Skip to main content
The Classify record block (AI step) assigns one or more tags to a record from a fixed list of options you provide. It reads the record’s attributes, picks the tags that fit best, and returns them as a run value. A run value lives only inside the run, so it is not written back to the record. To save the answer, pair this block with an Update record block.

Inputs

Outputs

Reference the chosen tag downstream as {{ steps.classify_record.classification }}. Then map it into the value field of a later Update record block.

Behavior

The block sends the record’s attribute values to the model and limits the answer to the Tags list. The output is always one of your options, never free text. With Allow multiple tags off, it returns exactly one tag. With it on, it returns zero or more tags as a list. The output lives only in the run. Nothing is saved to the record until you add a write block. So a workflow that ends at Classify record classifies silently and stores nothing. Each run sends fresh tokens to the model. The cost grows with two things: how much attribute data the record holds, and how many tags you offer.

Gotchas

  • The classification is never saved on its own. The block returns a run value, not a record write, so the tag disappears when the run ends. Add an Update record block and map {{ steps.classify_record.classification }} into the target attribute.
  • Writing to a select attribute fails when the option does not exist. A select or multi-select attribute only accepts values that are already defined as options. Create the matching options first, spelled exactly as the tags you pass in. Do this before the Update record block runs.
  • Overlapping tags produce inconsistent single answers. When two tags describe the same record and you need one answer, the model has no rule for breaking the tie. Keep tag names mutually exclusive whenever Allow multiple tags is off.

Examples

Sorting new companies into ICP tiers

This workflow runs on a freshly created company. It classifies the company into an ideal customer profile (ICP) tier, the segment of buyer you most want to win. It then writes the tier to a select attribute so reps can filter by it. The block returns Tier 1. A following Update record block maps {{ steps.classify_record.classification }} into the company’s ICP tier select attribute.

Tagging an account with multiple product interests

This workflow classifies an account against a set of product lines and keeps every match. A multi-select attribute can then hold all of them at once. The block returns the list ["Analytics", "Integrations"]. A following Update record block maps {{ steps.classify_record.classification }} into the account’s Product interest multi-select attribute, whose options already include both values.

Credit cost worked example

The block is token-priced, so each run costs a variable number of credits. The number depends on the record’s attribute volume and the size of your Tags list. Suppose you classify every new company into four ICP tiers. Suppose you also create 400 companies a month. That is 400 token-priced runs. You then add an Update record write to save each tier. Record writes cost 1 credit each, so that is 400 more credits. Compare that monthly total against your plan’s allowance on Credits & pricing explained, which holds the per-plan numbers.

Source

Attio Workflows block library. Last reviewed against the Attio Help Center, June 2026.

See also