Enrichment and AI columns

Enrichment columns look data up

Add the column, run it, and each row is filled from the data sources behind it. Email finding runs several providers in a waterfall rather than one, so a contact that the first source does not know can still be found by the second. Validation is a separate step that checks an address actually accepts mail.

AI columns write a field

An AI column takes your instructions and writes one value per contact. Four jobs cover almost every use:

• Classify: is this company a software vendor or a reseller? Is this person a decision maker?
• Score: how well does this contact match the ICP, from 1 to 5, with the reason in a second column.
• Clean: turn a messy job title or company name into something you can use in a greeting.
• Personalise: one opening line from the company's website, a recent post or a job ad.

How to run an AI column well

1. Write the instruction like a brief

   Say what the answer should be, what it should never be, and the exact format: a word, a number, one sentence.

   

2. Give it the right inputs

   An AI column reads the columns you reference. Personalisation needs something to be personal about: a website summary, a post, a job ad.

3. Test on ten rows

   Run the first ten rows and read every answer before you run the rest. Mistakes repeat identically across a thousand rows.

   

4. Constrain the output

   "Answer only Yes or No" is filterable. "Tell me about the company" is not.

5. Change the instruction carefully

   Saving an edited instruction clears the column and recalculates every row, billed like a fresh run. Test the new version on the first ten rows first, and use Run all empty or out-of-date rows in this list to fill gaps, not to apply a change.

Keeping lists fresh

Columns can fill automatically for new rows, so contacts arriving from a routine are enriched without anyone pressing a button, and a list can be re-enriched on a schedule so data does not go stale between campaigns.