Four settings decide whether HubSpot's Prospecting Agent books meetings or quietly burns credits: the outreach mode, the review setting, the email count, and which plays you launch first. We configure this agent for clients every week, and the same defaults keep winning: Adaptive outreach, five emails, send without review behind real guardrails, and all three core plays launched together. Here is why, decision by decision, as of November 2026.
Should the Prospecting Agent use Adaptive outreach or Sequences?
Adaptive, almost every time. In Sequences mode the agent drops prospects into pre-written sequences, which means you are paying AI credits to do what a free workflow already does. Adaptive mode is the actual product: the agent researches each prospect, writes each email from that research, and adjusts timing and messaging based on how the prospect responds.
The emails are visibly different. Sequence sends read like templates because they are templates. Adaptive sends reference the prospect's company, role, and the signal that enrolled them, which is what earns replies in 2026 inboxes. If you want templated sends, use a sequence and skip the agent. If you bought the agent, let it do the thing you bought it for. We walk through this trade-off in more depth in our comparison of AI prospecting agents.
Is it safe to let the agent send without review?
Yes, and it is the setting that determines whether the agent actually runs. Teams that keep every email in a review queue stop reviewing within two weeks, the queue backs up, and the agent is effectively off. The agent only compounds when it sends on its own.
Send without review is safe when the guardrails underneath it are real:
- Enrollment rules that exclude the wrong people. Customers, open deals, active support tickets, competitors, and anyone your team is already working. The biggest failure mode is not a bad sentence, it is a good email to the wrong contact.
- A clean sending identity. A dedicated sender with proper authentication, warmed up, with sending volume capped.
- Clear product context. The agent writes from the AI context you give it: brand voice, what you sell, who it is for, and what you never claim. Thin context produces generic emails. That is a setup problem, not an AI problem.
- A weekly audit habit. Spot-check sent emails once a week instead of gating every send. You catch drift without becoming the bottleneck.
Review mode is fine for the first few days while you confirm the context holds up. Treat it as training wheels, not a destination.
How many emails should the agent send per prospect?
Five. One or two touches is not enough for cold outbound, and the long eight-plus touch cadences that worked in 2021 now read as harassment and hurt deliverability. Five gives the agent room for an opener, two follow-ups that add something new, a different angle, and a clean close. Replies and meetings come overwhelmingly from the first five touches anyway, so pushing past that spends credits on diminishing returns.
Which plays should you launch first?
Launch the three core plays together rather than easing in with one:
- Inbound follow-up. The agent works your form fills, content downloads, and pricing-page visitors within minutes. This is the highest-intent, lowest-risk play and it usually pays for the agent by itself.
- Re-engagement. Closed-lost deals and gone-quiet prospects from the last year. The research is already in your CRM, and these contacts already know you.
- Targeted outbound. Net-new prospects matching your ideal customer profile. This is the slowest play to ramp, which is exactly why it should start on day one instead of month three.
Launching all three matters because they ramp at different speeds. Inbound follow-up produces meetings in week one while outbound builds. Teams that start with outbound alone judge the agent on its hardest job and turn it off before the easy wins ever run.
What does this cost to run?
The agent itself is driven by HubSpot credits, and the real number depends on your enrollment volume and research depth. We broke the math down in our Prospecting Agent cost guide. One framing note: do not sign up for HubSpot just to trial this on a free portal. If you are buying HubSpot or HubSpot Credits anyway, buy through a Solutions Partner and get the setup done free as part of the purchase, then treat the weeks after as your testing period. That is the service behind our Breeze agent setup and The Prospector builds, and the fastest way to see your own numbers is the free personalized Breeze starter pack we generate for your company.
Frequently asked questions
Should the HubSpot Prospecting Agent use Adaptive outreach or Sequences?
Adaptive. Sequences mode spends AI credits enrolling prospects into templated sequences a workflow could run for free. Adaptive mode researches each prospect and writes each email individually, which is what the agent is for.
Is it safe to let the Prospecting Agent send without review?
Yes, once your guardrails are real: enrollment rules that exclude customers and open deals, a dedicated authenticated sender, solid AI context, and a weekly spot-check of sent emails. Permanent review queues stall within weeks and the agent stops running.
How many emails should the Prospecting Agent send per prospect?
Five. It is enough touches to earn a reply without the deliverability and brand damage of long cadences, and nearly all replies come inside the first five touches.
Which Prospecting Agent plays should you turn on first?
All three core plays together: inbound follow-up, re-engagement, and targeted outbound. They ramp at different speeds, so launching them together means inbound meetings land in week one while outbound builds.
Want to watch these exact settings get configured live? We are building a HubSpot AI Prospecting Agent from scratch, live, this Wednesday, November 18 at 12 PM CT. It is free, and you will leave with the setup checklist we use on client builds. Save your seat for the live build.