Yes, for a short list of purposes: checking that a fix held, keeping a promise made in the ticket, or asking for feedback. It must respect the customer's contact preferences and a frequency cap, and it must not turn a follow-up into a sales message. Follow-ups on complaints, bereavement or vulnerability cases are a person's decision.
Which follow-ups are worth sending?
A good follow-up has a reason the customer would recognise. Three reasons cover most cases.
- Did the fix hold? After an intermittent fault, a check a few days later catches problems the customer might not bother to report.
- A promise made in the ticket. "We'll confirm once the credit appears" is a commitment, and the agent is well placed to keep it.
- Feedback. A short question about how the contact went, sent once.
What is not a follow-up: a discount offer, a new product, a "while we have you" upsell. Data protection law expects personal data to be used for the purpose it was collected for, and contact preferences may forbid marketing altogether. A follow-up that is really marketing breaks both.
What conditions decide it?
- The ticket notes say what was promised. The agent follows up on what is in the record, not on what it guesses. Good notes come from writing case notes after every contact.
- The customer's preferences allow it, on that channel.
- A cap on frequency, so one ticket does not produce a stream of messages.
- No sensitive flag on the ticket: complaint, bereavement, vulnerability, financial difficulty or a fraud case.
What may the agent do, and what stays with a person?
| Follow-up | The agent | A person |
|---|---|---|
| Check a technical fix held | Sends one check-in | None |
| Confirm a promised refund or credit arrived | Checks the record and confirms | Chases a failed credit |
| Ask for feedback | Sends one survey | Reviews negative feedback |
| Customer replies that the problem is back | Reopens the original ticket with notes | Takes it if it is the second recurrence |
| Follow-up after a complaint | Never by default | Decides whether and how |
| Follow-up after bereavement or vulnerability | Never | Decides |
| Promote a product | Never in a follow-up | Marketing, under its own rules |
What does this look like after a fault?
Take a broadband provider. A ticket about a line dropping every evening closes after the agent resets the customer's equipment remotely and the line tests clean. The case notes say: "Check in after three evenings."
On the fourth day the agent sends one message asking whether the drops have stopped. The customer replies that they happened again last night. The agent does not open a new ticket. It reopens the original one, so the history stays together, runs the line test again, and finds a fault outside the home. It books an engineer from available slots, adds the result to the notes, and tells the customer. If the customer had not replied, the agent would have sent nothing more.
Why reopen rather than start again?
Because a new ticket hides the failure. Resolution figures look better, the person who picks it up lacks the history, and the customer explains everything again. When and how the agent can bring a ticket back is covered in whether an AI agent can reopen a closed ticket, and the closing step that comes before it in whether an AI agent can close a support ticket on its own.
Follow-up replies arrive as new messages, so they should pass through ticket triage and be linked to the closed ticket rather than treated as fresh contacts.
Which controls make it safe?
- A fixed list of follow-up purposes, each tied to a ticket type.
- Contact preferences checked at the moment of sending, not when the ticket closed.
- One follow-up per purpose, with no chains.
- Complaint, bereavement, vulnerability and fraud flags blocking automatic follow-ups.
- Each message filed with the ticket and its purpose.
The broader approach to what support agents may say and when is in the guide to AI customer support agents.
Questions people ask
Can the agent follow up by phone?
If it is certified for voice and the customer's preferences allow calls. Most follow-ups work better as a short message the customer can answer when convenient.
What if the customer is annoyed by the follow-up?
Record it and stop follow-ups for that customer. Annoyance at follow-ups is useful evidence that the rules are too broad.
Should negative feedback reopen the ticket?
Not automatically. A person should read it first, because unhappy feedback may concern the process rather than an unresolved problem.