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.
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.
Key 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.
Check first: Confirm sponsored or co-authored work, disclosure and review terms, and recent post reach.
Main ideas
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
Bosun Liu’s posts cover two connected areas: cross-border logistics between Uzbekistan and China, and practical commentary on AI in business.
The logistics posts describe visits to clients and long-standing partners; the AI posts focus on the costs and operating choices that arise when teams rely on models and generated creative work.
Ask whether Liu has signature books, shows or other work that would explain a wider public profile if that matters to your campaign.
Bosun Liu turns AI talk into operating questions.
In one post, a freight budget becomes a model for tracking AI subscriptions, assigning owners and setting alerts when usage rises.
Others shift the focus from model quality to the cost of producing a usable video, or from replacing creators to making creative tests cheaper.
Practical steps and questions make the arguments concrete.
Logistics posts add context, grounding their perspective in client visits and partner relationships along the Uzbekistan–China corridor.
Follow Liu for trade context and practical thinking about AI costs and production.
Ask for a broader name-search picture if signature books, shows or businesses matter to your campaign.
Read together, Bosun Liu's posts make one argument: the interesting question about AI is not how clever the models are, but what it costs to get usable work out of them.
Read more
Model limits, token spend, half-finished video drafts and forgotten subscriptions all become production or procurement line items.
The argument draws on a working day spent moving goods between China and Central Asia, which informs both the cost-discipline habit and the audience the posts seem written for.
The priority is control: know what you are paying for, route work to the right model and design a pipeline that holds up through usage limits and cheap failures.
AI needs a bill of materials before it needs a strategy
Bosun Liu contrasts two kinds of company spending.
Read more
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.
Freight line items signed off by procurement, set against a few chat seats, a design subscription on someone's personal card and an API bill that doubled last quarter without anyone noticing.
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.
Read more
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.
Teams are building their own utilities to carry context from one model to another when a limit hits—a sign the switching problem is real enough to engineer around.
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.
Read more
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.
The full production chain runs from idea through image-to-video, voiceover and editing. Most outputs are drafts, making cost per usable video the number to watch.
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?
Read more
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.
Five human-tested UGC angles, most of which fail after paid creators, reshoots and revisions, compared with one product turned into dozens of cheap AI variants that mostly fail at a fraction of the cost.
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.
Read more
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.
Several days spent visiting two logistics partners of several years' standing, and a client visit to yarn factories in Uzbekistan. Both posts focus on service scope and relationship length, not pricing or volumes.
These logistics posts show the day job behind the cost-discipline framing in the later AI posts. The freight analogy for AI budgets comes from years of working with line-item costs and long-term carriers.
This is the clearest answer to what Bosun Liu thinks about, written for operators rather than AI specialists.
Read more
Start here to see how the freight-cost discipline of their day job applies to AI spending, and whether that framing is useful to you. It is short, concrete and ends with actionable questions rather than a thesis.
Bosun Liu sets two kinds of spending side by side.
A shipment moves, and every dollar has a name — ocean freight, terminal handling, customs brokerage, last-mile — with procurement signing off line by line.
Ask the same company what it spends on AI each month, and the answer is usually a shrug: a few chat seats, a design subscription on someone's personal card, an API bill that doubled last quarter without anyone noticing.
The post ends with three questions teams can ask this week: how many AI subscriptions are being paid for, including those expensed on personal cards; which have an owner who can explain what they are for; and whether anyone would get an alert if usage doubled next month.
The recommendation is not an AI strategy but a bill of materials.
Finance teams already know how to build one; they just have not been invited into the conversation.
For brands planning a campaign
For a consultancy selling AI cost management, this post is a natural anchor: shadow AI spend, unclear ownership and missing alerts are audit-shaped problems. The freight analogy offers a way to pitch an AI cost review to finance and procurement audiences, not just marketing teams.
Opens with a concrete, checkable observation and turns it into a claim about where AI operations are heading.
Read more
Read it to see what Bosun Liu sees as the real constraint on AI use inside a company, and whether model routing matters to your team.
Bosun Liu searched the phrase "Claude usage limit" and treated the results as a market signal, not a complaint thread.
The argument: people are not saying the model is bad; they are saying it is good enough to be part of daily work, making a usage limit a business problem.
The post lists what teams are already doing: planning the workday around reset windows, avoiding token waste, switching between Claude, ChatGPT and Gemini when one model caps out, and building tools to move context between models.
The next bottleneck, the post argues, is not adoption but access, cost and workflow continuity.
Winning teams will combine models efficiently rather than pay for the most expensive one for everything.
For brands planning a campaign
Supports a thought-leadership angle on model routing and workflow continuity: choosing the right model for each task, avoiding token waste and keeping work moving across providers. The content is aimed at agencies, e-commerce teams and AI builders who already pay for several tools.
Explains the production reality behind AI video, not the marketing version.
Read more
Read it for the concrete list of steps and costs between an idea and a finished ad, and why Bosun Liu sees model choice as a production-economics decision.
Bosun Liu walks through AI video ad production: idea, script, image generation, style testing, image-to-video, multiple shots, voiceover, music, editing, regeneration and more variations.
Most output is draft, not finished work.
The post argues that the key business metric is not how good the model is, but how much it costs to get one usable video.
Expensive tests slow creativity; expensive variations stop teams experimenting; and a model that cannot handle every style forces manual jumps between tools.
The proposed answer is a pipeline that combines different models for concept images, product shots, image-to-video, voice, music, captions and editing support, selected for speed, flexibility and cost at real production volume.
For brands planning a campaign
Useful for a content angle on pipeline design and cost per usable asset: the cost of producing one finished video, where variation drains budgets, and how to structure a multi-model workflow for e-commerce and short-form teams.
Makes the demand-side case for creative volume in one page, with the cost structure spelled out.
Read more
Read it if you care about testing throughput rather than model capability, or want to see how Bosun Liu frames AI UGC without claiming it replaces creators.
The post replaces the question "can AI replace a real creator?" with another: how many bad creatives can you afford to test before finding a winner?
Bosun Liu describes the human UGC workflow as expensive before validation — paying creators, waiting for footage, revising scripts, asking for reshoots, testing five angles and watching most of them fail.
That failure rate is normal but costly.
AI UGC changes the arithmetic: one product can become fifty variations across hooks, faces, voices, locations, scripts and emotional angles.
Most will still fail; the point is that failure gets cheaper.
AI UGC is less about replacing creators than making creative failure affordable.
Winning teams test more ideas and kill losers faster, before the budget runs out.
For brands planning a campaign
For DTC and e-commerce briefs, this supports content on creative testing economics — variation count, cost per test, and when AI-generated variants are worth producing at volume versus when a human creator still earns the spend.
Establishes the day-to-day business behind the AI commentary.
Read more
Read it to see what Bosun Liu's logistics work looks like in practice and how they describe long-running partner relationships — the clearest statement of their operating base in these posts.
A short operational post about international logistics between Uzbekistan and China.
Bosun Liu writes that finding a reliable partner matters, that the partner handles everything door to door, and that they spent several days visiting two long-standing partners, describing the team as powerful and efficient.
For a campaign, ask for client, pricing and route details beyond the two countries and door-to-door service model.
For brands planning a campaign
Limited as a campaign hook on its own, but a credible foundation for logistics-audience content: cross-border partner selection, door-to-door service coverage and the value of multi-year carrier relationships.
Shows the physical geography of the work and how it is communicated bilingually.
Read more
Read it to understand where the business operates and who it speaks to; treat it as context, not analysis.
A travel-and-work note written in both English and Chinese.
Bosun Liu describes the scenery on the way back from Xinjiang and the purpose of the stop: meeting a client who owns yarn factories in Uzbekistan, for whom they handle cross-border logistics.
The two versions carry the same content, suggesting a bilingual audience across the trade corridor.
Ask for client, volume and arrangement details if those affect the brief.
For brands planning a campaign
Little direct campaign application. Its value is as an audience signal — a bilingual readership across China and Central Asia — relevant to briefs targeting cross-border trade or manufacturing clients in the region.
Cross-border logistics between China and Central Asia
Two posts document hands-on work moving goods across borders between China and Uzbekistan and the wider region.
Read more
One describes travelling back from Xinjiang to meet a client who owns yarn factories in Uzbekistan and helping with cross-border logistics.
The other explains why finding a reliable partner for international logistics from Uzbekistan to China matters, noting door-to-door services and visits to two long-standing partners.
The details—factory ownership, door-to-door handling and multi-year relationships—show operational experience rather than outside commentary on the trade sector.
Reports from client and partner visits, naming the service performed—cross-border logistics or door-to-door—and the relationship behind it.
For brands planning a campaign
A credible practitioner voice for trade, freight, and supply-chain audiences, particularly on China–Central Asia corridors and what makes a logistics partner trustworthy. Build campaigns around supplier relationships and route-level detail, not general market commentary.
Bosun Liu argues that AIGC is as much a cost problem as a creativity problem, especially for AI video.
Read more
A finished video ad, in this account, rarely comes from one prompt: it passes through idea, script, image generation, style testing, image-to-video, multiple shots, voiceover, music, editing and regeneration, with most outputs remaining drafts.
The key metric is the cost of producing one usable video.
The proposed response is a pipeline that assigns different tasks to different models—concept images, product shots, image-to-video, voice, music, captions and editing support.
Walks through a production process stage by stage, then shifts the evaluation from model quality to cost per finished asset.
For brands planning a campaign
Strong material for thought leadership aimed at agencies, e-commerce teams and short-form video producers focused on production economics. The argument centres on building workflows, not reviewing individual models, making it a fit for consultancies selling process design rather than tool selection.
This post rejects the question "can AI replace a real creator?" and asks instead, "how many bad creatives can you afford to test before finding one winner?"
Read more
Human UGC is described as expensive before an idea is validated: creators are paid, footage is awaited, scripts are revised, reshoots are requested, and most tested angles fail.
AI UGC is presented as changing the math by turning one product into dozens of variants across hooks, faces, voices, locations, scripts and emotional angles.
Most will still fail, but at lower cost.
The advantage is making creative failure affordable, not eliminating human creators.
Reframes the question, then compares two workflows using the same variable: cost per failed test.
For brands planning a campaign
Useful for DTC brands and startups balancing creative-testing volume against budget. The angle is testing capacity and kill-rate discipline; pair it with a brief on how many variants a team can produce and evaluate per cycle.
Model access limits and continuity across processes
Bosun Liu observes that complaints about Claude usage limits reflect not dissatisfaction with the tool but its status as daily infrastructure.
Read more
Teams are said to plan their workday around reset windows, avoid token waste, switch between Claude, ChatGPT, Gemini and other models when one hits a limit, and in some cases build tools to move context between models.
The conclusion: the next AI bottleneck is not adoption but access, cost and continuity across processes.
Teams need to route tasks to the right model, preserve context across switches and control costs once AI is part of daily operations.
Starts with a search result about usage limits and generalises from patterns in user complaints, framing the issue as a shift in the bottleneck rather than making a product recommendation.
For brands planning a campaign
A natural angle for multi-model orchestration and continuity: choosing models for each task, avoiding expensive token waste and keeping work moving through limits. It suits operations and platform audiences more than buyers looking for a single-tool endorsement.
Comparing AI budgeting to freight, Bosun Liu notes that every dollar of a shipment has a name and is signed off line by line, while the same company often cannot say what it spends on AI each month—a few chatbot seats, a subscription on someone's personal card, an API bill that doubled unnoticed.
Read more
The post offers three diagnostic questions: how many AI subscriptions are being paid for, including expensed personal ones; which have an owner who can explain their purpose; and whether anyone would be alerted if usage doubled next month. The suggested first step is a bill of materials, with finance brought into the conversation.
Draws an analogy with trade cost discipline, then turns it into a short list of questions a team can ask in a week.
The governance argument draws on the author's trade and logistics background rather than AI-specific research, and the examples are observed patterns, not audited figures. Ask for audited figures before using them to support a quantitative claim.
For brands planning a campaign
Well suited to finance, operations and procurement audiences at trading and e-commerce companies, especially in Southeast Asia. A thought-leadership angle could pair spend visibility with the cost-per-output arguments in the production-economics posts.
Bosun Liu writes about helping a client in Uzbekistan that owns yarn factories with cross-border logistics, but the client is unnamed and their connection to the logistics service is not stated beyond “we.”
A second post refers to long-standing logistics partners on the Uzbekistan–China route, also unnamed.
The AI posts name tools such as Claude, ChatGPT, Gemini and Midjourney as examples in commentary about model limits, costs and creative production—not as sponsorships, clients or partnerships.
One post notes years spent in trade without naming an employer.
Ask Liu to confirm the client, logistics partners and any commercial relationships before booking.
Bosun Liu writes short, operational posts about costs and how work gets done.
Read more
Their posts cover two threads: moving goods across borders between China and Central Asia, and getting usable output from AI tools without losing control of spend.
The two logistics posts provide working context for the AI commentary: a client with yarn factories in Uzbekistan and door-to-door freight partners of several years' standing.
The four AI posts form a connected argument: AI video is a cost-per-finished-asset problem rather than a model-quality problem; model limits are a process-continuity problem rather than an adoption problem; and unsupervised AI subscriptions are a procurement problem.
The most practical entry point is the post that asks three diagnostic questions a team can answer in a week.
For adjacent angles, see the comparison accounts on ad-creative diagnosis, AI UGC production guides and creative-testing cadence.
Soriano treats ad creative as a numbers problem rather than an aesthetic one.
Read more
Both see a weak ad as diagnosable: Soriano reads hook rate, hold rate, click-through and conversion to locate where viewers leave, and argues that most ads should be re-cut rather than killed.
Bosun Liu frames creative volume as a cost question: how many variants a team can afford to test before one works.
Both focus on the opening seconds and on reusing footage already paid for.
Soriano focuses narrowly on diagnosing and re-editing existing ad footage, with a checklist linking each metric to a fix.
Bosun Liu takes a wider view, examining the cost of the production pipeline and whether teams track AI spend.
For advice on which part of a specific ad to change, Soriano is the more direct resource.
Grace Eva covers the same shift toward AI-generated UGC as Bosun Liu.
Read more
Both present production capacity — more variants, faster testing, no reshoots — as the technology's payoff. They describe AI avatars, voiceovers and scripted variants as ways to remove the filming bottleneck rather than replace creators outright.
Grace Eva's posts are how-to and tool-selection guides, mapping a process from idea through script, video, testing and optimisation, and rounding up specific AI tools, including a named platform.
Bosun Liu writes about the economics of that process without recommending products, focusing on what one usable video costs and how spend should be governed.
Choose Grace Eva for a starting process and Bosun Liu for the budget and pipeline argument around it.
Smith treats creative throughput as a supply-chain problem, a move Bosun Liu makes in reverse.
Read more
Both argue that teams win by producing and testing more concepts rather than chasing a single winning asset, and accept a low hit rate as normal: Smith says eight failures out of ten is the system working, while Bosun Liu says AI UGC makes failure cheaper so teams can kill losers faster. Both borrow logistics language to describe advertising.
Smith writes as an operator of AI and creative-testing products and an agency, drawing on accounts running paid social.
Much of his advice concerns cadence, briefing and tracking concepts entering testing.
Bosun Liu draws on international trade, applying that cost discipline to AI tooling and production pipelines without a product pitch.
Smith is the closer match for campaign-execution advice; Bosun Liu for the cost and procurement argument behind it.
The only measured audience figure stated for Bosun Liu is a LinkedIn following of 2,384 people, observed October 9, 2026.
Read more
Their posts do not break that number down by country, age, gender, job function, employer or seniority.
The content points to an intended audience.
Posts from May 2021 address cross-border logistics between Uzbekistan and China, including door-to-door services and yarn factory clients.
Posts from June and July 2026 address e-commerce teams, agencies, DTC brands, startups and AI builders on model usage costs, AI UGC testing volume, AI video production economics and AI tool budgets.
That subject matter does not establish who is in the following, so treat audience relevance as unverified.
For a placement decision, ask for recent post reach and engagement or a LinkedIn analytics view of who engaged with the last few posts.
If job function or industry matters, ask whether they have audience survey results or another documented source covering those attributes.
A following of roughly 2,400 also caps per-post reach unless content travels beyond the following, so request the reach of two or three recent posts rather than relying on the follower total alone.
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.
Read more
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.
The subject-matter fit is strong; distribution is unproven. Commission the series only if the consultancy is buying subject-matter credibility rather than reach, and settle the format with Bosun Liu before committing a budget.
Read more
Brief
A B2B AI consultancy wants a named practitioner voice for thought leadership on AI cost management and creative production economics, aimed at e-commerce and logistics teams. It also needs a content angle on model cost discipline.
Reasons
Four of the six posts directly address the consultancy's need for thought leadership on AI cost management and creative production economics. They cover the cost of producing a usable AI video, how creative volume changes testing economics, how model usage limits become an operations problem, and why companies cannot say what they spend on AI each month. The subject is the consultancy's own, grounded in day-to-day work.
The posts offer usable structures, not just opinions: a step-by-step production pipeline, a testing-volume argument about cost per failed creative, three diagnostic spend questions a team can answer this week, and an argument for matching models to tasks. Each could become a series instalment with a worksheet, without the consultancy having to supply the thinking.
The author's day job is cross-border logistics between China and Central Asia; two posts show client and partner visits and door-to-door service relationships. That work connects the AI cost arguments to the logistics half of the brief. The freight-versus-AI budgeting post comes from someone who has signed off freight line items—the audience the consultancy wants to reach.
The tone suits B2B thought leadership. The posts explain a mechanism, name the relevant metric and end with questions or a next step rather than a product recommendation. They do not push a tool, leaving room for the consultancy to add its own methodology without contradicting the author's position.
Caveats
Bosun Liu's posts do not say whether they take on sponsored or co-authored work, or specify an offer, rate, collaboration format, disclosure or review process. Confirm directly whether they take on this work, and agree on disclosure and review before proceeding.
The only measured audience figure is a LinkedIn following of 2,384, observed October 9, 2026. The posts provide no reach, engagement, impression or demographic data. Ask for reach figures for two or three recent posts before promising the series an audience.
The available material is written commentary; no video, audio or transcript material was available. On-camera or recorded-interview work is therefore untested.
Post dates are missing from the text here. Two logistics posts date to May 2021 and the four AI posts to mid-2026, suggesting a recent shift toward AI economics. Ask for a full content inventory to assess how consistent that focus is.
When to choose a specialist
Poor fit based on the available posts. Decline or park the campaign until the research below is completed and verified; the mismatch is structural, not a matter of taste.
Read more
Brief
A consumer skincare brand launching direct-to-consumer in North America wants short-form video demonstrations, broad consumer reach through paid amplification and tracked conversion against spend.
Reasons
The reviewed work is written B2B commentary for operators, not consumer product demonstration. None of the six posts addresses consumer audiences, beauty, retail or a physical product trial. The closest material covers how teams budget for and produce AI-generated ads: advice for buyers, not a demonstration for shoppers.
The campaign needs measured reach and conversion, but the posts contain no engagement, reach, impression, click or conversion figures, and no audience breakdown by country or interest. A following of 2,384 sets a low ceiling on organic reach, so paid amplification would carry almost all the load. Ask for audience and performance data before committing spend.
The campaign calls for short-form video, but the available posts are text-only. There is no basis for judging on-camera performance or production quality; ask for a video sample.
The campaign assumes a commercial format that can be bought, but the posts include no rate card, media kit, sponsored-post example or inquiry route. Cost, rights, exclusivity and turnaround remain unknown; ask for these terms before negotiating.
Caveats
The available posts do not support committing consumer launch budget. Confirm directly whether Bosun Liu takes on consumer or sponsored work before proceeding.
A documented sponsored post, rate card or video sample with performance data could change the assessment for a consumer brief. Revisit the fit if any of these become available.
A worked campaign example
Angle
Build the series around one argument: once AI becomes part of daily production, the key metric is cost per usable output—and most teams have never priced one. Each instalment turns a structure already visible in the work into an exercise readers can complete.
Brief
Develop the B2B AI consultancy's need described above into a three-part thought-leadership series on model cost discipline for e-commerce and logistics teams.
Deliverables
Part one: the AI bill of materials. Start with the contrast between a shipment, where every dollar is named and signed off, and AI spend, scattered across seats, personal-card subscriptions and an API bill nobody watches. Deliver a one-page template listing subscriptions, owner, purpose and monthly cost, plus three questions worth asking this week. Close by noting that finance teams already know how to build this and have not been invited in.
Part two: cost per usable video. Trace the pipeline from idea through image generation, style testing, image-to-video, multiple shots, voiceover, music, editing and regeneration, marking where drafts become finished work. Give readers a simple calculation: total production spend divided by finished assets, with a worked example at two plausible volume levels. The argument: model choice is a production-economics decision, not a taste decision.
Part three: the testing math. Replace the question of whether AI replaces creators with a more useful one: how many failed creatives can a team afford before finding a winner? Show what the human version costs before validation, then what changes when one product becomes dozens of variants across hooks, faces, voices and scripts. End with a kill-rate discipline: how many variants to produce per cycle and what evidence retires one. The response ends with a short brief template and a route to a cost review, presented as the consultancy's methodology built on the author's framing.
Measurement
Track worksheet downloads or completions and brief-template submissions as the primary signal, not follower growth. Before publishing, ask for the reach and engagement of two or three recent posts on the author's account to establish a baseline for comparison.
Next step
Confirm with Bosun Liu whether sponsored or co-authored content is available, how authorship and disclosure would work, and what review the consultancy gets before publication. If the answer is yes, start with part one: the bill-of-materials template can draw on material already published. If no collaboration format is available, cite the public posts as source material for the consultancy's own bylined series instead.
Why it works
The posts already make the argument, so the series needs editing and packaging, not invention. It gives the consultancy three concrete assets to hand over instead of three essays, speaks to the logistics half of the audience through the freight comparison, and leads to a service the consultancy sells.
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.
Read more
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.
The argument is narrower and more operational.
Soriano writes about ad creative on Meta and TikTok for DTC apparel and consumer brands, not AI model costs, subscription visibility or freight-style budgeting.
The logistics half of the consultancy's audience gets little from that focus.
His posts also close with a direct service handoff: an offer to rewrite the hook on one live ad.
That makes the content a lead-in to his own offer, rather than neutral framing for the consultancy's methodology.
A commissioned series with him would likely compete with, rather than extend, the consultancy's service line.
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.
Read more
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.
The posts are promotional summaries, not worked arguments.
Each describes what a full guide covers and routes readers to it; one names a specific tool in a commercial tie, leaving the substantive detail outside the posts.
There's no costing, worked example of output price or mechanism-level explanation of why one configuration beats another—the layer the consultancy needs for a series on AI production costs.
The heavier hashtag framing and checkmark lists also read as vendor marketing, so pairing Grace Eva with a neutral methodology would take more editorial work than pairing Liu's posts, which reach a mechanism and stop.
Choose Nathan Smith when the series wants an operator-founder voice and a supply-chain frame rather than cost accounting.
Read more
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.
Much of the argument is borrowed or self-promotional.
The defensibility post attributes its core framework to a named third party and points readers to that framework's site; the scaling post ends with an explicit pitch for agency help.
The consultancy would be buying a founder narrative plus a service ad, rather than an independent cost method.
The creative-as-inventory post also doesn't price anything: it argues for testing cadence and hit rates, not what a finished asset or subscription actually costs.
It can complement, but not replace, the bill-of-materials and cost-per-usable-video analysis the consultancy needs.
This page covers six LinkedIn posts by Bosun Liu: four on AI costs, model usage limits and creative production economics, and two on cross-border logistics between Uzbekistan and China, one of them written in both English and Chinese.
The posts here include no video, audio, transcript, comment or engagement data, and do not state reach or audience breakdown; the only audience figure is a LinkedIn following of 2,384 observed on 9 October 2026.
The post text carries no publication dates, so nothing here is ordered chronologically.
Three adjacent creator accounts are included for comparison.
The posts do not mention signature work such as a book, show or business, and do not describe sponsorship, rate or collaboration format, so commercial terms should be confirmed directly.
