Matik automates insights by using AI to analyze your dynamic content results to generate summaries, trends, key takeaways, and other insights. Automated Insights is available as an add-on to your Matik plan. Please reach out to your Technical Account Manager or Relationship Manager if you are interested.
Automated Insights work by taking the results of your dynamic content, combining them with your prompt and relevant business context, and sending them to a LLM (Large Language Model) to create clear cohesive sentences. The output is returned in the format you specify, making it easy to insert insights directly into your presentations or other Matik templates.
Automated Insights can be output as:
- Text: Automate elements like headlines, sub-headlines, callout boxes, and full length paragraphs, including summaries, trends, key takeaways, next steps, and other insights, all tailored to the data and context of your specific template generation.
- Tables: Output your automated insights as structured rows/columns that can be inserted as nicely formatted tables in your template. Automate elements like summary tables, milestone updates, or recommendation overviews.
When you generate content with Automated Insights, Matik executes your chosen dynamic content with the specified inputs, analyzes the returned data, and uses that context to create insightful, accurate, and relevant text.
Output Formats
Automate Text Insights
Use Text Insights whenever you want to populate text elements within a presentation, document, or email. For example, you could:
- Automate the headline or subheadline on a slide based on the slide content
- Automate the callout box next to a chart based on the data
- Automate the text body of an email. (When using the Code Editor, you can even include HTML formatting as part of your Text output.)
- Automate the key takeaways within a document
Text Insights can be tagged and used just like any other Text-type dynamic content in Matik, so you can apply text box formatting/styling and leverage rich text tags as usual. A single Text Insights DC can also return several separately taggable pieces of copy at once. See Sub-content: Multiple Outputs from One Prompt.
To start:
- Create a new piece of dynamic content
-
Select Insights
- Select "Text" as your output format
- Instruct Automated Insights on what content to generate. You have two options for doing this:
- Use the Form for a guided setup. The form helps you select a piece of dynamic content, define what you insights you want to automate, and structure the output into a defined format.
- The Free Prompt allows you to create an open-ended prompt to the LLM. You can customize instructions for how the LLM should handle your dynamic content results, specify the format and style of the output, and include any additional context needed to reliably produce your desired result. You can also reference dynamic content and inputs directly in the prompt to pull in the required data.
Automate Table Insights
Use Table Insights when you want to output your content as structured rows/columns that can be inserted as nicely formatted tables in your template. Table Insights are a good choice when you want to summarize, format, or organize information for easy comparison or analysis.
Example use cases include:
- Summarize key risks or opportunities based on call transcripts and CRM data.
| Item | Type | Description | Recommendation |
|---|---|---|---|
| Stakeholder disengagement | Risk | Key decision-makers haven’t joined recent meetings. | Contact stakeholders to confirm involvement. |
| Expansion interest | Opportunity | Customer wants to expand usage to the Sales team. | Share tailored proposal for Sales team use case. |
- Identify key concerns based on customer support ticket data.
| Topic | Example Ticket Quote | Frequency | Suggested Action |
|---|---|---|---|
| Login issues | "I can't reset my password after multiple attempts." | High | Improve password reset workflow and update help articles. |
| Feature requests | "Is there a way to schedule automated reports?" | Medium | Review requests with product team for roadmap consideration. |
| Billing questions | "My invoice is missing last month’s usage adjustment." | Low | Clarify usage adjustment in billing FAQs. |
- Summarize next steps based on meeting notes in a summary document or follow-up email
| Action Item | Owner | Due Date | Status |
|---|---|---|---|
| Share Q2 roadmap with client | Alex | 2024-06-10 | Pending |
| Schedule follow-up meeting | Jordan | 2024-06-15 | In progress |
| Send billing summary | Sam | 2024-06-12 | Completed |
Table Insights can be tagged and used just like any other Table-type dynamic content in Matik, so you can apply table formatting as usual. If you want the same column headers on every generation, declare them as sub-content fields. See Sub-content: Multiple Outputs from One Prompt.
To create Table Insights:
- Create a new piece of dynamic content
-
Select Insights
- Select "Table" as your output format
- Use the Free Prompt to instruct Automated Insights on what to generate. See Using the Free Prompt to create Insights Dynamic Content. For optimal results, make sure that you specify what columns you want your Insights Table to have as part of the output requirements.
Configuring Automated Insights
Using the Form to Create Insights Dynamic Content
When creating Text Insights, you have the option to use the Form to be guided through your Insights setup. After selecting Text as your output type:
-
Select Form
- Choose your function under I want to: Describe, Analyze, Summarize, or Compare
- Describe: Provides a description of your dynamic content output
- Analyze: Identifies trends or patterns when your dynamic content returns complex data
- Summarize: Provides a high-level summary of the data returned by dynamic content
- Compare: Looks at two different types of dynamic content to highlight similarities or differences
- Select the Dynamic Content you want to analyze. Note that if you select Compare, you will need to select at least 2.
- Select the length and type of output you desire: One Sentence, One Paragraph, Bulleted List, or Numbered List. For list outputs, specify the desired list length.
- Add context to guide the output. Use this section to explain how Automated Insights should interpret your dynamic content, clarify your objectives, and include any background information that will help it generate the result you need.
- Tag Automated Insights dynamic content as you would any other text in your Matik templates.
Using the Free Prompt to create Insights Dynamic Content
The Free prompt option is supported for Text and Table Insights. The Free Prompt option offers maximum flexibility for tailoring your Automated Insights. While you’re not limited to a specific format, following these guidelines will help you create prompts that produce the most accurate and useful results:
Clear Task Instructions: Explain exactly what you want Automated Insights to do with the provided data.
- Example: “Summarize this sales data into three short bullet points highlighting key performance trends.”
- Include fallbacks: “If any data point is missing, state ‘Data not available’ instead of guessing.”
Structure Your Prompt with Explicit Sections: Break your prompt into clearly labeled sections (such as Task, Data, Output Requirements, and Context) to improve consistency and reduce misinterpretation as dynamic content is inserted.
Why this helps:
- Reduces “bleed” between instructions, data, and formatting requirements.
- Makes prompts easier to maintain and reuse across templates.
- Provides a stable framework so you don’t rewrite prompts for different datasets.
How to do it:
- Use visible labels like
### Taskor=== Data ===to separate sections. - Add short descriptors in the Data section so Matik understands what each value or table represents (e.g., “Total Revenue (USD): …” or “This table lists regions in column A and total sales in column B.”).
Dynamic Content and Inputs: When building the prompt in Matik, insert dynamic content and inputs using the following format:
- Dynamic content:
{{dynamic_content_name}} - Inputs:
&:input_name
Examples:
Total Revenue (USD): {{total_revenue}}-
Sales by Region (table): {{sales_by_region_table}}
Follow with a brief note: “This table lists regions in column A and total sales in column B.”
Output Formatting Requirements: Define exactly how you want the output returned:
- Length: “Under 300 characters” or “No more than 2 sentences.”
- Format: “3 bullet points with bolded headers” or “single paragraph of plain text.” You can also specific exact formats to for metrics or other data. See example below.
- Tone: “Professional and concise” or “Friendly and persuasive.”
- (For Table Insights): make sure you specify what columns should be included in your final table output.
-
Rich text: Matik supports basic rich text formatting, which you can direct Matik to apply in the custom outputs you create. This can often be achieved by specifying you want the output formatted in HTML. See example below. Please note that Matik's support of Rich Text is more limited than full HTML. See Rich Text Formatting to check which tags are supported in which template types.
Business Context: Provide background so Matik can interpret the data correctly:
- Example: “Matik is a B2B SaaS platform. Audience: CSMs presenting quarterly performance updates to enterprise clients. Focus on retention and upsell metrics.”
- Include KPIs, goals, and themes that matter most so Automated Insights emphasize the right points.
Example Prompt
Additional Tips
If you are creating multiple pieces of insights dynamic content that will have the same business content or share other elements, you can add the text to a static text piece of dynamic content and tag that in the prompt in your content section.
Sub-content: Multiple Outputs from One Prompt
Beta Feature: Insights sub-content is currently in beta. Contact your Technical Account Manager for access.
Sub-content lets a single Insights prompt produce several separate pieces of copy, each with its own template tag.
Say a slide needs a sub-headline, a body paragraph, and a callout about customer impact. You can write one prompt that analyzes the data once, then declare the pieces of copy you want out of it: opening_subheadline, executive_summary, and customer_impact. Each one becomes its own tag, like {{QBR_Narrative.opening_subheadline}}, that you can place anywhere in your template.
Benefits include:
- Less to manage. When your analysis needs to change, you update one prompt instead of keeping several of them in sync.
- Copy that agrees with itself. Every field comes from a single read of the data, so your sub-headline, body copy, and callout all reinforce the same story.
Each field holds as much or as little copy as you need, from a few words to a full multi-paragraph write-up, and you can ask Matik to apply rich text formatting within a field. See Rich Text Formatting for which tags are supported in each template type.
How Insights sub-content works
With query-based Dynamic Content, sub-content fields come from the columns your query returns, so Matik detects them by running the query. Insights has no query to inspect, so you declare the fields yourself. Matik builds those declared names into the request it sends to the AI, which keeps the same set of fields coming back on every generation.
Sub-content is available on Insights DC with Text or Table output. Image output does not support it.
-
Text Insights: each declared field holds its own piece of copy, and
{{DC_name.field_name}}inserts one of them. - Table Insights: your declared fields become the table's column headers. Matik pins the header row to exactly the fields you declared, in the order you declared them, so your table keeps the same shape on every generation.
Setting up Insights sub-content
- Open or create an Insights DC and select Text or Table as the output format.
- Write your prompt on the Free Prompt tab. Describe the analysis you want and the separate pieces of copy you want out of it.
- Turn on sub-content. In the Sub-content section, switch on Use as subcontent.
- Define your fields. Add a row for each field under Define column names, or click Suggest Columns to have Matik propose them from your prompt.
- Name your fields in the prompt. Matik flags any declared field your prompt never mentions. Those fields are still generated, but naming them in the prompt is the most effective way to improve what comes back.
- Test before you save. Click Test Insights to see the value Matik returns for each field.
- Tag your template. The Template tags list at the bottom of the Sub-content section shows the exact tag for each field, ready to copy.
Note: Field names cannot contain spaces or the characters . : | { } ( ) ,, and each name must be unique. Settle on your names before you tag a template: tags match on the field name, so renaming a field afterward breaks the tags that reference it.
Using Suggest Columns
Suggest Columns reads the prompt you have written and proposes a field name for each distinct piece of output it thinks you are asking for. It reads your instructions, not your data, so the suggestions are only as specific as your prompt.
- Write your prompt first. "Summarize this account's performance" produces generic names. "Give me an executive summary, the top three risks, and recommended next steps" produces the fields you actually wanted.
- Suggest as often as you like. Matik adds new names and leaves your existing ones alone, so you can refine your prompt and suggest again.
- Close the loop. Once your fields are settled, mention them by name in your prompt.
When to Consolidate into One Prompt with Sub-content
Sub-content is most useful when your outputs are different cuts of the same underlying analysis. When two outputs answer genuinely different questions, keep them in separate Insights DC. Ask yourself:
If one of these outputs changed, would the others have to change too? If yes, they belong in one prompt as separate fields. If no, use separate DC.
Example Use Cases
One analysis, several cuts of copy. A quarterly business review deck needs a sub-headline on the opening slide, a full summary in the slide body, a callout on adoption, and a closing list of next steps. All four describe the same thing: how the account did this quarter and what they should do next. If your read on adoption shifts, every one of those pieces has to change with it.
That is one Insights DC, named QBR_Narrative, over the account's usage, adoption, and support data, with four declared fields:
-
opening_subheadlinefor the title slide -
executive_summaryfor the slide body -
adoption_calloutfor the callout box -
next_stepsfor the closing slide
Tag them as {{QBR_Narrative.opening_subheadline}}, {{QBR_Narrative.executive_summary}}, and so on. One prompt to maintain, and the sub-headline and the body copy are telling the same story because they came out of the same analysis.
Two analyses in one deck. An executive business review draws on two different bodies of data that call for two different reads. Consolidate each one, and keep them apart from each other:
-
Account_Narrative, over product usage data, with the fieldsheadline,usage_story, andexpansion_opportunity. -
Support_Health, over support ticket data, with the fieldsticket_summaryandtop_issue_theme.
That is five outputs from two DC rather than five. They stay separate because they answer different questions from different data, and because the support commentary should stand on its own rather than being shaped to fit the expansion story.
Same data, still separate. Support ticket data can feed both a customer-facing "here is how we supported you" QBR narrative and an internal churn-risk assessment. Same rows, but different questions for different audiences, and you do not want the internal risk read softened to match the customer narrative. Keep these in separate Insights DC. Sharing a dataset is not on its own a reason to share a prompt.
Sub-content limitations
- Text and Table output only. Insights DC with Image output cannot use sub-content.
-
Field names have to be tag-safe. No spaces and none of the characters
. : | { } ( ) ,. Names must be unique within the DC. - Renaming a field breaks its existing tags. Tags resolve by field name, so update your templates if you rename a field.
Limitations, Testing, and Optimizing
If your dynamic content returns very large datasets, Automated Insights may not be able to process all of the data at once. The exact limit isn’t based on a fixed number of rows or characters—it depends on the overall size and complexity of the data—so the point at which limits are reached can vary.
We recommend stress-testing your Insights Dynamic Content before rolling it out broadly to make sure it runs smoothly.
If limits are reached while testing, you’ll see an error message.
If limits are reached during generation, a warning will appear in the presentation generation logs.
💡 Best Practice: For large tables, consider aggregating, filtering, or summarizing the data before passing it to Insights. Smaller, more focused datasets not only avoid processing limits but also lead to clearer, more relevant outputs.
Best Practices
-
Always proofread and validate results before sharing—Matik is not responsible for errors produced by Automated Insights Dynamic Content.
If you plan to use Insights Dynamic Content in situations where it will not be reviewed first, make sure to include a note that the text was AI-generated and may contain inaccuracies.
Matik's AI cannot scrape URLs, browse the web, or perform internet searches. It is limited to the data returned by your Dynamic Content and the context you provide.
Thoroughly test your Automated Insights dynamic content under a range of inputs and dynamic content values.
- Ensure the dynamic content being used by Automated Insights has a clearly written description.
- Test multiple variations of descriptions to see which one produces the best results for your use case.
- Add context to the significance of different value ranges in your dynamic content description if you would like them to appear in your generated insights.
- Name your dynamic content, inputs, and data columns with easy-to-understand names to give the Automated Insights context to your data.
- To generate Automated Insights, Matik must first run the dynamic content being fed into Automated Insights and then generate the insights. If you find Automated Insights is taking a while to generate look into optimizing your dynamic content runtime.
Security
All AI-powered features in Matik and add-ons created by Matik use GPT-5.1 from Open AI under the hood to provide various functionality. When Automated Insights Dynamic Content is created, it will have access to the Dynamic Content data returned from any input set as well as the queries, description, and other setup information about that referenced Dynamic Content to provide content on the return data. As specified in the Open AI API policy, any data fed into the API will not be used by Open AI except to provide the responses to the API prompts. More information can be found in the Open AI Terms of Use and Matik’s Privacy Policy. Matik does not store any data used in these features and add-ons at this time. We’ll notify our users if this changes.
Comments
0 comments
Please sign in to leave a comment.