ChatGPT + HubSpot: spot the issue behind the tickets

When to use it

A service lead has two questions: which customers need attention now, and which recurring problem should be fixed so fewer customers need to write in. The prompts work from the ticket data the connector can retrieve. Engagement access can be limited, for example when Sensitive Data is enabled in the account, so both prompts ask the model to state its scope.

Prompt 1: group recurring issues and list open priorities

Define the pipeline and the period. Grouping by category and description gives the service lead a concrete question: is an onboarding step unclear, is something broken, or are customers arriving with the wrong expectations? Each answer leads somewhere different: a help article, an onboarding change or a product fix.

Using the HubSpot connector, review tickets created in [SUPPORT PIPELINE] between [START DATE] and [END DATE]. Group recurring issues using the available ticket categories and descriptions. For each group show the count, two or three representative ticket links, and how many tickets in the group remain open. Separately, list up to five currently open high-priority tickets in that pipeline, including older ones, with owner and age in days. State how many records you reviewed, which fields were missing, and any retrieval limits. If ticket conversations or engagements are not accessible, say so and work from ticket properties only. Do not reply to any ticket or update any record.

Prompt 2: prepare a handover for one ticket

This helps the next person understand where things stand without asking the customer to explain everything again. Check the sources before assuming the summary includes every customer message.

For ticket [TICKET ID OR SUBJECT], summarize the reported problem, the documented troubleshooting so far, the current status, and the questions that remain unanswered, using only the HubSpot information you can access. Draft an internal handover note with the next action to consider and why. Link to the sources for each point. Do not invent a resolution, do not draft a reply to the customer, and do not change the ticket.

How to use it

1. 1Run prompt 1 with your support pipeline name and a period of two to four weeks. Open one example ticket per group to confirm the grouping makes sense.
2. 2Pick the recurring issue with the most open tickets and decide who investigates the root cause.
3. 3Run prompt 2 for one open high-priority ticket and hand the note to the owner.
4. 4Share recurring pre-sale confusions with marketing; they belong in the content you create next.

Tips

• Label the scope. If the connector cannot see conversations, the summary is built on properties, and the reader should know that.
• Do not let the model resolve anything. Its job is to organise what is recorded so a person can act.
• Run the grouping prompt on a schedule, monthly for instance, and compare the groups over time.

Related resources