How AskHumans Works
From raw responses to weighted themes: the full pipeline.
Survey data enters as thousands of individual text responses and leaves as a small numbered set of stakeholder-weighted themes, each tagged by which parts of the organization raise it most. Here is how each step works.
Request Early AccessStep 1
Data ingestion
Bring your survey export in three ways: upload a CSV directly from your survey platform, connect via Qualtrics export format, or push responses programmatically through the REST API. AskHumans accepts responses in any column mapping; you do not need to reformat your data before upload.
Once your file lands, the system validates the response columns, strips personally identifiable fields you mark as excluded, and records the metadata you need for stakeholder weighting: respondent role group, department, and tenure band. If your export does not include tenure, you can leave that dimension unmapped; the weighting will run on the dimensions that are present.
Step 2
Clustering engine
AskHumans does not count keywords. It reads each response at sentence level, converts the language to a dense semantic vector, and groups responses by meaning similarity, not by the words used. Two responses saying the same thing in different vocabulary land in the same cluster.
The underlying method is closer to the HDBSCAN family of density-based clustering than to topic modeling (LDA) or keyword frequency. This matters: it handles the messy, multi-topic open-ended answer without forcing artificial topic counts, and it does not penalize responses that are short or colloquial.
Step 3
Stakeholder weighting
Not every theme means the same thing to every part of the organization. A concern voiced by 80% of managers in one department carries different weight than the same language surfaced by 3% of the company across every function. AskHumans assigns each theme a proximity score for each respondent group.
You configure weighting rules by setting which dimensions matter for your organization: role band, department, tenure, or custom metadata from your survey. The output shows not just which themes exist, but which groups cluster most densely around each. Leadership decisions can then be targeted rather than broadcast.
Step 4
Output and delivery
Results arrive in the format your team actually uses. An executive PDF brief with numbered themes, source quotes, and group-weight annotations. A CSV mapping each response to its assigned theme. A Slack or Microsoft Teams push that delivers the summary digest to the right channel after each survey cycle closes.