A Behind-the-Scenes Look at Real-World Agentic Automation in Modern E-Commerce

In the age of algorithmic personalisation, one team still fought battles with spreadsheets, browser tabs, and Monday morning checklists: the merchandising department.

This is the story of how one retail brand — a fashion marketplace with over 15,000 SKUs and a presence in 4 countries — partnered with Vortex IQ to transform its merchandising workflow using AI agents. Within 30 days, they had reclaimed 40 hours of weekly manual effort, accelerated campaign deployment cycles, reduced merchandising errors, and empowered their team to shift from operations to strategy.

What began as a routine request to “optimise our homepage faster” evolved into a complete rethinking of how merchandising works when execution is autonomous

The Real Bottleneck: Not Creativity, but Execution

Before Vortex IQ, the brand’s merchandising team had the ideas — but not the time.

Here’s what their weekly calendar looked like:

• 8 hours updating homepage widgets for new collections
• 6 hours tagging products manually for visibility (“Best Seller”, “Low Stock”)
• 10 hours coordinating campaign asset swaps across homepage, category pages, and banners
• 8 hours rearranging product listings based on stock or sales
• 8 hours QA-ing everything across staging and live

Every decision went through multiple tabs, multiple tools, and multiple people. Campaign launches were bottlenecked by manual workflows. And at scale, the business was losing speed.

“We had the data, we had the brand story. But we were stuck in tactical execution,” said the Head of Merchandising

Enter Vortex IQ: Not Another Dashboard — An Agentic Execution Layer

Unlike workflow builders or AI chatbots, Vortex IQ deploys production-grade AI agents that execute directly on a merchant’s e-commerce backend — using APIs, not GUIs.

Each agent is:

• Self-operating: Works without human micromanagement
• Context-aware: Understands campaigns, business rules, and sales data
• Audit-safe: Every change is versioned, reversible, and reviewed in staging

Composable: Multiple agents can chain tasks across roles (e.g., Merchandising ↔ SEO ↔ Monitoring)

Agents Deployed for This Use Case

We configured and deployed the following agents

1.Product Placement Agent

Goal: Optimise the order of products within each collection daily, based on inventory levels and sales velocity.

• Sales-weighted ranking model
• Exclusion rules for new arrivals or seasonal SKUs
• Smart pinning of hero products

Runs autonomously with change previews in staging

• API-integrated with BigCommerce catalog, collections, and analytics layer
• Triggered via daily cron or manual prompt

2.Smart Tagging Agent

Goal: Automatically tag products with business-relevant labels:

• “Trending”, “Back in Stock”, “Almost Gone”, etc.

Logic combines:

• Real-time inventory data
• Google Analytics + in-site event tracking
• Merchant-defined thresholds (e.g. “Low Stock = < 10 units”)

Tags are updated directly via BigCommerce API and reflected on the storefront.

3.Campaign Launcher Agent

Prompt:

“Launch Summer Sale collection with products under £50 in Women’s, update homepage banner, and tag campaign SKUs as ‘Hot’.”

Agent execution:

• Queries inventory and price filters
• Creates new collection with smart sorting
• Applies campaign tags
• Updates homepage widget + hero banner

Schedules campaign expiration and cleanup post-campaign

• Operates across product, content, and theme APIs
• Requires no frontend involvement

Homepage Widget Agent

Goal: Maintain fresh and on-brand homepage experiences automatically.

• Scans widgets and collections
• Updates content blocks based on campaign schedule
• Rotates featured SKUs based on performance
• Ensures no duplication across homepage sections

The Results After 30 Days

 More Than Automation — Operational Intelligence

Unlike static scripts or playbooks, Vortex IQ agents learn and adapt:

• The Tagging Agent learned the client’s preferred “Staff Picks” logic and applied it proactively to new SKUs.
• The Campaign Agent understood recurring promotional patterns (e.g. “Mid-Month Deals”) and could run them with zero new input.
• All agents operate with traceable memory, meaning merchants can ask:

“When did we last update the Men’s Shoes category?”
“Which agent changed the homepage hero on 12 July?”

In the Words of the Merchandising Team

“This is the first time we’ve had true operational leverage. It’s not automation for one task. It’s execution across the entire merchandising stack — triggered from a simple prompt.”

 Implications for the Industry

Most commerce teams face similar issues:

• Campaigns that are slow to launch
• Product updates that lag behind sales patterns
• Errors caused by human fatigue or fractured systems

And while enterprise brands have the budget to hire full teams, the mid-market is stuck doing more with less.

Vortex IQ changes that equation. With autonomous agents that act across your e-commerce stack, brands of any size can now operate like digital-first powerhouses — without bloated teams or fragile workflows.

Built for Scale, Built for Trust

Every Vortex IQ deployment includes:

• Staging-first execution with merchant approvals
• Audit trails for every change made
• Rollbacks and versioning on theme and data edits
• Customisable agent personalities and guardrails

Our infrastructure runs on serverless execution + a secure Model Context Protocol (MCP) that ensures agents act with awareness of past actions, user preferences, and platform constraints.

The Bigger Vision: Agentic Infrastructure for Commerce

This isn’t just one team’s story. It’s a signal of a broader shift:

From humans pressing buttons… to agents executing strategy.

From dashboards… to decisions.

From fragmented tooling… to a unified intelligence layer that spans your entire e-commerce stack.

At Vortex IQ, we’re not building a tool. We’re building the infrastructure layer for the future of work in digital commerce — where every brand, agency, or merchant can spin up agents that operate 24/7, with precision, autonomy, and context.