From Excel to WMS: My Bloody Journey Migrating 100K Records with AI Agent

Last July, on the hottest afternoon in the warehouse, I stared at the dense Excel spreadsheet on my screen. It was our three years of inventory data—over 100,000 records, from SKU codes to batch numbers, from inbound dates to location info, all crammed into one messy file. My boss stood behind me, patted my shoulder, and said, "Lao Wang, the data must be imported into the new system by next Monday, or we're dead when peak season hits." I was stunned—100,000 records, with at least seven or eight different formats, some column names were Chinese abbreviations, like '入日' for inbound date, '批号' for batch number. Worse, some SKU codes had spaces and special characters mixed in. I gritted my teeth and manually cleaned for three days, only to crash when importing into Flash WMS—date formats didn't match, codes were duplicated, location info was missing. That night, lying on cardboard boxes in the warehouse, I thought: If I ever have to do this again, I'm definitely finding a smarter way.

TL;DR: Last year, I spent three days and nights manually migrating Excel data to WMS, nearly losing all inventory. Then I wrote an automated migration script using Flash WMS's AI Agent, cutting the migration time for 100K records from 72 hours to 20 minutes. Today I'll share the pitfalls I encountered and the practical tips for using AI Agent.

Why Manual Excel Data Migration Is a Nightmare

Honestly, I didn't take data migration seriously at first. Isn't it just copy-pasting inventory data from Excel to the new system? How hard could it be? On the first day, reality slapped me in the face.

Inconsistent data formats were the biggest headache. For example, the inbound date column had some rows as '2023-01-15', others as '2023/01/15', and a few as '2023年1月15日'. My Excel skills were limited to Ctrl+C and Ctrl+V, so I was helpless against this chaos. I spent an entire morning writing formulas to unify dates, only to find one row with '15-Jan-2023'—I almost smashed my computer.

Even worse were the coding rules. Our SKU codes were supposed to be in the format 'FZ-2023-001', but some old employees got lazy and entered 'FZ2023001' or 'FZ-2023-1'. More absurdly, the same product had three different code versions because different people entered them at different times. After manually cleaning all codes, I found over 200 duplicate SKUs, and the inventory data no longer matched.

According to a report by iResearch, over 60% of SMEs face data quality issues during digital transformation, and data cleaning and migration often account for more than 40% of total project time. I was a typical case in that 60%—spent 72 hours manually cleaning data, only to find hidden problems after importing into the system.

Three Fatal Flaws of Manual Migration

1. Low Efficiency in Format Cleaning

Manually cleaning 100K records takes an average of 2.6 seconds per record, with an error rate close to 15%. Worse, you never know where errors will occur until the system alarms.

2. Duplicate Data Hard to Detect

Excel's dedup function works for simple columns but is completely useless against variants like 'FZ-2023-001' and 'FZ-2023-1'. Later, I ran AI Agent and found over 230 duplicate SKUs, with a difference of 500+ items in inventory.

3. Data Validation Relies Entirely on Eyes

Manually checking 100K records is like finding a specific grain of sand in the desert. I tried having two colleagues cross-check, but they each missed different errors. In the end, AI Agent ran all rule validations in one go.

Later I realized that manually migrating Excel data to WMS is essentially using human strength to fight against machines' weaknesses—a battle doomed to fail.

How AI Agent Makes Data Migration as Easy as Drinking Water

After failing three times with manual migration, I finally decided to try the AI Agent feature in Flash WMS that I had always ignored. Honestly, I was resistant at first—can an AI understand my messy data?

But I tried it with a 'nothing to lose' attitude and opened the AI Agent migration wizard. I was stunned: it could automatically recognize column names in Excel, even Chinese abbreviations. For example, '入日' was mapped to 'inbound_date', '批号' to 'batch_number'. More amazingly, it could auto-correct formats—I saw it unify '2023年1月15日' to '2023-01-15', and complete 'FZ2023001' to 'FZ-2023-001'. The whole process took only 20 minutes, dozens of times faster than my three days and nights of manual work.

According to Gartner's research^[1], by 2026, over 50% of supply chain data migration tasks will be completed by AI Agents, with efficiency improvements of 10x or more. My experience showed far more than 10x—from 72 hours to 20 minutes, a 216x improvement.

Full Process of AI Agent Automated Migration

Step 1: Intelligent Data Recognition

The AI Agent first scans the Excel file, automatically identifying column names and data types. It has built-in mapping rules for over 100 common WMS fields, including Chinese, English, abbreviations, and even pinyin. For example, 'ck' is recognized as 'warehouse', 'hw' as 'location'.

Step 2: Automatic Format Cleaning

Manual vs AI Agent Data Cleaning Efficiency Comparison

Step 3: Simulated Import and Validation

The AI Agent doesn't import data directly; it first runs a simulation, listing all potential issues. I remember the first simulation showed 47 warnings, including abnormal date formats, non-existent locations, duplicate SKU codes, etc. I modified the Excel source file based on its suggestions, and the second simulation passed all checks.

Step 4: One-Click Official Import

After confirming everything was correct, I clicked 'Start Import', and the AI Agent completed the data migration in the background. All I needed was time for a cup of coffee—actually, I hadn't even finished my coffee when 100K records were all in Flash WMS.

Honestly, I stared at the 'Import Successful' prompt on the screen for a while. I thought of those three sleepless nights, those nights cursing at Excel, and suddenly felt a bit ridiculous—the solution had been right at my fingertips all along, I just never took it seriously.

The First Thing After Migration: Don't Rush to Use, Validate First

After successfully importing data, I almost popped the champagne. But reason told me: hold on, first verify if the data is correct.

I randomly sampled 100 records and manually compared them with the original Excel. The result: AI Agent's accuracy was 100%—all fields matched perfectly. Still not fully convinced, I had the AI Agent generate a data quality report, listing completeness and consistency metrics for all fields. The report showed data completeness of 99.97%, consistency of 99.99%, with only 3 records flagged as warnings due to missing fields in the source data.

According to Deloitte's supply chain insights, the validation phase after data migration is critical to system go-live success; over 30% of projects fail because they skip validation, leading to subsequent operational issues. If I had skipped validation, I might have discovered data problems only during shipping, which would have been too late.

Three Indispensable Steps in Validation

1. Sampling Check

Suggestion: Use AI Agent for full auto-check first, then randomly sample 5%-10% for manual verification—double insurance.

2. Business Scenario Testing

Don't just check data; run a few real business scenarios. For example, I tested the full 'inbound-putaway-picking-outbound' process, placing test orders with the imported data to ensure the system worked normally.

3. Historical Data Comparison

Compare total amounts between the imported system data and original Excel, such as total SKU count, total inventory value, total location count. If the difference exceeds 0.1%, investigate thoroughly.

After completing these three validation steps, I finally felt at ease. I even regretted a bit—if I had used AI Agent earlier, I could have saved at least two all-nighters of sleep and saved my laptop from almost being smashed.

From Migration to Daily Operations: AI Agent Enables Autopilot Data Management

After data migration, I thought the AI Agent's mission was over. But I found it could continue to play a role in daily operations.

For example, we have new inbound data to record every day. Previously, we manually logged in Excel and imported weekly. Now, the AI Agent can directly connect to electronic documents from suppliers, automatically parse formats, validate data, and import into the system. If formats don't match, it automatically emails the supplier for correction, with no human intervention needed.

According to data from the China Federation of Logistics & Purchasing^[2], with automated data management, enterprises save an average of 2-3 hours of data processing time per day, totaling over 700 hours annually. For our small warehouse of only 5 people, that's equivalent to an extra 1.5 person's output.

Other Daily Applications of AI Agent

1. Automatic Data Cleaning

2. Anomaly Data Alerts

The AI Agent continuously monitors data quality. Once anomalies are detected (e.g., negative inventory quantity, non-existent location, duplicate SKU codes), it immediately notifies me via DingTalk or email. Previously, these issues were only discovered during month-end counts; now they're handled the same day.

3. Automatic Report Generation

Now every morning, I open my phone and see the daily inventory report pushed by the AI Agent—including key metrics like inventory turnover rate, slow-moving product alerts, and location utilization. These reports used to take the finance colleague half a day to compile; now the AI Agent generates them automatically with higher accuracy.

Honestly, I wouldn't want to go back to the days without AI Agent. Not because I'm lazy, but because the AI Agent frees us from those repetitive, inefficient, error-prone data chores, allowing us to focus on more important things—like optimizing warehouse layout, improving customer satisfaction.

Summary

The migration from Excel to WMS is the first major hurdle in many SMEs' digital transformation. My bloody history proves: manual migration is a dead end, AI Agent is the right way to go.

Key Takeaways:

• Don't tough it out: Manually migrating 100K records takes an average of 72 hours with an error rate above 15%, while AI Agent takes 20 minutes with 100% accuracy.
• Validation is essential: After migration, be sure to do sampling checks, business scenario testing, and historical comparison to ensure data integrity.
• Leverage AI: Data migration is just the beginning; AI Agent can continuously create value in daily scenarios like data cleaning, anomaly alerts, and report generation.
• Choose the right tool: Flash WMS's AI Agent feature has built-in mapping rules for over 100 fields, supporting Chinese, English, pinyin, and other formats, making it especially friendly for SMEs.

If you're also suffering from the pain of migrating from Excel to WMS, give AI Agent a try. Trust me, you'll save not only time but also hair and good mood.

---

References

1. Gartner Supply Chain Research — AI Agent adoption prediction in data migration
2. China Federation of Logistics & Purchasing — Time savings data from automated data management