Back to articles
Article · AI & Creative WorkflowMichele Meloni7 min read

AI in the creative process: where it helps and where it does not

A practical framework for using AI to accelerate research and production while retaining direction, control and responsibility.

Published · 4 September 2026Updated · 4 September 2026

Speed does not automatically create quality.

AI can broaden alternatives, synthesise material and automate repetitive steps. Its value still depends on the context supplied, the criteria used to select an outcome and verification in the real deliverable.

The prompt is one step, not the method.

Promptario grew from the need to make the work around generation easier to understand and reuse: preparing context, exploring alternatives, selecting with explicit criteria and verifying the outcome in its final channel.

  1. 01Context

    Goal, audience, materials and constraints.

  2. 02Exploration

    Genuinely different alternatives, not superficial variations.

  3. 03Selection

    Comparison against declared, traceable criteria.

  4. 04Verification

    Human review in the real format and journey.

Promptario website interface with a catalogue of resources for working with artificial intelligence
Promptario · from a single prompt to a system for discovery, use and verification.
01

Define the problem before the prompt

Goals, audience, constraints, available material and quality standards matter more than a clever prompt formula.

Practical actions

  • State the desired outcome in one testable sentence.
  • Separate known facts, assumptions and missing information.
  • Provide examples and anti-examples for specific directions.
02

Use AI to explore

Exploration helps compare structures, narrative angles and variants. Do not generate dozens of options without criteria for selecting them.

Practical actions

  • Request alternatives that are genuinely different.
  • Evaluate every proposal against the same criteria.
  • Keep the decisions, not every generation.
03

Keep editorial control

Tone, hierarchy and precision require human review. Output must fit the context, not simply be grammatically correct.

Practical actions

  • Check sources, names, numbers and attribution.
  • Remove generic language and repeated formulas.
  • Verify accessibility and rights for used material.
04

Validate in the real channel

Copy must be read in the page, an image in its final format and an interface in its journey. The model response is not the delivery.

Practical actions

  • Test intended sizes, states and devices.
  • Compare the result with the original objective.
  • Keep a record of meaningful revisions.

Responsible workflow check

  • The task and criteria exist before generation.
  • Sensitive data and confidential material remain protected.
  • Sources and claims are verified.
  • A person owns the final selection.
  • The result is tested in its intended context.

Key principle

AI expands the process; it does not replace direction.

Quality comes from combining generative capability, design criteria and human control. Automation without governance only makes mistakes faster.