Guide
Enable Fibery to act as the central hub for storing, processing, and analyzing customer and internal call data. The space should allow for automatic ingestion of meeting transcripts, structured enrichment (via AI), and actionable insights that connect directly with CRM and task management workflows.
1. Capture Call Data
Save your call transcript into the Calls database.
Call database
It will consist of two types of fields:
2. Process call with AI
When a new record appears in the Calls database with all details synced from your meeting tool, we will need to:
Generate a summary of the transcript.
Assign a sentiment score (positive, neutral, negative).
Suggest tasks/action items based on the discussion.
To do this, we can configure a button.
Button will consist of two actions:
Both of them are calling for AI.
Summary & Tasks suggestion
The first one has to update AI summary Rich text field - there is will make the summary and propose Tasks to be created and done. Here is how the prompt looks like - you are welcome to adjust it.
[ai]
As an expert product manager, create an ultra-concise, single-paragraph summary per topic of this conversation in no more than 150 words.
Focus solely on the most critical topics without using section titles, separate paragraphs, action items, decisions made, or next steps sections.
The section's name should be in the same paragraph as the content. Prioritize extreme brevity and clarity.
The summary must be significantly shorter than the original conversation while conveying its essence.
One section should only contain the problem and the solution if there was any.
Format the result as follows:
"Section name: section summary
Section name: section summary
Section name: section summary."
Ensure that section names and summaries are on the same line.
As a separate section, add potential Tasks that has to be done according to the call notes.
Tasks must be listed with in a bulleted list format. Here is the customer conversation below:
"""
{{Transcript}}
"""
[/ai]
How this prompt is written, what is the logic?
The [ai] ... [/ai] markers are delimiters that tell the system “the content inside should be processed by AI.” Everything between them is the instruction set for how the AI should handle the transcript.
The placeholder {{Transcript}} is a variable/field reference — it refers to the Fibery field, AI has to be base it's answer on. When the automation runs, the transcript text is dynamically injected into that spot, and the AI uses it as the source material for summarization.
Inside the block, the AI must follow strict rules:
Produce ultra-concise summaries (≤150 words per topic).
Each topic becomes one line: "Section name: section summary".
Only problems + solutions per topic are included.
Tasks (if found) go into a final Tasks section formatted as a bulleted list.
No other structure (no decisions, no action items, no multi-paragraph output).
In short: [ai] marks AI instructions, {{Transcript}} injects the raw transcript text, and the AI returns a compressed, standardized summary + tasks.
Sentiment score
This one populates Number field Sentiment score. Here is how the prompt looks like
Sentiment scores are a metric for measuring customer sentiment.
Scores can range from 0-100, where 100 is the most positive possible outcome and 0 is the least.
Generate a sentiment score for the text below and return result as a number in a format [sentiment score] without brackets
"""
{{Transcript}}
"""
This automation block works in the same way as the earlier one, but here it’s designed to generate a numerical sentiment score instead of a text summary.
[ai] ... [/ai] → Marks the section that AI must process according to the instructions inside.
{{Transcript}} → Refers to the Fibery field, AI has to be base it's answer on.
Instruction inside → AI is told to evaluate customer sentiment on a 0–100 scale (0 = most negative, 100 = most positive).
Output format → Must return only the raw number (e.g., 73), no extra text, no brackets.
Target field → The AI output is stored in a Number field in the database (e.g., Sentiment Score).
In short: [ai] defines the AI job, {{Transcript}} provides the input text, and the AI responds with a single numeric sentiment score that gets written into a Number field.
3. Manage Action Items as Tasks
Suggested tasks can be automatically converted into Task entities. Check out this guide for more details.
To create a Task from a piece of text in your Meeting notes:
Highlight the text you want to turn into a Task.
Click the second icon from the left (📝) to create a new Task.
What happens next:
A new Task will be added to your Tasks database.
It will automatically appear in all Views that display Tasks.
If your Task - Meeting relation is one-to-many, the Task will be automatically linked to the Meeting it was created from.
If your relation is many-to-many, you’ll need to manually assign the Meeting.
We recommend using a one-to-many relation between Tasks and Meetings for the smoothest workflow.
4. Send Follow-up to Participants
You can also create a Button, that will send this AI summary to Participants. Here you can read more about Send email functionality Send emails and reply to emails
Here is the button I suggest you to create:
Let’s break down how Markdown is used in that message:
# Summary of Call: {{Name}} → the # makes this a heading level 1 (largest title). {{Name}} is a variable placeholder that will be replaced with the actual name of your conversation
Blank lines → in Markdown, blank lines create paragraph breaks, which is why each section appears as separate blocks.
Hi, folks! and the following text are just plain text, which will render as normal paragraphs.
{{AI Summary}} → another placeholder variable; when replaced with text, it will appear as plain content in the rendered Markdown.
So overall, Markdown here is very simple: it’s being used to create a title heading and separate content into clear paragraphs, while the placeholders ({{ }}) will be substituted dynamically.
You can check Markdown Templates for more examples, if you will decide to add more variables
Please, note that you will also have to connect the email account.
When you click the button, the email will be sent from your account and under your name.
If you prefer, you can replace the button with a rule — this allows you to define a default account that all emails will be sent from automatically.
Here is how sent email looks like
5. Optional: reporting on Sentiment
You can generate sentiment reports to gain insights at different levels:
Per employee (e.g., how customers respond to a specific rep).
Per contact (e.g., which customers often show negative sentiment).
Global overview (e.g., sentiment trends across all calls).
It can be done with Reports and Context Views mix
Context reports
Here is how it may look like
6. Optional: Connect with CRM
Connect with Company
If you store your CRM in Fibery, you can automatically link participants (e.g., meeting attendees or call participants) to existing CRM records based on their email domain. This is useful for associating people with the correct company without manual effort.
How it works:
Extract the email domain from each participant’s email address using a formula. For example:
If(Find(Email, "@") >= 0, Right(Email, Length(Email) - Find(Email, "@")), "")
Enable auto-linking (Auto-linking (Set Relations Automatically) ) in Fibery, matching the extracted email domain with the company’s domain stored in your CRM.
Connect with CRM contact
If you want to associate meeting participants from external tools with contacts in your CRM, you can do it directly via their email addresses.
How it works:
Create a 1:1 relation between your Meeting Contact database and your CRM Contact database.
Turn on auto-linking. Fibery will automatically connect participants to CRM contacts when the emails match.
Benefits / Use Cases:
7. Advanced queries & Reporting
With data structured in Fibery, you can ask powerful questions such as:
“Which customers asked about Product X this month?”
“What kinds of calls is Rep Y having?”
“How many negative-sentiment calls occurred last quarter?
Try our Fibery AI Agent Ask mode. Or even configure AI Agent in automations to create regular reports.