A full qualitative research and evidence environment inside Microsoft Excel. Import survey data, code responses with progressive AI learning, challenge findings with adversarial review, and generate publication-ready reports — all without leaving the workbook.
Four stages inside Excel from raw spreadsheet data to defensible findings.
Select a range in Excel, create an import job, and map columns with AI assistance.
Code 20 responses by hand. The AI learns your patterns and auto-codes the rest.
Red-team your findings, check for bias, test claims against the data with evidence functions.
Generate narratives, cross-tabulations, and theme charts — written back into the workbook.
Type =CHOKMI.
in any cell to access these functions. Single cells return rich entity data types with confidence
and reasoning. Ranges batch-process automatically.
| Function | Description | Returns |
|---|---|---|
| Core AI | ||
CHOKMI.SENTIMENT(text) |
Classify sentiment: Positive, Negative, Neutral, Mixed | Entity |
CHOKMI.CODE(text, method?) |
Assign a qualitative code using the specified method | Entity |
CHOKMI.CONFIDENCE(text) |
Return a 0-1 confidence score | Number |
CHOKMI.SUMMARIZE(text) |
Produce a 1-2 sentence summary | String |
CHOKMI.PII_CHECK(text) |
Detect personal identifiable information | Boolean |
| Research | ||
CHOKMI.LEARN(texts, codes) |
Train the AI on your text-code pairs | String |
CHOKMI.EVIDENCE(claim) |
Test a research claim against survey data with chi-square | Matrix |
CHOKMI.BIAS_CHECK() |
Scan coding patterns for demographic bias | Matrix |
CHOKMI.RED_TEAM(claim) |
Adversarial peer review of a research finding | Matrix |
CHOKMI.NARRATIVE(audience?) |
Generate a research narrative for executive, academic, or policy audiences | Matrix |
CHOKMI.VOICES(theme, count?) |
Find the most vivid, representative quotes for a theme | Matrix |
| Semantic | ||
CHOKMI.SIMILAR(ref, targets, threshold?) |
Find semantically similar texts using embedding similarity | Matrix |
CHOKMI.CLUSTER(texts, n?) |
Group texts into clusters by embedding similarity | Matrix |
CHOKMI.BETWEEN(textA, textB) |
Cosine similarity score between two texts | Number |
CHOKMI.ANCHOR(texts, codes, input) |
Few-shot coding using human-coded anchor pairs | String |
CHOKMI.CONTRADICT(text) |
Find responses that semantically contradict a statement | Matrix |
| Analysis | ||
CHOKMI.CROSSTAB(colA, colB, chart?) |
Cross-tabulate two columns with chi-square statistics | Matrix |
CHOKMI.FINGERPRINT(text) |
Discourse analysis: metaphors, register, formality, agency | Entity |
CHOKMI.HYPOTHESIZE() |
Generate novel testable hypotheses from coded data | Matrix |
CHOKMI.SILENCE(topic?) |
Detect what respondents did not say (blind spot analysis) | Matrix |
| Live (streaming) | ||
CHOKMI.SATURATION_LIVE() |
Live-updating saturation percentage (refreshes every 15s) | String |
CHOKMI.CODING_LIVE() |
Live-updating coded response count (refreshes every 10s) | String |
CHOKMI.PROGRESS_LIVE() |
Live-updating import progress (refreshes every 5s) | String |
The task pane organises the research workflow into focused panels. Each tab maps to a stage of the qualitative research process.
| Tab | Purpose |
|---|---|
| Connection | Authenticate with API token or username/password. Manage session state. |
| Survey Context | Select organisation and survey. View survey health and workflow status. |
| Workbook Import | Scan the active range, create an import job, choose AI-assisted or manual mapping. |
| Current Job | Run AI mapping, write preview sheets, resolve blockers, and finalise the import. |
| Column Mapping | In-pane visual editor for unresolved column mappings before import proceeds. |
| Workbook Coding | Create codebook templates, sync codes, select method, run qualitative coding. |
| Workflow Automation | Write workflow plan, quality triage, and health snapshot sheets into Excel. |
| Research Pipeline | One-click macro: import, code, cross-tab, format, chart, and narrative generation. |
| AI Assistant | Chat interface to ask questions about your survey data from within Excel. |
| Activity Log | Real-time log of API calls, import progress, and diagnostic messages. |
What makes the Excel add-in different from a generic AI plugin.
Code 20 responses by hand with CHOKMI.LEARN(). The AI learns your coding patterns,
then auto-codes the remaining responses with confidence scores and reasoning. Each correction
improves the next batch.
Functions like CHOKMI.SENTIMENT and CHOKMI.CODE return expandable Excel
data types. Click the cell icon to see confidence, reasoning, and alternative labels alongside the
primary result.
SATURATION_LIVE, CODING_LIVE, and PROGRESS_LIVE auto-refresh
in the cell, giving you a real-time dashboard of coding progress and thematic saturation without
leaving the sheet.
One click runs the full sequence: auto-code uncoded responses, build a theme frequency pivot, apply confidence formatting, insert a chart, and generate an executive narrative. All results are written back as structured worksheets.
When AI mapping encounters ambiguous columns, the task pane surfaces an inline editor. Resolve each mapping decision without switching to the browser, then continue the import from the same panel.
RED_TEAM stress-tests your findings with attack vectors. EVIDENCE tests
claims against the data with chi-square statistics. BIAS_CHECK scans coding patterns
for demographic disparity. Challenge your own work before reviewers do.
Two paths depending on your setup.
Open the Chokmi web app, go to Integrations Hub, and click Install Excel Add-in. The add-in is registered automatically via the Office dialog API. Recommended for most users.
Deploy via the Microsoft 365 admin centre for org-wide availability. Contact your Chokmi account manager for the centralised add-in catalogue package.
Show us how your team codes responses, tests claims, and reports country-level findings today. We will show you how Chokmi structures it into one defensible operating workflow.