From ¥200 to ¥150K Saved: A Complete Enterprise Local AI Cost-Reduction Story

From ¥200 to ¥150K Saved: A Complete Enterprise Local AI Cost-Reduction Story

📖 Glossary

AI Box (also known as Agent Computer / Agent PC), is a dedicated local hardware device that runs AI Agents. Pre-installed with an AI agent management system, plug-and-play, running 24/7. Users can remotely command AI to work via Discord, Slack, Telegram, WhatsApp, and more.

A real user's cost ledger. A 45-person e-commerce agency deployed Nizwo D1 in Q4 2025. By Q1 2026, their AI spending dropped from 十余元K/month to ¥3.1K — saving over ¥100K annually.

Background

"Weijing E-commerce" (pseudonym) is a 45-person agency managing Tmall and Douyin stores for brands. Their AI needs are extremely high-frequency: product copywriting, customer service scripts, competitor price monitoring, livestream scripts, and data reporting.

Their original setup: ChatGPT Plus for all ($20/month/person) + Jasper Pro for the content team ($99/month/person). Monthly AI spend: ~十余元,000.

Three-Phase, Five-Month Migration

Phase 1: Pilot (October 2025)

Three content team members switched from Jasper to one Nizwo D1.

Results: Chinese copy quality slightly better than Jasper. The biggest change: no more "rationing" usage. Unlimited local generation meant they stopped hoarding requests.

Month 1 savings: ¥2,100.

Phase 2: Scale (Nov-Dec 2025)

Deployed three D1s (content, operations, customer service). Kept only 5 ChatGPT Plus subscriptions. The biggest challenge wasn't the AI — it was that employees didn't know how to use it. Only 5 out of 45 could write decent prompts.

Solution: The CTO spent two weekends creating 50 Prompt templates covering 5 business scenarios. After templating, usage jumped from 15% to 70%.

Phase 3: Automation (Jan-Mar 2026)

Built three automated workflows: daily competitor price monitoring, pre-livestream script generation, and auto-updating customer service knowledge base.

Cost Trajectory

Annual savings: ~¥107,000. Minus ¥18,000 for three D1s. Net savings: ~¥89,000.

Three Takeaways Beyond Cost

1. Data security peace of mind: Product data, pricing strategy, customer info — all stay inside the network. "We stopped asking 'can I send this to the AI?'"

2. Usage frequency exploded: From 15-20 copy generations/day (metered) to 50-60/day (unlimited). "Creativity tripled when it stopped costing money."

3. Prompt capability internalized: From 1 person to 10 people writing quality prompts in 3 months. "That skill is worth more than the cost savings."

Advice for SMEs

The CTO's three rules: 1. Don't chase ROI on Day 1. Let one or two people find their "aha moment." ROI follows. 2. Template prompts beat training. Write the templates. They fill in blanks. 3. Calculate a 3-year ROI, not 3-month. The upfront cost pays back in Year 1. Years 2-3 are pure savings.

The Case Studies column collects real AI deployment stories. Have one to share? Contact us.