Triody

Cross-Border ｜ Agent Team for E-Commerce

CBEC

Problem & Demand

Most cross-border e-commerce sellers don't fail because of bad execution. They fail because they picked the wrong product to begin with. The real bottleneck isn't operations — it's the research that happens before a single dollar is spent on inventory.

Traditional market research takes 2–3 weeks: manually reading hundreds of competitor reviews, scrolling social media for trend signals, comparing prices one ASIN at a time, making Excel charts for the boss. It's slow, it's expensive, and it still ends with a gut-feel decision. One bad pick costs ¥70K–350K in dead stock, wasted ads, and lost time.

Product sourcing is no better. Hours of scrolling 1688 and Alibaba, manually comparing MOQ, price, and supplier reliability — hoping you don't get scammed. The entire pre-launch workflow was built on manual labor and intuition. It didn't scale, and it didn't need to be this way.

Solution

We designed and deployed an agent team on sloke.ai — two specialized agents that work in concert, each owning a distinct part of the cross-border e-commerce workflow. The architecture draws from Alibaba's Accio Work skill-collaboration patterns, adapted for the realities of cross-border selling.

Here's what this looks like in practice. A real scenario:

The boss asks: "Can we still make money on bluetooth earphones?"

I (Market Research) pull Jungle Scout weekly search volume — ~250K, stable. Exa reports CAGR at 5.8%. Competitive landscape: Apple, ZMI, Samsung dominate the top end. But the mid-to-low tier has gaps. 30 minutes in, the boss has data he can act on.

Argos (Product Sourcing) is already running in parallel on 1688: three suppliers found with the same mold, ¥28–45 factory price, Alibaba at ¥18–25 with MOQ 500. Landed cost: $6.80. Recommended retail: $24.99 — positioned at 60% of Anker's price point with clear differentiation.

30 minutes from question to decision-ready data. That's the system working.

Market Research Agent dashboard — bluetooth earphone selection scenario: search volume + trends + competitors + supply chain, decision-ready in 30 min
Market Research Agent dashboard — bluetooth earphone selection scenario: search volume + trends + competitors + supply chain, decision-ready in 30 min

What each agent does, specifically:

Argos — Product Sourcing:

1688 image search: one screenshot of an Amazon bestseller → 19 fields auto-extracted (price, sales volume, repurchase rate, shipping origin). What took 2–3 hours of manual scrolling now takes 5 minutes.

Jungle Scout deep category analysis: 8-dimension report with 3-tier product recommendations. What took 2–3 days of pulling data and building spreadsheets now runs in a single pass. At ¥2,000/day, that's ¥6,000–10,000 saved per analysis.

Alibaba supplier matching + screening: one-click search → Markdown table with thumbnails, MOQ, pricing, lead time. Supplier shortlisting from 1 day → 1 minute.

Scientific product selection: multi-source cross-validation (JS search volume + Exa trends + competitive analysis) replacing 1–2 weeks of manual trial. One wrong pick costs ¥70K–350K. This is how you avoid it.

AliExpress seller evaluation: 5-indicator scorecard (store age, followers, response rate, positive review rate, review quality) + anti-fraud checklist + landed cost calculation. 1 hour per supplier → minutes. This isn't efficiency — it's survival.

Agent capability overview — 7 core functions, from product selection to supplier evaluation, each with quantified savings
Agent capability overview — 7 core functions, from product selection to supplier evaluation, each with quantified savings

Argos-Marketing — Market Research:

Competitor review mining: 4 agents run in parallel, analyzing hundreds of reviews to find market gaps — unmet needs, recurring complaints, feature requests. What used to take 1–2 days and cost $3K–5K/month in outsourced research now runs in a single pass.

Category entry feasibility: Jungle Scout search volume trends + Exa cross-validation. Days of social media scrolling and gut-feeling → 10 minutes of data-backed analysis.

Best Seller deconstruction: pricing sweet spot, visual strategy, and copywriting logic extracted in one pass. Half a day of manual comparison → minutes.

Customer review hook extraction: 6-dimension analysis surfacing the exact language and pain points from real customer reviews. Hours of manual sorting → minutes.

Keyword trend tracking: Selenium auto-scrapes AMZ123 and JS search data → CSV. 30 minutes of daily manual work → zero.

Competitor background search: Exa pulls company and founder profiles in seconds. 1 hour per competitor → seconds.

Market data visualization: Python auto-generates charts and trend lines. 1–2 hours of Excel → automated.

@Argos + @Argos-Marketing overview — product sourcing saves 3–4 h/day, market research saves 2–3 h/day
@Argos + @Argos-Marketing overview — product sourcing saves 3–4 h/day, market research saves 2–3 h/day

Beyond the screenshots: we deployed the agent team on-site with the client's actual workflow, their actual data sources, their actual decision-making rhythm. The system isn't a demo — it's running in production, connected to their JS account, their 1688 supplier relationships, their daily operations.

On-site FDE with the cross-border team (shared with employer consent)
On-site FDE with the cross-border team (shared with employer consent)

The bottom line, in numbers:

Product sourcing: 3–4 hours saved per day. $2K–5K saved per month. $10K–50K in trial-and-error costs avoided per product decision.

Market research: 2–3 hours saved per day. $3K–5K saved per month.

Combined: 5–7 hours saved every day. $5K–10K saved every month. And the kind of mistake that kills a cross-border business — picking the wrong product — is systematically prevented.