional work. Others create unnecessary variation, inconsistent quality and repeated review cycles.

The answer isn't asking everyone to become an AI specialist.

It's creating an operating model that supports everyone.

At ICP, we encourage organisations to build prompting into their Creative Operations strategy by treating it as a recognised production capability.

That includes:

This mirrors the way creative teams already manage design systems, brand governance and production workflows.

The objective isn't simply better prompts. It's a better production system.

The next stage of AI maturity

The first wave of AI adoption was about understanding what the technology could do.

The next wave is about building the operational capability to use it consistently.

Organisations that succeed won't necessarily have access to every AI model, they'll have better operating models.

They'll understand how prompting fits into Creative Operations, how it supports governance, how it protects consistency and how it reduces production effort over time.

Most importantly, they'll recognise that prompting isn't an invisible task sitting alongside production.

It's becoming part of production infrastructure itself.

Final thoughts

Our previous article introduced the fact that prompting has become production because it now carries much of the work traditionally spread across briefing, art direction, refinement and quality control.

This is the next step in that conversation.

If prompting is production, it should be treated like every other production capability. It should be planned, documented, governed and continuously improved. It should have ownership, standards and measurable outcomes.

At ICP, we help organisations move beyond AI experimentation by designing the operational frameworks that allow AI to deliver real business value. That means creating repeatable prompt systems, embedding prompting into Creative Operations, establishing governance and building the foundations that enable teams to scale confidently.

Because the organisations that realise the greatest value from AI won't simply be those generating more content.

They'll be the ones building better systems around how that content is created