Guide: Automating lead list cleaning with HuggingChat and a simple Python script.

Everyone's raving about AI lead enrichment like it's a silver bullet. Spoiler: it's not. HuggingChat's free API is decent for parsing messy input, but expecting it to magically fix your data is a one-way ticket to "garbage in, gospel out."

Here's a pragmatic snippet I use to scrub lead lists *before* they hit Salesforce. It's not fancy, but it catches the obvious junk. You'll need the `huggingface_hub` library. This just standardizes company names from a CSV column—because "intl ltd" and "International Ltd." shouldn't be two accounts.

```python
import pandas as pd
from huggingface_hub import InferenceClient

client = InferenceClient(token="your_token")
def clean_company(raw_name):
prompt = f"""Return only the cleaned, canonical company name. Input: {raw_name}. Rules: Remove 'Inc', 'LLC', 'Ltd'. Keep the core name. Correct obvious typos."""
response = client.text_generation(prompt, max_new_tokens=20)
return response.strip()

df = pd.read_csv('your_leads.csv')
df['cleaned_company'] = df['company'].apply(lambda x: clean_company(str(x)))
df.to_csv('cleaned_leads.csv', index=False)
```

Run it, review the output manually, *then* import. It saves me a few hours of manual merging each month. The key is treating the output as a *suggestion*, not a truth.

— Laura