Guide
The Fibery AI Agent in Fibery is a built-in AI tool that can read structured or unstructured data and generate intelligent output — summaries, insights, transformations — based on natural language prompts.
The Smart Agent becomes especially powerful in automations when you have a repeating prompt that should be executed on a regular basis. Unlike ad-hoc analysis, this use fits situations where structured AI output can complement your ongoing reporting or workflows.
When to use it?
Use Smart Agent in automations when you want to:
Analyze performance regularly
e.g. Summarize how a person or team is progressing week-to-week (fast).
Continuously process feedback
e.g. Group and summarize feedback received in some timeframe
Generate periodic insights
e.g. Automatically create a weekly or monthly report from project updates, research notes, or changelogs.
Good Practice
Smart Agent is best used as a supplement, not a replacement, for structured reporting.
We recommend combining AI-generated reports with traditional visual dashboards or text-based summaries for full context and reliability.
Actions overview
When configuring automations in Fibery, you can invoke Smart Agent and put results into rich text field via [ai-agent] special tag, like this
[ai-agent]
{{Prompt}}
[/ai-agent]
In a rule it can look like this. Here we take text from Prompt field, put it into AI Agent and it will return result of the prompt when it is ready and put it into Description field (Below I will explain the setup for this use case)
Suggested structure
You can install ready-to-go template with this link.
To automate AI-powered reporting in Fibery, we recommend setting up two connected databases:
The AI Prompt database stores reusable prompts and serves as the source for automation. Each prompt includes a Prompt rich text field (used in automation). It also links to the AI Report database, where historical reports are saved. Automations (via rules or buttons) trigger the Smart Agent, which processes the prompt creates a new AI Report.
AI Prompt
This database stores reusable AI instructions and context. Each item defines:
Think of this as the template brain - what you want the AI to do.
AI Report
This stores the actual outputs produced by the Smart Agent. Each entry captures:
This acts as a history of executions, helping with:
Versioning and tracking changes over time
Comparing results from different prompts or inputs
Adding AI-generated insights to other workflows (dashboards, exports, reviews)
Relations Between Them
Each AI Report belongs to one AI Prompt
One AI Prompt can produce many AI Reports over time (1→∞)
This relation allows you to:
AI prompt fields overview
The AI Prompt database acts as the home for reusable AI instructions. Each prompt defines what kind of report or insight should be generated.
Button action logic
When:
Then - Step 2: Add AI Reports Item
What happens if you click the button?
A button click triggers the automation.
Fibery sends a static prompt to the Smart Agent, asking it to summarize recent feedback. The Smart Agent processes the data (based on the Highlights DB), generates a markdown summary, and returns the result in Description field.
The automation creates a new report entry in the AI Reports database with that content.
That’s it. Use this setup to automate recurring insights, track progress over time, and summarize unstructured data - like feedback or updates. It gives you a consistent, repeatable way to generate AI-powered reports without manual effort. By separating prompts and results, you keep things clean, reusable, and easy to manage.
Prompt examples
Fresh feedback
We capture feedback via Highlights database.
Create a clever report for all Highlights created for last 30 days. Try to group feedback by problems, ideas, and find repeating patterns.
Do not mention any customer data.
Do not add recommendations.
Here you can check sample report generated with such a prompt.
Create a report about Features "In Progress".
I want to know
Amount of features
Open bugs for every feature and overall evaluation of quality for every feature
When feature is going to be released (check feature release, and if it empty, try to guess)
Who works on the feature
Total MRR of all customers that requested these features (calculate it via Highlights linked to feature)
Do not generate PDF. Generate HTML and Text reports.
Here you can check sample report generated with such a prompt.