How AI Supercharged Our Design Workflow for High‑Impact Delivery Campaigns

In early August, our team received a business request for two delivery‑service campaigns – “Express Thursday” and “Crazy Weekend Shipping.” Data analysis showed Thursday’s peak stemmed from recipients wanting timely delivery, while the weekend peak was driven by users’ shipping time and location preferences. The goal was to attract users through activity operations, increase coupon redemption, and boost order volume.

Because the schedule was tight, with two themes per week and regular activity updates, we decided to use AI‑assisted design to improve efficiency. AI has become a hot topic in the design community; after months of learning and small‑scale trials, we applied it fully to this project and saw clear user‑data improvements.

We share the four‑step process we followed: design decomposition, AI usage, design adjustment, and collaborative consolidation.

Design Decomposition

AI generation ≠ no thinking

With AI widely used in design, opinions differ. Some see it as a productivity boost that lets designers and non‑designers quickly create appealing visuals, while others fear being replaced. In real projects, specific requirements and constraints make it harder than simply prompting a tool.

We first analyzed business needs and translated them into design concepts, allowing AI to provide more thinking time and experimentation.

For the two campaigns:

Express Thursday : Positioned in a business office scenario to convey professionalism and brand strength. Visuals emphasized “door‑to‑door pickup,” “coupon benefits,” “insured service,” and “on‑time delivery.”

Crazy Weekend Shipping : Targeted a warm home setting for younger users, using vibrant dopamine colors and popular themes such as shopping, gaming, romance, and pets, with a “check‑in” visual motif.

AI Usage

AI as a designer’s tool

Using Midjourney, we generated two series of visuals quickly, though we faced challenges that we resolved as follows.

Stable character output

Initial attempts produced inconsistent characters. We created a localized reference library and saved usable images as “base images.” The AI learned preferences, gradually stabilizing facial features, age dynamics, and visual style.

We established a stable keyword prompt:

a Chinese young man in a shirt, holding a box, being happy, broad smile, clean background, clay, blind box toys, bryce 3d, best quality, 4K

, then varied the action description for each scene.

Scene‑person relationship

Combining characters with complex backgrounds proved time‑consuming. We split background and foreground generation, then composited them, ensuring consistent style and faster iteration.

Design Adjustment

Post‑processing is essential

After generating assets, we added brand elements—brand colors, logos, and contextual items like parcels, envelopes, and courier uniforms—to align visuals with the product identity.

Injecting small creativity

In the weekend‑shipping pet theme, we replaced a thumbs‑up icon with a cat paw, adding a playful touch that made the visual more lively.

Collaboration Consolidation

Team synergy

Team members collaborated on prompts, shared reference images, and standardized parameters, building an AI asset library that ensured a unified style across all visuals. This collaborative workflow has been extended to other projects, continuously improving design efficiency.

Conclusion

AI has dramatically increased our design efficiency, freeing designers’ “hands” while reminding us to retain independent “brain” thinking. Even as AI handles routine tasks, creative and critical thinking remain indispensable for continual innovation.