Feature · AI columns
AI columns that work row by row
AI is most useful when it understands your data. An AI column in Sheetlane runs a prompt against every row, references other cells, uses any context you've attached to the sheet, and writes the result back into the correct column.
- Reference other columns with {Column name} syntax
- Normalize outputs into select options, numbers, or dates
- Use sheet-level context (uploads or pasted markdown) for grounding
- Re-run a single cell or regenerate the whole column
- Per-row prompts
- Structured outputs
- Context-aware
What AI columns can do
An AI column is a prompt that turns one row of inputs into one row of output. It's the building block behind enrichment, classification, generation, and cleanup workflows.
- Generate text, emails, captions, briefs, summaries, outreach lines
- Clean messy data, normalize names, casing, country codes, currencies
- Classify rows, fit score, intent, sentiment, segment, priority
- Normalize select fields, keep enum cells strict against allowed options
- Summarize long inputs, meeting notes, transcripts, customer feedback
- Rewrite content, tone shifts, audience swaps, length changes
- Research from context, synthesize what you know about a company or contact
- Generate image prompts, feed downstream AI image columns
- Add more rows, append rows that match the existing schema
Example: lead personalization
A typical sales row starts with a few facts and needs structured outputs the team can ship.
Inputs (from your import)
- Company
- Website
- Role
- Industry
- Recent signal
AI outputs (one per column)
- Account summary
- Likely pain point
- Personalization line
- Email opener
- Suggested CTA
- Fit score (Select: A / B / C)
Research {Company} using {Website}. In 2 sentences, summarize what they sell and who their core customer is.Write a one-line personalization opener using {Recent signal} and {Role}. No greetings or sign-offs.Example: content repurposing
A blog row can fan out into a full week of channel-specific assets.
Inputs
- Blog title
- Target keyword
- Article summary
- Audience
- Offer
AI outputs
- LinkedIn post
- X thread
- Instagram caption
- Newsletter blurb
- Short video hook
- Image prompt
Why AI columns beat copy-paste AI
Copying rows into a chatbot breaks structure. Outputs come back as prose, get manually re-pasted, and lose the connection to the row they came from.
An AI column keeps the result attached to the row, typed into the correct column, and ready to be referenced by downstream AI columns, integrated columns, and action columns.
Realistic expectations
AI is great at structured, repetitive tasks where every row has consistent inputs. Set yourself up well by:
- Writing the prompt with concrete examples of the input you'll feed it
- Adding sheet context for product, audience, or brand voice you want reflected
- Picking the right output type (select for categories, text for prose)
- Reviewing the first 5-10 outputs before regenerating the whole column
Frequently asked questions
Can AI columns reference other columns?
Yes. Prompts use {Column name} placeholders. Sheetlane substitutes the cell value for the current row, so each row gets its own contextual prompt.
Can AI columns produce select values?
Yes. When a column is configured as a select or multi-select, AI outputs are normalized server-side against the allowed options. Free text never reaches an enum cell.
Can AI columns use uploaded context?
Yes. Attach context to the sheet by pasting markdown or uploading a .txt, .csv, .xlsx, .docx, or .pdf (up to 10 MB). The context grounds every AI column run on that sheet.
Can I regenerate an AI column?
Yes. Open the column menu and choose Regenerate to re-run the prompt across rows. Single-cell regeneration is also available from the cell menu.
Related reads
Add your first AI column
Start with a template or describe the workflow you want. No credit card required.