How a Design Company Cut Outsourcing Costs by 60% with Local AI
A 30-person UI design firm in Guangzhou deployed Nizwo devices for local AI in late 2025. Six months later, their monthly outsourcing spend dropped from ¥80K to ¥32K. Here's the real story.
Background
"Jianjie Design" (pseudonym) is a B2B UI design company with 30 designers and ~¥15M annual revenue.
Their pain point: spiraling outsourcing costs. Every project needed copywriting, competitive analysis reports, and design specification documents — work designers didn't want to do.
Monthly outsourcing costs: ~¥80K. Annually: nearly ¥1M.
The Trigger
Founder Chen saw a Nizwo demo at a design conference in October 2025. His immediate thought: "This could replace 70% of our copywriting outsourcing." He bought three Nizwo A1 units within a week.
Why local AI over cloud? Three reasons: 1. Client design files and business info cannot leave the device 2. Outsourced copywriting requires multiple revision rounds — cloud AI per-token pricing adds up 3. Designers' machines are already maxed out running design software
Implementation Timeline
Phase 1: Ramp-Up (Weeks 1-2)
- Designers learned to use AI for copywriting
- Chen created 10 Prompt templates for common document types
- Started with 2 early adopters, grew to 8 by Day 10
Phase 2: Efficiency (Weeks 3-8)
- Competitive analysis: outsourced 3 days → AI draft 1 hour → designer polish 30 min
- Design specs: templated prompts → one-shot output → minor edits
- Client proposals: AI-generated outline + copy → designer formats only
Monthly outsourcing: ¥75K (Month 1) → ¥48K (Month 2) → ¥32K (Month 3)
Phase 3: Automation (Weeks 9-24)
- Daily competitor design update briefing auto-pushed to WeCom
- Automated new-hire training manual generation
- Customer feedback auto-classification and prioritization
AI is no longer a "tool" — it's a team member.
Six-Month Results
| Metric | Before | After | Change |
|---|---|---|---|
| Monthly outsourcing | ¥80K | ¥32K | -60% |
| Competitive analysis cycle | 3-4 days | 0.5 day | -87% |
| Document delivery cycle | 2-3 days | 2 hours | -95% |
| Designer per-capita output | 100% | 135% | +35% |
Three Takeaways from Chen
- "Fix the most painful problem first. Don't boil the ocean." — They started with one scenario. Neighboring scenarios followed naturally.
- "Templated Prompts are the shortest path to adoption." — Designers aren't prompt engineers. Let them fill in blanks, not write from scratch.
- "Local AI ROI is calculable. Cloud isn't." — Three A1s cost under ¥10K. Six months saved ¥280K in outsourcing. Cloud AI's hidden cost is unquantifiable data security risk.
Why This Case Matters
Jianjie sits in the sweet spot for local AI: outsourcing costs too high to ignore, but not high enough for a full-time copywriting team; data sensitivity exists but at regulatory edge; team has learning willingness but limited time. This "in-between" state is exactly where local AI delivers the highest ROI.
Have a local AI case study to share? Contact us. Nizwo continues to track real-world AI deployment stories.