XGIMI · ERP + AI PIPELINE · 2024–PRESENT

Re-architecting enterprise operations with agentic AI.

Joined as XGIMI’s 7th US employee to rebuild the Americas ERP system and introduce AI-powered automation. Deployed a multi-agent system on OpenClaw — with Claude as the LLM backend and n8n as the workflow layer — that automated financial reconciliation, demand forecasting, and inventory optimization.

$15M
REVENUE
18%
ABOVE
FORECAST
85%
RECONCILIATION
AUTOMATED
4
AI AGENTS
IN PRODUCTION
OpenClaw + Claude + n8n
AGENT
STACK
XGIMI ERP + AI Pipeline hero image

XGIMI makes premium home projectors — $800–$2,500 devices sold through Amazon, Best Buy, and direct-to-consumer. The US operation runs lean: seven people managing a $15M+ revenue stream across sales, operations, and finance.

When I joined in July 2024, the ERP system was a patchwork of SAP modules, manual spreadsheets, and tribal knowledge. Financial reconciliation took 3 days per month. Demand forecasting was gut feel. I was brought in to fix the foundation and build the AI layer on top.

PM + builder — half strategy, half code
What I owned

I’m the sole PM for both the ERP re-architecture and the AI pipeline. I report to the VP of Operations and work directly with our SAP consultant, a contract data engineer, and the finance team. I also write production code — I built and deployed the agent pipelines myself — writing production code with Claude Code, deployed on OpenClaw.

How I worked

Mornings are ERP — SAP configuration, data migration, process mapping with Finance and Ops. Afternoons are AI — designing agent architectures, writing prompts, testing reconciliation outputs, and building the n8n automation layer. I ship code alongside specs.

Four agents on OpenClaw, one automation platform.

4 Agents

Reconciliation Agent

OpenClaw + Claude + SAP BAPI
Matches invoices, POs, and bank statements across three systems. Reduced 3-day manual process to 4 hours with 99.2% accuracy.

Forecasting Agent

OpenClaw + Claude + historical data
Analyzes 18 months of sell-through data, seasonality, and promotional calendars to generate weekly demand forecasts by SKU.

Inventory Agent

n8n + warehouse API
Monitors real-time stock levels across 3 warehouses and 2 FBA locations. Triggers reorder alerts and generates transfer orders.

Reporting Agent

OpenClaw + Claude + Google Sheets API
Generates weekly executive reports — P&L summaries, channel mix, margin analysis — and distributes to leadership automatically.
SYSTEM ARCHITECTURE
How the agent system works under the hood
ORCHESTRATION
OpenClaw as the agent platform, Claude as the LLM backend. Each agent receives structured context windows with task-specific prompts, data schemas, and confidence thresholds. n8n handles scheduling and inter-agent handoffs.
DATA PIPELINE
Amazon SP-API, Best Buy EDI, Shopify webhooks, and SAP BAPIs feed into a unified data layer. Agents consume normalized inputs and write structured outputs back to SAP and Google Sheets.
GUARDRAILS
Confidence scoring on every output. Actions above 95% auto-execute. Below that, agents surface exceptions with dollar-impact ranking and one-click resolution paths for human review.
Data Sources
SAP BAPIs
Warehouse
Google Sheets
Historical Data
OpenClaw · Claude LLM
Reconciliation Agent
Forecasting Agent
Inventory Agent
Exception Agent
Outputs
Reports
Alerts
Forecasts

The hard part of AI in enterprise isn’t the model — it’s deciding what the model should never decide alone.

3
DAYS TO RECONCILE

The Problem

XGIMI’s Americas operation was growing faster than its systems could handle. Financial reconciliation — matching Amazon payouts, Best Buy remittances, DTC orders, and SAP records — took the finance team 3 full days every month. Demand forecasting was a monthly meeting where people guessed. Inventory allocation across 5 locations was managed in a shared Google Sheet.

3-day monthly reconciliation

No automated demand signals

5-location inventory blind spots

From patchwork to platform.

I mapped every data flow, identified every manual handoff, and rebuilt the system in layers — ERP foundation first, then AI automation on top.

DATA FOUNDATION
01 SAP master data cleanup — Standardized 2,000+ SKU records, fixed pricing hierarchies, reconciled vendor codes
02 Channel integration — Built API connectors for Amazon SP-API, Best Buy EDI, and Shopify webhooks
03 Financial mapping — Created unified chart of accounts across all sales channels and cost centers
AUTOMATION LAYER
04 Reconciliation pipeline — Claude-powered agents on OpenClaw match transactions across bank, marketplace, and ERP systems
05 Demand forecasting — Historical analysis + promotional calendar + seasonal patterns → weekly SKU-level forecasts
06 Inventory optimization — Real-time stock monitoring with automated reorder points and transfer suggestions
INTELLIGENCE
07 Executive reporting — Automated weekly P&L, channel mix, and margin analysis delivered to leadership
08 Anomaly detection — Agents flag pricing discrepancies, unusual returns, and margin compression in real-time
Full pipeline operational — 85% of reconciliation now automated

An AI-powered operating system that runs itself.

The system I built on OpenClaw isn’t a dashboard — it’s an autonomous multi-agent pipeline. Four AI agents work in coordination: the reconciliation agent processes financial data nightly, the forecasting agent updates demand signals weekly, the inventory agent monitors stock continuously, and the reporting agent synthesizes everything into executive-ready outputs. Human review is built in at every critical decision point.

Claude over fine-tuned models.

We evaluated building custom ML models for reconciliation. Claude won — it handled the messy, unstructured matching (partial payments, split shipments, promotional adjustments) that rule-based systems and fine-tuned models couldn’t parse. And I could iterate on prompts in hours, not retrain models over weeks.

Agents over monolith.

Instead of one large automation, I built four specialized agents that communicate through structured handoffs. Each agent has a clear scope, its own error handling, and can be updated independently. When the reconciliation logic needed to change for a new marketplace, I updated one agent without touching the others.

Human-in-the-loop by design.

Every agent output goes through a confidence scoring system. High-confidence actions (>95%) execute automatically. Everything else gets flagged for human review. This isn’t a limitation — it’s the architecture. Finance trusts the system because they know it won’t make decisions it isn’t sure about.

ARCHITECTURE TRADE-OFF
I chose four narrow agents over one general-purpose model because enterprise finance has zero tolerance for hallucination. A reconciliation agent that’s 99% right is worse than a spreadsheet — that 1% erodes all trust. Specialization let me tune each agent’s confidence threshold independently.

I built too much before validating with Finance.

The first version of the reconciliation agent was technically impressive but practically useless. I had optimized for matching accuracy without understanding how the finance team actually worked. They didn’t need perfect matches — they needed the agent to surface the 5% of transactions that were genuinely ambiguous, with enough context to resolve them in under a minute. I rebuilt the output layer to match their workflow: a daily exception report ranked by dollar impact, with one-click resolution paths. Adoption went from skeptical to dependent in two weeks.

Three horizons. Compounding returns.

From automation to intelligence.
01
Foundation
H2 2024
  • SAP re-architecture
  • Master data cleanup
  • Channel integrations
  • Basic reporting
02
Automation
H1 2025
  • Multi-agent system on OpenClaw
  • Reconciliation pipeline
  • Demand forecasting
  • Inventory optimization
03
Intelligence
H2 2025
  • Predictive margin optimization
  • Automated pricing recommendations
  • Cross-channel inventory balancing

Selling AI to a team that didn’t ask for it.

Nobody at XGIMI requested an AI system. I had to prove value before I could propose it.

01

Start with the pain

Shadowed Finance for a week. Documented every manual step in reconciliation. Made the cost of the status quo undeniable

02

Prototype in production

Built the first reconciliation agent in 3 days — Claude Code as the dev tool, OpenClaw as the runtime. Ran it against real data and showed Finance the results side-by-side with their manual output

03

Expand through trust

Once reconciliation worked, the forecasting and inventory teams asked to be next. Pull, not push

04

Document everything

Built an internal wiki documenting every agent’s logic, inputs, outputs, and failure modes. Made the system transferable, not dependent on me

What the AI system delivered.

$15M
REVENUE (18% ABOVE FORECAST)
◆ FORECASTING AGENT
85%
RECONCILIATION AUTOMATED
◆ RECONCILIATION AGENT
3d→4h
MONTHLY CLOSE TIME
◆ OPENCLAW PIPELINE
99.2%
MATCHING ACCURACY
◆ CLAUDE + OPENCLAW
4
AI AGENTS IN PRODUCTION
◆ MULTI-AGENT SYSTEM
KEY TAKEAWAY

The goal was never “build an AI system” — it was fix an ERP that couldn’t scale. Once the work broke down into pattern matching and data transformation, agents were the obvious architecture. Shipping it solo with Claude Code compressed months of data engineering into weeks.

What I’d do differently.

I would have involved Finance from day one of the agent build, not after the first version was done. The technical architecture was right, but the output format was wrong because I hadn’t watched them work closely enough. I’d also invest earlier in monitoring and alerting — we had a silent failure on the inventory agent for 48 hours before anyone noticed. The system needs to tell you when it’s broken, not wait for you to check.

Want to discuss this project?

Get in Touch →
← Back to all projects
Next Project
Dealer Portal
0-to-1 B2B commerce platform for Segway's SMB dealer network