# Bosun Liu

Bosun Liu’s posts describe work helping clients with cross-border logistics between Uzbekistan and China, alongside practical commentary on AI costs, model use, and creative production workflows.

- Profile: https://b2binfluencer.marketing/creators/bosun-liu/
- LinkedIn: https://www.linkedin.com/in/bosun-liu-b561b5135
- LinkedIn followers: 2,384
- Topics: artificial-intelligence, supply-chain, creative-testing
- Based on 12 dated posts, latest 2026-10-05

## Summary

They are most useful for practical commentary on AI costs, model use, creative production workflows, and cross-border logistics between Uzbekistan and China.

- Best for: Fits a B2B AI consultancy wanting AI cost and creative production thought leadership for e-commerce and logistics teams.
- Not for: Consumer skincare launch in North America needing short-form video demos, broad consumer reach, paid amplification, tracked conversion.
- Check first: Confirm sponsored or co-authored work, disclosure and review terms, and recent post reach.

### Takeaways

- AI needs a bill of materials before it needs a strategy.
- The AI bottleneck is access, cost and continuity, not adoption.
- Judge AI production by cost per usable output, not by model quality.

## Main ideas

- **AI needs a bill of materials before it needs a strategy**: Bosun Liu contrasts two kinds of company spending. When a shipment moves, every dollar has a name—ocean freight, terminal handling, customs brokerage, last-mile—and procurement signs off line by line. Ask the same company what it spends on AI each month and the answer is usually a shrug. Finance teams already know how to itemise costs; they simply have not been brought into the AI conversation. Within a week, a team can ask three questions: how many AI subscriptions is it paying for, including those expensed on personal cards; who owns each subscription and can explain its purpose; and would anyone be alerted if usage doubled next month? The first step is not an AI strategy but a bill of materials.
- **The AI bottleneck is access, cost and continuity, not adoption**: After searching for "Claude usage limit", Bosun Liu reads the results as a market signal, not a complaint thread. Users are not saying the tool is bad; they are saying it is good enough to become part of daily work, making a usage limit a business problem rather than an annoyance. Teams are already planning work around reset windows, learning to avoid token waste, switching between Claude, ChatGPT and Gemini when a model caps out, and in some cases building tools to move context between models. The next constraints are access, cost and workflow continuity. Teams need to route each task to the right model, keep context intact across switches and control spend once AI becomes part of daily operations.
- **Judge AI production by cost per usable output, not by model quality**: Bosun Liu argues that generated content is a cost problem as much as a creativity problem, especially in AI video. A finished video ad rarely comes from one prompt: it passes through idea, script, image generation, style testing, image-to-video, multiple shots, voiceover, music, editing, regeneration and more variations. Most of that output is drafts. The metric that matters is how much it costs to reach one usable video. If every test is expensive, experimentation slows; if every variation costs too much, teams stop trying. Liu proposes a pipeline that assigns different jobs to different models: concept images, product shots, image-to-video, voice, music, captions and editing support. Choose each for speed, flexibility and cost at real production volume.
- **Make creative failure affordable instead of asking whether AI replaces creators**: This post rejects the question "can AI replace a real creator?" in favour of another: how many bad creatives can you afford to test before finding a winner? Human UGC is expensive before an idea is validated: creators are paid, footage is awaited, scripts are revised, reshoots are requested, several angles are tested and most fail. That failure rate is normal but costly. AI UGC changes the arithmetic, not the goal: one product can become dozens of variants across hooks, faces, voices, locations, scripts and emotional angles. Most will still fail, but failure gets cheaper, giving teams an advantage in testing more ideas and killing losers faster before the budget runs out. Liu says this is not about replacing human creators completely.
- **Reliable partners and long relationships are the priority in cross-border logistics**: The operating base behind the AI commentary is practical trade work between China and Central Asia. In these posts, Liu prioritises finding a partner for reliable international logistics from Uzbekistan to China, with door-to-door service, and maintaining that relationship. They describe visiting two partners they have worked with for several years, calling the team powerful and efficient, and separately stopping on the way back from Xinjiang to meet a client who owns yarn factories in Uzbekistan and receives help with cross-border logistics. The details are operational—factory ownership, door-to-door service and multi-year relationships—rather than commentary from outside the trade. The same cost discipline shapes the AI posts.

## Work to read

- [Nobody budgets for AI the way they budget for freight](https://www.linkedin.com/posts/bosun-liu-b561b5135_nobody-budgets-for-ai-the-way-they-budget-activity-7485222643458105344-1M-u)
- [Searching "Claude usage limit" and reading it as a signal](https://www.linkedin.com/posts/bosun-liu-b561b5135_i-searched-claude-usage-limit-today-the-activity-7478429313273536512-dKbG)
- [AIGC is a cost problem, not only a creativity problem](https://www.linkedin.com/posts/bosun-liu-b561b5135_aigc-is-not-only-a-creativity-problem-it-activity-7472546192954945536-wpjf)
- [Most brands are asking the wrong question about AI UGC](https://www.linkedin.com/posts/bosun-liu-b561b5135_most-brands-are-asking-the-wrong-question-activity-7474723079244378112-HmnX)
- [Finding a reliable logistics partner for Uzbekistan to China](https://www.linkedin.com/posts/bosun-liu-b561b5135_its-very-important-to-find-a-reliable-partner-activity-6796321542356762624-BfHa)
- [Field note from a client visit: yarn factories in Uzbekistan](https://www.linkedin.com/posts/bosun-liu-b561b5135_amazing-views-along-the-way-back-from-xinjiang-activity-6797036134858936320-JCyZ)

## Partnerships

Bosun Liu's posts offer operating commentary on cross-border logistics between Uzbekistan and China, and on the costs and production side of AI tools for e-commerce and agency teams, rather than pitches to brands. They don't describe sponsorship offers, rates, media kits, or partnership formats. For campaign planning, the posts suggest they work with clients on cross-border supply chains and write in detail about AI production economics, model usage limits, and creative testing costs. Brands focused on AI cost discipline, creative pipeline efficiency, or trade and logistics operations may find a plausible subject-matter fit, though the posts don't define an ad format. Before committing budget, ask whether they take sponsored content, which formats they offer, whether disclosure is routine, and what turnaround and review involve.

## Alternatives

- John Soriano: Choose John Soriano when the series needs to show a team how to cut waste in creative spend it has already committed, not how to budget new AI production. His two posts tackle the same buyer task from opposite ends of the pipeline: diagnose a live ad before switching it off, using hook rate, hold rate, click-through and conversion to pinpoint whether the problem is the opening three seconds, the middle, the offer or the landing page. That's an exercise the consultancy's e-commerce readers could run next week. A second post offers a five-cut re-edit routine that turns one paid-for UGC clip into five testable ads by swapping the opener, leading with problem, product or proof, and rewriting captions to match. If the angle is cost discipline through reuse of existing assets rather than model routing, Soriano is the more directly useful practitioner. He reports doing the work at volume: 45 variations a week for a DTC apparel brand.
- Grace Eva: Choose Grace Eva when the reader's task is tool selection rather than cost accounting: deciding which AI avatar, voiceover, script and product-to-video tools belong at each stage of a UGC-style ad production process. Both posts point readers into that territory: one covers making a first AI UGC ad without a creator, camera, lights or a recorded clip; the other lists the top AI tools for UGC-style creative in 2026, organised by tool type, then lays out a sequence from idea through script, video, testing and optimisation. If the consultancy wants an instalment that helps marketers assemble a production stack, this is closer to the deliverable than anything in Liu's posts.
- Nathan Smith: Choose Nathan Smith when the series wants an operator-founder voice and a supply-chain frame rather than cost accounting. His account of building Arrim to six-figure ARR solo in under four months argues that generation is commoditising and durable value lies in proprietary data, outcomes and compounding learning. That overlaps with the consultancy's defensibility-adjacent positioning. The second post makes the mapping almost literal: scaling brands treat creative like inventory, forecasting demand weeks ahead, measuring how many new concepts entered testing rather than ROAS per ad, decoupling production from performance data, and accepting a low hit rate by design. For a logistics-and-e-commerce audience already fluent in inventory and cadence, that framing transfers with little translation.

---
AI-assisted summary of public posts. Audience demographics are not available. Engagement appears only where we checked recent posts. Method: https://b2binfluencer.marketing/methodology/