← Back to projects
Self-directed concept · AI creative workflow

AdForge Creative Ops

A working AI-assisted creative-operations tool that turns a product, audience and goal into three structured ad directions, then keeps human review and selection in the loop before production.

AdForge Creative Ops AI-assisted concept generation workflow
Live interactive concept built as portfolio evidence, not client work.
Problem

A realistic product problem, not a decorative dashboard brief.

Teams can generate a large volume of AI copy or imagery quickly, but volume alone does not create a useful creative system. The product problem is to structure generation around a real brief, make outputs comparable, preserve human judgment and avoid turning AI suggestions into unsupported performance claims.

My role

Product framing, interface design and working prototype.

I designed the brief-to-review workflow, implemented the product interface and wired a real backend AI generation route with structured output, validation and a visible failure state.

Core workflow

The interface is organized around decisions and state changes.

The concept is intentionally interactive so the design can be judged against a real task flow instead of only a polished screenshot.

01 / BRIEFDefine product, audience and goal

The generator starts from three business inputs instead of an open-ended prompt box.

02 / GENERATECreate distinct structured directions

The backend returns exactly three concepts with hook, promise, visual system and CTA fields so outputs can be reviewed consistently.

03 / REVIEWCompare concepts side by side

Cards keep the strategic fields aligned, making differences easier to judge than free-form AI text.

04 / SELECTMove one direction to production

Selection exposes a handoff state rather than treating generation as the final deliverable.

Design decisions

System choices are tied to operational use.

The goal is not to claim invented conversion or efficiency gains. The case shows the reasoning and interaction model that can be evaluated directly in the live prototype.

01 / DECISIONAI is a collaborator, not the claim

The interface explicitly frames generated concepts as starting points and avoids invented metrics, endorsements or product capabilities.

02 / DECISIONStructured output improves review

A fixed schema makes each angle comparable and production-useful while still allowing creative variation.

03 / DECISIONFailure stays visible

Required-field validation and a backend error state prevent silent failure and keep the workflow recoverable.

What this demonstrates

Client-relevant product thinking without pretending this was paid work.

AI product UXStructured generationCreative operationsHuman-in-the-loop reviewBackend AI integrationGuardrail designProduction handoff