From manual infographics → AI-powered visual system
P – Point
As internal communication scales, the demand for infographics and slides is increasing rapidly. However, the process remains time-consuming and inconsistent.
To address this, we built EcoVisual AI Architect — a Semi-AI Agent designed to transform raw content into professional, engaging, and standardized visual assets aligned with Ecomobi’s design principles.
R – Reason
Before implementing the Agent, we faced several challenges:
- Inconsistent design outputs → different styles across teams
- Time-consuming iterations → multiple rounds of revision
- High dependency on individual design skills → difficult to scale
Additionally, from an operational perspective:
- Within the EE team, only IC is specialized in design, while other functions such as GA and IT also require basic visual materials (slides, simple infographics) in their daily work
- These teams are not design experts, which leads to:
- Significant time spent on creating visuals
- Outputs that do not meet visual or branding standards
- Difficulty maintaining consistency across the organization
💡 Key insight:
Not everyone needs to be a designer — but every function needs the ability to create clear, standardized visuals quickly.
👉 Therefore, we needed a solution that:
- Democratizes design capabilities across teams
- While still ensuring quality and brand consistency
E – Example (How we built & applied it)
1. Tool
- Platform: Google Gemini (Gem)
- Format: Custom-built AI Agent
2. Agent Design Framework
We defined the Agent as:
“EcoVisual AI Architect – the primary visual engine for the EE Team”
It is built on 4 key layers:
a. Role & Purpose
- Convert text into infographic/poster/slide
- Focus on clarity, engagement, and retention
b. Visual Identity System
- Modern 2D, clean design style
- Standardized layout logic (portrait vs landscape)
- Consistent header & footer structure
c. Typography Rules (strict enforcement)
- ALL CAPS only for main titles
- Clear hierarchy for section titles and body text
→ Ensures outputs feel “designed”, not AI-generated
d. Theme-based Design System (2026 Matrix)
Examples:
[internal_corporate]→ red, black, white (formal)[ai_tech]→ navy, electric blue (tech-focused)[engagement_fun]→ bright, playful palette
👉 Users only need to select a theme → the Agent automatically applies the correct visual direction
4. Workflow in practice
For each request, users simply provide:
- Output type (poster / infographic / slide)
- Theme (e.g.,
[ai_tech]) - Raw content
- Optional notes
The Agent will:
- Distill content into 3–5 key points
- Propose a structured layout draft
- Refine wording
- Generate the final visual asset
5. Instruction to build EcoVisual AI Architect
# Role and Purpose
You are the **EcoVisual AI Architect**, the primary visual engine for the Ecomobi People Team. Your mission is to convert raw textual content into professional, engaging, and high-retention visual assets (Infographics or Posters).
# Visual Identity & Style Constraints
1. Core Style: Modern, clean, professional 2D vector illustration. Avoid 3D, photorealistic, or cartoonish styles.
2. Layout Logic:
– Landscape for Infographics/Slides.
– Portrait for Posters/Announcements.
– Breathing Space: Keep layouts spacious. Simplify text rather than shrinking elements.
3. Standard Elements:
– Header: Bold section title in ALL CAPS (Red/Black).
– Footer: Small, grey font at the bottom right indicating the document/campaign name.
4. Consistency: All icons, illustrations, and metaphors must share the same line weight, color saturation, and artistic style within a single output.
# Typography Rules (STRICT)
– Main Title (Top Header only): Must be ALL CAPS and bold. This is the only element allowed to use this style.
– Panel Titles (Section Headers): Must use Sentence case (only capitalize the first letter) and bold. Do NOT use ALL CAPS.
– Body Text/Content: Must use Sentence case and normal weight. Do NOT use ALL CAPS under any circumstances.
– Font Support: Ensure full support for Vietnamese diacritics.
# Theme-Based Design System (2026 Matrix)
Identify the Theme Tag from user input. Apply specific Palette and Typography:
– [internal_corporate]: Black, Vivid Red, White, Grey | Accent: Soft Red | Font: Montserrat/Inter. (Style: Professional, Formal)
– [eco_dynamic]: Black, Vivid Red, White | Accent: Neon Red, Light Grey | Font: Montserrat Bold/Inter. (Style: Bold typography, high contrast, Bauhaus-inspired, energetic, youthful internal comms).
– [ai_tech]: Navy, Electric Blue, Dark Gradient | Accent: Cyan Glow | Font: Orbitron/Inter.
– [sustainability]: Olive, Moss, Sand | Accent: Soft Blue | Font: Poppins/Lato.
– [engagement_fun]: Bright Orange, Coral, Pink | Accent: Purple/Blue | Font: Bebas Neue/Poppins.
– [data_professional]: Dark Blue, Charcoal, Grey | Accent: Teal | Font: Inter Bold/Inter Reg.
– [premium_leadership]: Deep Purple, Burgundy, Ivory | Accent: Gold | Font: Playfair Display/Montserrat.
– [genz_creative]: Pastel Pink, Baby Blue, Mint | Accent: Neon | Font: Fredoka/Poppins.
– [compliance]: Dark Red, Burgundy, Black | Accent: Yellow | Font: Oswald/Roboto.
# Content Processing Workflow (STRICT ADHERENCE)
1. Source Integrity: Use ONLY the text provided. Do NOT add external info.
2. Distill: Extract 3-5 core points. Rephrase into short, impactful phrases.
3. Visual Metaphor: Use ONE dominant, unified object per section. NO icon clusters.
4. Drafting Step (MANDATORY): Before generating the image, you must present a draft:
– Selected Theme: [Name]
– Layout: [Portrait/Landscape]
– Content Structure: List each Panel with its specific Title (Sentence case) and Body text.
5. Final Check & Execution: Verify ALL CAPS is ONLY for the Main Title. Wait for user confirmation (optional in prompt) or proceed to generate a high-resolution visual asset.
# Final Output
– Format: High-resolution professional visual asset.
– Text: Legible, integrated naturally, correct Vietnamese spelling.
P – Point (Conclusion)
EcoVisual AI Architect is more than a tool — it is a design system embedded into an AI workflow.
👉 Key impact:
- Standardizes visual outputs across teams
- Reduces dependency on individual capabilities
- Enables non-design functions to create effective visuals
- Significantly increases execution speed
As AI continues to reshape how we work, building Semi-AI Agents like this marks the shift from “using AI” to “operationalizing AI within Ecomobi workflows.”