Overview
Workflows are one of AskElephant's most powerful features. They automate tasks so you don't have to repeat the same steps over and over. But like any automation, things can go wrong. A workflow might not run, produce incomplete results, show errors, or behave unexpectedly.
The good news: most workflow problems are fixable, and they usually fall into a few predictable categories. This guide walks you through diagnosing and solving workflow issues so you can get back to automating your work.
Key Terms to Know
- Workflow: An automated process you build that runs AI-powered tasks without you doing them manually each time.
- Node: A single step or action in your workflow. Think of nodes as building blocks stacked together to accomplish a goal.
- Trigger: The event that starts your workflow. Examples: a new meeting is recorded, a deal is updated in HubSpot, or a specific word is mentioned in a call.
- Variable: A container that holds information from one node and passes it to the next. Variables let your workflow remember and use data throughout its steps.
- Prompt Node: A step in your workflow where you give AI instructions to analyze or process information.
- Workflow History: A log showing every time your workflow has run, whether it succeeded or failed, and details about what happened.
Before You Troubleshoot: Save Your Work
This is critical. Workflows exist in your browser as you build them. If you close your browser, turn off your computer, or navigate away from the page, unsaved work will be lost.
Save frequently:
- Every time you make a change, click the Save button in the top right corner
- You'll see a number next to "Save" (like "Save 5")—that number shows how many changes haven't been saved yet
- Once saved, your workflow is safe in the cloud
Pro tip: If you see the Save button showing a high number (like "Save (25)" or higher), you're not saving frequently enough. Save after every few changes.
The Most Common Workflow Problem: Missing or Incorrect Variables
This is the #1 reason workflows fail or produce empty results.
What variables are and why they matter:
- Each node in your workflow produces output—information that the next node might need
- A variable is how you reference that output
- If you don't include variables in your prompts, the AI has no data to analyze
Example of what goes wrong:
- You create a workflow that should summarize a meeting
- You set up a "New Meeting" trigger (this gives you access to the meeting)
- You add a "Run Prompt" node with instructions like "Summarize this meeting"
- But you never tell the prompt which meeting to summarize
- Result: The AI generates a summary with no actual content
How to fix it:
- Identify what data you need
- What information should your workflow analyze or use?
- Is it from the trigger (like a meeting transcript)?
- Is it from another node (like data pulled from HubSpot)?
- Find the variable name
- In your workflow, click on each node
- Look at the output section—it shows what variable names are available
- For a "New Meeting" trigger, you might see variables like
meeting_transcript,meeting_title,attendees, etc. - Add the variable to your prompt
- In your "Run Prompt" node, write your instructions
- When you want to reference data, hit the curly brackets key twice:
{{ - A menu will appear showing available variables
- Select the variable you need
- The variable will be inserted into your prompt like:
{{meeting_transcript}} - Double-check all your prompts
- Go through every prompt node in your workflow
- Make sure each one includes the variables it needs
- If a prompt is supposed to analyze data, that data should be referenced with variables
Red flag: If your workflow produces an email that just says "I don't have the information to process this" or similar empty output, you're missing variables in your prompts.
Workflow Won't Save or Shows Errors
Problem: You see error messages when you try to save, or your workflow won't save at all.
What to check:
- Look at the error message
- AskElephant will tell you what's wrong
- Common issues: a node isn't fully set up, a connection is missing, or required fields are blank
- Read the error carefully—it usually points to the exact problem
- Use "Fix with AI"
- Click the "Fix with AI" button (usually near the error message)
- AskElephant will analyze the problem and attempt to fix it automatically
- Review the changes it made to make sure they make sense
- Check each node
- Click on each node individually
- Make sure all required fields are filled in
- For prompt nodes, make sure you've written actual instructions
- For integration nodes (sending to HubSpot, Slack, etc.), make sure you've selected the correct destination
- Verify connections are set up
- If your workflow sends to Slack, make sure Slack is integrated at both the workspace and personal level
- If it sends to HubSpot or Salesforce, verify those integrations exist
- If it uses a web search or accesses your CRM, make sure those tools are enabled
- Save after each major change
- Don't make 20 changes and then save once
- Save after setting up each node or making significant changes
- This helps you pinpoint which change caused an error
Workflow Runs But Produces Empty or Incorrect Output
Problem: Your workflow completes without errors, but the results are blank, incomplete, or wrong.
Step 1: Check for missing variables (see section above)
- This is the #1 cause of empty output
- Review all your prompt nodes and verify they include the variables they need
Step 2: Verify your trigger is working
- Make sure your workflow is actually being triggered
- For a "New Meeting" trigger: record a test meeting and confirm the workflow runs
- For a "Deal Updated" trigger: go into HubSpot and update a test deal to trigger it
- For a signal trigger: make sure the signal is set up correctly
Step 3: Check your data exists
- If your workflow pulls information from HubSpot or Salesforce, make sure that data actually exists there
- The AI can't create data from nothing—it can only pull what's already in your systems
- Example: If your prompt asks "What company is this from?" but the contact doesn't have a company assigned, the AI will say it doesn't have that information
Step 4: Review your prompt instructions
- Is your prompt clear and specific?
- Are you asking for something realistic?
- Vague instructions produce vague results
- Example of vague: "Tell me about this deal"
- Example of specific: "Summarize the deal size, expected close date, and next steps from our last meeting with this customer"
Step 5: Check your workflow history
- Click "Workflow History" in the top right
- Look at recent runs and check their status
- If a run shows "Failed," click on it to see which node failed and why
- If a run shows "Completed" but output looks wrong, the issue is likely with your prompt (not the workflow itself)
Workflow Runs Partially Then Stops
Problem: Your workflow starts running but stops partway through and shows a "Failed" status.
How to diagnose:
- Go to Workflow History
- Click the "Workflow History" button
- Find the failed run
- Click on it to expand the details
- Identify which node failed
- You'll see a list of all nodes in that run
- Look for the one marked as "Failed" or showing an error
- Click on that node to see the specific error message
- Common failure points:
- You're sending a message to Slack but didn't select which channel
- You're creating a HubSpot record but didn't fill in required fields
- Fix: Open the failed node and double-check all settings
- Your prompt asks the AI to do something, but the data isn't available
- Fix: Make sure you're passing the right variables to the prompt
- Use the error message
- AskElephant will tell you what went wrong
- Read it carefully—it usually explains what to fix
Incorrect node configuration:
AI can't access required data:
Your Workflow Output Isn't Going Where You Expected
Problem: Your workflow completes but the output isn't in Slack, the email doesn't arrive, or the HubSpot field isn't updating.
Step 1: Verify the output node is set up correctly
For Email:
- Did you specify which email address to send to?
- Is it spelled correctly?
- Does that email account exist?
For Slack:
- Did you select the correct channel?
- Did you make sure AskElephant is added to that channel?
- (AskElephant won't send to channels it's not a member of)
- Is the Slack integration actually connected?
For HubSpot or Salesforce:
- Did you select which object type (Deal, Contact, Company, etc.)?
- Did you select which record to update?
- Do you have permission to update that type of record?
- Ask your workspace owner or CRM admin to verify permissions
For Conversation:
- The output should appear in AskElephant chat
- Check the conversation you specified
- Refresh the page if you don't see it
Step 2: Check your workflow history
- Did the output node actually complete?
- Or did the workflow stop before reaching the output node?
- Look at "Workflow History" to confirm it reached the final step
Step 3: Verify integrations
- Go to Workspace Settings and check integrations
- Go to your Preferences and check personal connected accounts
- Make sure the service you're sending to is connected
Performance Issues: Workflow Is Too Slow
Problem: Your workflow takes a long time to run or seems stuck.
Normal wait times:
- Most workflows complete in 30 seconds to 2 minutes
- If your workflow processes lots of data (like 50+ meetings), it may take longer
- AI processing itself takes a few seconds per step
If it's taking longer than expected:
- Check if it's actually running
- Go to Workflow History
- Is there a run marked as "In Progress"?
- If yes, it's still working—give it a few more minutes
- Simplify your workflow
- Do you have multiple prompt nodes doing similar analysis?
- Can you combine them into one?
- Fewer nodes = faster execution
- Optimize your prompts
- Long, complex prompts take longer to process
- Break down complex analysis into multiple simpler prompts
- Be specific (vague prompts require the AI to think longer)
- Reduce data volume
- Are you processing 100 meetings when you could filter to 10?
- Use triggers and filters to process only the data you need
- Smaller datasets = faster processing
Testing Your Workflow: Best Practices
Before you run your workflow on real, important data, test it thoroughly.
Step 1: Choose a low-stakes test case
- Don't test on your most important client meeting or deal
- Use a practice recording, old meeting, or test data
- If something goes wrong, it won't affect real work
Step 2: Run the full workflow
- Click "Test"
- Select your test data (meeting, deal, etc.)
- Check the output carefully
- Look at Workflow History to make sure all nodes completed
Step 3: Verify the output
- Did the email send to the right address?
- Does the summary in HubSpot look correct?
- Is the Slack message formatted properly?
- Check every aspect of the output
Step 4: Make adjustments
- If something isn't right, edit the workflow
- Save your changes
- Test again
- Repeat until you're happy with the results
Step 5: Activate it
- Once testing is complete, turn on the workflow for real use
- Monitor it for the first few runs to make sure it behaves correctly
Workflow Setup Mistakes to Avoid
Mistake 1: Not understanding your trigger
- Before building anything, write down: "This workflow should start when [specific thing happens]"
- Make sure your trigger matches that description
- Test it to confirm it triggers at the right time
Mistake 2: Forgetting to consider your output destination
- Before building, know exactly where the output should go
- Make sure that destination (email, Slack, CRM) is set up and integrated
- Don't build a workflow that sends to a tool you haven't integrated yet
Mistake 3: Assuming the AI knows what you mean
- AI needs clear, specific instructions
- "Summarize the meeting" is vague
- "Create a 3-sentence summary of the customer's main pain points discussed in this meeting" is clear
- Vague prompts = vague results
Mistake 4: Not including required variables in prompts
- Every prompt node should reference the data it's supposed to analyze
- Use variables to pass data from node to node
- Without variables, the AI has no context
When to Reach Out to Support
You've tried everything above, but your workflow still isn't working. Time to get help.
Contact support:
- Email: support@askelephant.ai
- Or use the chat bubble in the bottom right corner of AskElephant
Include in your message:
- The name of your workflow
- A link to the workflow (copy the URL from your browser)
- What you're trying to accomplish
- What you've already tried to fix it
- Any error messages (screenshot them if possible)
- When the issue started (was it working before?)
What support can do:
- Review your workflow setup
- Identify bugs or issues on our end
- Help you reconfigure nodes correctly
- Verify your integrations are set up properly
- Suggest better approaches to what you're trying to build
Quick Troubleshooting Checklist
Before you dig deep, run through this checklist:
Quick Recap
Workflow troubleshooting usually comes down to:
- Variables: Make sure prompts include the data they need to analyze
- Setup: Verify all nodes are configured correctly and integrations are connected
- Testing: Always test on safe, practice data before going live
- History: Use Workflow History to see exactly where things went wrong
- Specificity: Give clear, detailed instructions to AI; vague prompts produce vague results
Most workflow issues are fixable with these tools. If you're still stuck, support is ready to help.
Need Additional Help?
If you have questions or need further assistance, the AskElephant support team is here to help!
You can reach our support team in several ways:
- click the chat button in the bottom right corner of your screen,
- email us at support@askelephant.ai
- or use @askelephant support in your dedicated Slack channel.
We're committed to getting you the answers you need as quickly as possible.
