AI Task Runner for Google Sheets
Overview
The AI Task Runner is an AI-powered execution engine that turns a natural-language description into a sequence of actual data cleaning operations. Instead of telling you what to do, it does it for you - profiling columns, standardising text, running fuzzy matches and more, all from a single description.
Unlike the Sheets AI Assistant which advises you on formulas and functions, the AI Task Runner executes multi-step workflows directly on your sheet. It is purpose-built for Google Sheets data operations and orchestrates all ten Flookup Data Processing Functions.
How to Open It
In Google Sheets, go to Extensions > Flookup Data Wrangler > Data Processing Functions . The sidebar opens with a grid of function cards. Select the AI Task Runner card.
Before describing your task, select the data range you are working with. The AI Task Runner uses your active sheet name and range context to construct accurate A1 notation references (e.g. `Sheet1`!A1:A500) in each step.
How to Use It
Once the AI Task Runner layout is open, you will see a single input field and a Run button:
- Describe your task. In the text area, describe the full data cleaning pipeline you need. For example: "Profile column A for issues, standardise the names to remove diacritics, then fuzzy match against column C at 85% threshold."
- Click Run. The AI analyses your description and generates a step-by-step action plan. Each step maps to a specific Flookup function with its required parameters.
- Watch the execution. Steps appear sequentially with progress indicators (✓ for complete, ↻ for running). Each function processes your data and writes results to the sheet before the next step begins.
If a step fails, the runner stops and shows which step encountered the error. All successfully completed steps are preserved.
Tools It Orchestrates
The AI Task Runner can plan and execute any combination of these ten functions:
| Function | What It Does |
|---|---|
| Profile Column | Scans a column for data quality issues and returns a report with one-click fix suggestions |
| Standardize Data | Removes diacritics, punctuation, stop words or learns a transformation pattern from examples |
| Fuzzy Match Data | Searches a lookup value against a table column and returns the best match with a similarity score |
| Fuzzy Merge Rows | Finds the best match for each lookup value and returns the entire matching row |
| Deduplicate Data | Removes duplicate rows from a data range using exact or fuzzy similarity matching |
| Smart Deduplicate | Detects duplicate groups using four silent strategies (exact, normalised, phonetic, fuzzy) with confidence ranking |
| Extract Unique Rows | Filters a dataset to return only distinct rows based on text similarity |
| Enrich Data | Matches values against built-in reference tables (countries, US states, suffixes, TLDs) or uses AI-powered custom enrichment |
| Fuzzy Conditional Formatting | Applies colour-coded highlighting to groups of similar rows with configurable colour schemes |
| Compare Similarity | Computes text similarity scores between two columns of strings |
How It Generates Steps
When you describe a task, the AI Task Runner parses your natural-language description into a structured JSON action plan. Each step includes the exact function name and its required parameters using your the sheet actual A1 notation.
For example, the description "Profile column A, standardise names, then fuzzy match at 85%" generates:
- Profile Column with range set to your selected column
- Standardize Data to remove diacritics from the profiled column
- Fuzzy Match Data with a threshold of 0.85 against the specified target range
The AI uses your active sheet context (sheet name, row count, selected range) to construct accurate range references. The parameter parsing is a single AI operation - all subsequent execution is pure deterministic code with zero additional AI cost.
Custom (AI) Enrichment
In addition to the AI Task Runner, the Enrich Data function includes a Custom (AI) option in the entity type dropdown. This lets you enrich any column with information beyond the four built-in reference tables.
For example, select Custom (AI) and type "Industry" to enrich a column of company names with industry classifications. Or type "Stock ticker" to add trading symbols to a list of public companies. The AI enriches each value in one batch operation and writes the results next to your input column.
Custom (AI) Enrichment is also available as a step within the AI Task Runner.
Data Privacy
Your data privacy is a priority:
- Minimal data sharing: For workflow planning, only your task description and sheet metadata (sheet names, active range) are sent to the AI service. Your full dataset never leaves your spreadsheet. For Custom (AI) Enrichment, only the values you want to enrich are sent - not the entire sheet.
- No model training: Your data is used solely to generate your requested workflow or enrichment. It is not stored or used to train or improve AI models.
- Purpose-limited: The AI only processes requests related to Google Sheets data operations. Requests unrelated to data cleaning or spreadsheet management are declined.
Usage and Limits
The AI Task Runner and Custom (AI) Enrichment are included in the Data Nova plan. Each workflow plan consumes a small number of AI credits. Column profiling is always free and consumes no credits.
| Plan | AI Credits | AI Features |
|---|---|---|
| Free Forever | None | Not available |
| Data Nova | 10,000 – 250,000 | AI Task Runner, Sheets AI Assistant, Custom (AI) Enrichment |
Credits never expire. If your credits are exhausted, AI features stop working and you revert to Free Forever limits. Purchase a new plan to restore full access and get more credits.
The AI Task Runner is included in the Data Nova plan.
Use AI to execute multi-step data cleaning tasks automatically with credits that never expire.
Best Practices
- Start with profiling. Always include a Profile Column step at the beginning of your workflow. It detects issues before you clean, so you know what to standardise or deduplicate.
- Be explicit about column ranges. Instead of "match column A to my master list", say "fuzzy match column A against column C in Sheet1, returning column D". The more specific your description, the more accurate the generated steps.
- Specify thresholds when needed. If you need a specific similarity threshold, include it in your description (e.g. "match at 90%"). Otherwise the default of 0.75 is used.
- Test on a sample first. If you are describing a complex multi-step workflow, test it on a small subset of your data first to verify the results before running on the full dataset.
- Check step progress. Each completed step shows a green checkmark. If a step fails, review the error message and adjust your description before retrying.
Frequently Asked Questions
How is the AI Task Runner different from the Sheets AI Assistant?
The AI Assistant advises you by returning formula suggestions that you apply manually. The AI Task Runner executes cleaning steps directly on your sheet - describing a pipeline like "profile, standardise, then fuzzy match" and watching all three steps run sequentially without further input.
Does the AI Task Runner modify my data automatically?
Yes, it writes results to your sheet after each step completes. You can cancel at any time and partial results are preserved. Always work on a copy of your data if you want to preview the results first.
What happens if a step fails?
The runner stops at the failed step and shows the error. All successfully completed steps remain on your sheet. You can adjust your task description and retry.
How many steps can I run in a single workflow?
There is no hard limit but the total execution time must fit within Google Sheets' 330-second processing window. For large datasets, consider scheduling individual functions via Schedule Functions instead.
Can I use Custom (AI) Enrichment without the AI Task Runner?
Yes. Open the Enrich Data function in the sidebar, select Custom (AI) from the entity type dropdown, describe what you want to know and click Run. It works as a standalone function.