Steal this HubSpot workflow to automatically summarize the buyer’s next step in your deals using OpenAI and write the result back to the deal record for quick pipeline review context.
You'll also be subscribed to the Happy Customers newsletter.
HubSpot recently launched a new OpenAI app in public beta, and it includes a new workflow action that lets you send deal data to GPT-4 and use the response however you want: write to a property, post to Slack, trigger follow-ups, create a task, and more.
This post walks through how to build a workflow that generates a short summary of the buyer’s next step in the sales process based on recent deal activity. You can use this summary in pipeline reviews to reduce time spent scanning timelines and guessing about next steps.
This setup takes about 20 minutes, requires no process changes, and works surprisingly well right out of the box.
💡 Note: HubSpot’s OpenAI app is currently in public beta and only available only on Enterprise tiers (Sales, Service, Marketing, or Ops).
→ In your HubSpot account, go to Product Updates and search for “OpenAI.”
→ Enroll your portal in the public beta if you haven’t already.
You can read HubSpot's knowledge base article for deeper instructions on the OpenAI setup.
Set the enrollment trigger to:
This ensures the workflow runs only on open deals with recent activity.
💡 Note: make sure re-enrollment for this workflow is ON so that your deals can re-enroll any time a new activity happens!
Here's the exact prompt we used:
Read the attached sales activity. Summarize the buyer’s current step in the deal in under 10 words, using a clear sentence fragment (not a full sentence).
No filler. No background. Just the current status.
Use the buyer’s name if known.
Examples:
“Sarah testing with team”
“Internal review after demo”
“Waiting on legal”
“Budget approval in progress”
Do not say “the prospect.”
Do not include past steps - only what’s happening now.
Do not use full sentences. Just direct fragments.
→ Inside the OpenAi workflow action, click into the prompt editor, and inside the right sidebar select "Edit available data"
→ Click on "Add records"
→ Name your data source. In this example, I created two and named them "Most recent note" and "Most recent email"
→ For the most recent note example, in the "Choose a record type" dropdown, select "Note"
→ In the "Choose a condition to filter notes" dropdown, select "If associated to the enrolled deal"
→ In the "Choose a condition to filter notes down to one" dropdown, select "Most recently updated"
→ Do the same for any other data sources you want to reference (Emails in this example)
→ Go back to the "Choose data variable" sidebar within the OpenAI prompt editor, and under the "View properties or action outputs from" dropdown, select the "Most recent note" source we just created
→ Click on "Text properties" → "Note body"
→ Again, follow the same steps for adding the "Most recent email" source.
→ Make sure both properties are added to your prompt!
Since this is a fairly simple prompt with a short and concise output, these are the advanced settings we used:
Model selection: gpt-4o
Reasoning effort: Low
Temperature: 0.0
→ Add a new step in your workflow with the "Edit record" action
→ Select your "Buyer next step" property to edit
→ When it asks you to "Choose a value", click into the "View properties or action outputs from" dropdown and select "Action outputs"
→ Select "Response" and this will include the output of your AI response as the property value
Just review and publish your workflow, and the "Buyer next step" property value should automatically update any time a new activity happens on the deal!
Not sure where to start? Download the 7 playbooks you need to scale across the customer journey.
You'll also be subscribed to the Happy Customers newsletter.