Tray AI Workshop: Intro to infusing processes with AI
This is a 'Project' template which means that it contains a group of workflows that work together to achieve a particular aim
Tray AI Workshop: Infusing Processes with AICopy
OverviewCopy
You don't need to build a full agent to harness the power of AI in your business processes. This self-paced workshop demonstrates practical ways to infuse AI into existing workflows using Tray's out-of-the-box AI connectors—collectively known as the "Merlin AI palette."
What You'll LearnCopy
Through three hands-on lessons, you will:
Explore key AI connectors in the Merlin AI palette
Use the Merlin generative AI connector to analyze qualitative data
Apply Intelligent Document Processing (IDP) to extract information from files and images
Use Cases CoveredCopy
Basic Summarization:
Create and email a call summary from a transcript
Advanced Summarization:
Extract targeted insights from a call transcript using custom prompts
Document Processing:
Extract financial data from invoice files
PrerequisitesCopy
Access to a Tray workspace
Basic familiarity with the Tray workflow builder
An email address to receive workflow outputs
Getting StartedCopy
Import the Workshop TemplateCopy
Navigate to your personal workspace in Tray
Select Create new project from template
Search for and import the "Tray AI Workshop: Intro to infusing processes with AI" project template
Once imported, you'll see three pre-built workflows corresponding to the three lessons below.
Lesson 1: Basic SummarizationCopy
Objective: Use out-of-the-box connectors to create a call transcript summary and email it to yourself.
Part 1: Get a TranscriptCopy
In this section, you'll simulate retrieving a call transcript from an API and inspect the data.
Open the "Lesson 1: Basic summarization" workflow
Click Run workflow
Click Show logs in the top left
In the logs panel, select the grey tile at the bottom of the list
Choose the "Get call transcript" log in the second panel that opens
Expand the output log and locate the transcript property
Close the logs panel
Part 2: Generate Email ContentCopy
Now you'll configure Merlin connectors to generate a subject line and summary from the transcript.
Move the Terminate step below the first Merlin connector (labeled "Create a subject line")
Click the "Create a subject line" Merlin connector to open its configuration
Select "Short Title" from the Format dropdown
Click Run workflow
Open logs and inspect the output of the 1st Merlin connector
Move the Terminate step below the second Merlin connector (labeled "Create a summary")
Click the second Merlin connector to open its configuration
Map the "Text" input field to the "transcript" field from the "Get call transcript" step
Tip: Click into the Text field and select the transcript value from the data tree
Select "Short sentence" from the Format dropdown
Click Run workflow
Open logs and inspect the output of the 2nd Merlin connector
Delete the Terminate step entirely
Click Run workflow one final time
Check your email inbox for the generated summary
Expected Result: You should receive an email with a generated subject line and brief summary of the call transcript.
Lesson 2: Advanced SummarizationCopy
Objective: Extract targeted, structured insights from a call transcript using custom prompts with the Merlin Text Generation connector.
Create a Custom SummaryCopy
In this lesson, you'll write a custom prompt to generate a more detailed, structured call summary.
Open the "Lesson 2: Advanced summarization" workflow
Click the + icon below the "Create a subject" Merlin connector
In the connector search, type "Generate text" and select the Merlin Text Generation connector
In the connector configuration:
Select Claude from the Model dropdown
Click this link to access the custom prompt (copy it to your clipboard)
Paste the prompt into the "Instructions" field
Set Temperature to 0.1
Set Max tokens to 2000
Click Run workflow
Check your email inbox for the results
Expected Result: You should receive an email with a more detailed, structured summary that follows the format defined in your custom prompt.
Key Takeaway: Custom prompts give you precise control over AI outputs, allowing you to extract specific insights in your preferred format.
Lesson 3: Intelligent Document ProcessingCopy
Objective: Use Intelligent Document Processing (IDP) to extract structured data from an unstructured invoice file.
Extract Invoice DataCopy
In this lesson, you'll configure IDP to answer specific questions about an invoice document.
Open the "Lesson 3: Intelligent Document Processing" workflow
Click Run workflow
Click Show logs in the top left
Select the most recent grey log tile and choose the "Create file" log
In the log output, hover over the "URL" field and click Follow link
The invoice will open in a new browser tab
Take note of the invoice amount shown on the document
Return to the workflow builder
Click the "Get invoice information" step to open its configuration
In the "Queries" list, add a new query:
"How much was this invoice?"
Delete the Terminate step
Click Run workflow
Open the logs and inspect the output from the "Get invoice information" step
Expected Result: The IDP connector should return the invoice amount extracted from the document.
Try Additional QueriesCopy
Experiment with IDP by adding these queries to the "Queries" list:
"What is the invoice number?"
"What is the invoice date?"
"Who is the invoice addressed to?"
Run the workflow after each addition and inspect the outputs.
Key Takeaway: IDP can extract specific information from unstructured documents without manual data entry, making it ideal for automating document-heavy processes like accounts payable, contract review, and compliance.
Next StepsCopy
Congratulations! You've completed the AI Infusion workshop. You now have hands-on experience with:
✅ Basic AI text summarization ✅ Custom prompt engineering for targeted insights ✅ Intelligent document processing for data extraction
Apply What You've LearnedCopy
Consider how you might apply these techniques to your own workflows:
Customer Support: Automatically summarize support tickets or call transcripts
Sales: Extract key insights from discovery calls and send follow-up summaries
Finance: Process invoices, receipts, and financial documents at scale
HR: Analyze interview transcripts or extract data from resumes
Additional ResourcesCopy
Tray Academy - Explore more AI-focused courses
Tray Community - Connect with other builders
Merlin AI Documentation - Deep dive into AI connectors