How do we run outbound AI calling campaigns without annoying leads or breaking the rules?
An outbound AI calling playbook: consent and disclosure, calling windows, retries and voicemail, do-not-call checks from the CRM, and writing every outcome back to HubSpot or GoHighLevel.
- Time needed
- 1 to 2 days
- Steps
- 6
- Level
- Intermediate
In short: Outbound AI calling works when the people on the list expect to hear from you, such as leads who filled in a form, customers with an appointment or renewal, or past clients who opted in. Before the first call, set calling windows by the contact's time zone, a retry and voicemail policy, a do-not-call check against the CRM on every dial, a clear disclosure that the caller is an AI agent, and a write-back of every outcome to HubSpot or GoHighLevel. Do not use it for cold lists you have no consent to call. This is general guidance, not legal advice, so check the rules where your contacts live.
When outbound AI calling fits, and when it does not
| Good fit | Poor fit |
|---|---|
| Speed-to-lead calls to people who just requested contact | Cold lists bought or scraped without consent |
| Appointment confirmations and reminders | Complex negotiations or complaints |
| Renewal and re-engagement calls to opted-in customers | Calls where the person cannot reach a human |
| Qualifying inbound leads before a human call | Regions or industries where you have not checked the rules |
Platforms now ship campaign tools for this. ElevenLabs launched ElevenAgents Campaign Manager on November 9, 2026: the agent calls during the hours you set, carries context between attempts, retries after no answer or voicemail, and stops when the outcome is reached or the person asks to stop (ElevenLabs). The rules below apply whichever platform you use.
Step 1: Consent, disclosure and the law
- Know your basis for calling. In the US, the FCC confirmed in February 2024 that AI-generated voices count as "artificial" voices under the Telephone Consumer Protection Act, so the consent rules for artificial or prerecorded voice calls apply (FCC 24-17).
- Store consent in the CRM. Record when and how each contact agreed to be contacted, and only enrol contacts with a valid consent value.
- Disclose at the start. The agent says who it is calling for and that it is an automated assistant, and offers a human.
- Honour opt-outs immediately. "Stop calling me" ends the call, marks the contact and removes them from every campaign.
Step 2: Calling windows and pacing
| Setting | Recommended starting point |
|---|---|
| Time zone | Use the contact's time zone from the CRM, not yours |
| Window | Business hours on weekdays in the contact's time zone, then narrow to the hours that work |
| Concurrency | Start low and raise it once hand-offs keep up |
| Holidays | Skip public holidays in the contact's region |
Step 3: Retries and voicemail
- Set a maximum number of attempts per contact and a minimum gap between them.
- Vary the time of day for retries.
- Decide whether to leave a voicemail and keep it short: who is calling, why, and how to reach a person.
- After the last attempt, hand the contact to a human task or an email follow-up instead of calling again.
Step 4: Do-not-call checks from the CRM
- Check a do-not-call property in HubSpot or a DND flag in GoHighLevel immediately before each dial, not only when the list is built.
- Include company-wide suppression lists and any national registry your rules require.
- Write every opt-out back to the CRM within the call, so other tools see it too.
Step 5: Write outcomes back to the CRM
| Outcome | CRM update |
|---|---|
| Booked | Meeting created, owner notified, lifecycle stage updated |
| Interested, call back | Task for the owner with the time the person asked for |
| Not interested | Reason recorded, removed from the campaign |
| Opt-out | Do-not-call set, removed from all campaigns |
| Wrong number | Number marked invalid |
| No answer after final attempt | Hand-off task or email sequence |
Store the transcript or summary and the recording link on the contact. See our voice agent production playbook for logging and monitoring, and HubSpot services for CRM setup.
Step 6: Hand-offs and latency
Give the agent a live transfer to a person during business hours and a booked callback outside them. Slow responses make calls feel robotic, so test response time before launch; our note on voice agent latency covers where delay comes from.
Example flow: real estate enquiry follow-up
This is an illustrative flow, not a client project. A buyer submits an enquiry on a listing. Within minutes, the agent calls in the buyer's calling window, confirms the property, asks budget and timing questions, and offers viewing slots from the agent's calendar. A booked viewing goes to the listing agent with the summary; "just browsing" gets an email with similar listings; an opt-out is recorded at once. See construction and real estate automation.
Launch checklist
- Consent field required for enrolment
- Disclosure and opt-out lines tested
- Calling windows set by contact time zone
- Retry limit and voicemail script approved
- Do-not-call check runs before each dial
- Every outcome writes back to the CRM
- Human transfer works during business hours
- A person reviews a sample of calls every day for the first two weeks
FAQ
Is it legal to call leads with an AI voice agent?
It depends on where your contacts are and what consent you hold. In the US, AI-generated voices are treated as artificial voices under the TCPA. Get legal advice for your situation before launching.
How many times should an AI agent retry a contact?
Set a small maximum, space attempts out and vary the time of day. After the last attempt, switch to a human task or email instead of more calls.
Should the agent say it is an AI?
Yes. Disclose at the start and offer a human. It builds trust and is required in some places.
Can outbound AI calls update HubSpot or GoHighLevel?
Yes. Write the outcome, summary, recording link and any opt-out to the contact after every call, and create tasks or meetings for the owner.
Do we need a separate inbound receptionist?
Inbound and outbound are different jobs. If you also want calls answered, see AI receptionist for small business.
How Octacer helps
Octacer builds voice agents on platforms such as ElevenLabs and Retell AI, with CRM write-back, do-not-call checks and human hand-offs. We start with one campaign and a review loop before scaling.
Next step
Choose one list where every contact asked to hear from you, and write down the outcome you want from each call. Then book a call and we will scope the campaign.
Sources
- ElevenLabs, ElevenAgents Campaign Manager works every contact to an outcome (November 9, 2026): https://elevenlabs.io/blog/elevenagents-campaign-manager
- FCC, Declaratory Ruling FCC 24-17 on AI-generated voices under the TCPA (February 2024): https://docs.fcc.gov/public/attachments/FCC-24-17A1.pdf
Keep going
More playbooks
Want it running on your stack?
We implement this playbook for clients, tailored to your stack, your data, and your team's workflow, and handed over documented.