Write complete prompts to get clearer results with the Analytics Agent. This article answers frequently asked questions about the details to include in your prompts, such as the time period, grouping, visual format, and filters or exclusions.
Use Case
Use this article when you want to ask Analytics Agent about workflow or contract data and would like to work on drafting efficient prompts to get precise outcomes.
Prerequisites
| Features |
Ironclad Assistant |
| Permissions |
Users must belong to the Insights group. |
Writing Complete Prompts
When relevant, include these four details in your first message:
- Time period: The date range or recurring period to analyze.
- Grouping: How to group the results, such as department, role, record type, or owner.
- Visual format: The chart or output format you want.
- Filters and exclusions: The records to include or exclude, such as canceled workflows, specific agreement types, or test data.
Example Prompt
How many workflows did we complete each month over the last 12 months, broken down by record type, as a bar chart, excluding canceled workflows?
You don't need to include each detail every time. Just include the details that are relevant to the answer you're looking for.
Prompt Template
Use this template when you already know the metric, scope, and presentation you need:
[How many / What is the / Show me] [metric] for [record type or workflow],
[time period], grouped by [grouping], excluding [filter], as a [chart type].Example Prompt
What is the average number of days it takes Legal to approve a Customer Agreement workflow in the last two months, grouped by initial approver, excluding reassigned workflows, as a bar chart?
FAQs
How specific should my question be?
Be more specific about the parts of the result that matter to you. Adding those details to the first prompt helps Analytics Agent return the intended answer without requiring a separate refinement.
Less specific:
How many workflows did we complete?
More specific:
How many workflows did we complete each month over the last 12 months, broken down by record type, as a bar chart, excluding canceled workflows?
Should I specify a date range?
Yes, if you care about a specific timeframe.
For a one-time analysis, a fixed date range is usually clearest:
What was our average NDA turnaround time from April 1 through June 30, 2026?
For a recurring analysis, state the recurring period explicitly:
What is our average NDA turnaround time this quarter?
Without a date range, Analytics Agent may apply a default time period that doesn't match your intent.
Avoid leaving room for ambiguity:
What's our average NDA turnaround time?
Should I specify the chart I want as the output?
Yes, if you have a format in mind. Analytics Agent selects a format based on your data structure, but naming your preference helps it return the format you want immediately.
Common formats include:
- Bar chart
- Line chart
- Pie chart
- Stacked bar chart
- Table
- Downloadable CSV
Example prompt:
Show me contract volume by department as a stacked bar chart.
How do I ask for a breakdown by department, role, record type, or owner?
Specify how you want to group the results in your first message. Common groupings include record type, department, cost center, approver role, and owner.
Less specific:
How many NDAs did we sign this year?
Then, as a follow-up:
Split these by record type.
More specific:
How many NDAs did we sign this year, split by record type?
If you want to group the same data in more than one way, request all groupings at the same time:
How many NDAs did we sign this year, broken down by both record type and department?
What if I want to include or exclude specific records?
State your filters and exclusions in the original prompt. You can exclude canceled workflows, specific agreement types, test data, or other values supported by the relevant field.
Less specific:
How many workflows launched with legal involvement?
Then, as a follow-up:
Remove agreements from DPA, Intercompany, and Ventures workflows.
More specific:
How many workflows launched with legal involvement, excluding DPA, Intercompany Agreement, and Ventures Agreement workflows?
If you do not know which values exist for a field, ask a discovery question first:
What record types do we have?
You can use the results of that question to build a more specific analysis prompt.
What should I do if the result looks wrong or I do not understand the logic?
Ask Analytics Agent to explain its methodology, or define the criteria you want it to use. Ambiguous terms can refer to different fields or workflow events. For example, “approved” could refer to a specific approval step or the final workflow status.
Define the criteria in the original prompt when possible:
How many contracts have a limit of liability greater than $500,000? Show me which field you are using for “limit of liability.”
If a result surprises you, ask what is driving the number instead of rerunning the full query with different wording:
What is driving this number?
You can also define how to measure a metric:
Measure turnaround time from workflow launch to completion.
Can I ask Analytics Agent to combine charts or compare periods and segments?
Yes. Describe the comparison or combined view in one prompt instead of building separate charts and asking Analytics Agent to merge them.
Less specific:
Show workflows launched by month.
Then, as a follow-up:
Now also show completed workflows on the same chart.
More specific:
Show workflows launched and completed each month for the last 12 months on one chart.
You can also compare fixed periods:
Compare workflow volume from September 2024 through August 2025 with September 2025 through August 2026. What was the percentage change?
Is it ever better to ask a short question first?
Yes. A short question is useful when you are exploring the available data and do not yet know which grouping or filter makes sense.
For example:
What record types do we have?
What fields do you have for contract value?
Use the answers to build a more specific prompt. The goal is to avoid re-asking the same question because a relevant detail was missing, not to discourage legitimate step-by-step exploration.