the email warmup schedule your agent is probably skipping

Your agent just got a fresh email address on a brand-new domain. It has 300 contacts to reach. So it starts sending. By the afternoon, half those emails are sitting in spam folders. By tomorrow, the domain's reputation is cooked.

This is the most common failure mode for AI agents that send email. Not a code bug. Not a configuration mistake. Just sending too much, too fast, from an address that nobody has seen before.

Gmail, Outlook, and Yahoo don't trust new domains. Trust is earned through a process called domain warmup, and it follows a specific schedule. Skip the schedule, your agent's emails vanish. If you'd rather not manage this manually, LobsterMail handles pacing for new inboxes automatically. But understanding how warmup works matters regardless of what infrastructure you use, because the rules come from receiving servers, not your sending platform.

what email warmup actually does#

When you register a new domain and start sending email from it, inbox providers have zero data on you. No sending history. No engagement patterns. No reputation score. You're a blank slate, and in email terms that means suspicious until proven otherwise.

Warmup is the process of gradually increasing your sending volume over several weeks while keeping engagement healthy: opens, replies, low bounces. This builds what the industry calls sender reputation, a score that Gmail, Microsoft, and other providers calculate behind the scenes to decide whether your messages hit the inbox or the spam folder.

The score factors in:

• Volume consistency (sudden spikes are red flags)
• Bounce rates (high bounces signal bad list hygiene)
• Spam complaint rates from recipients
• Authentication records like SPF, DKIM, and DMARC
• Engagement signals (opening, replying, marking as not-spam)

For agents, engagement is where things get tricky. An agent sending transactional notifications or automated outreach doesn't always generate the same open rates as a personal email between two humans. That makes the warmup period even more important, because you have less margin for error on the signals you can control.

a realistic warmup schedule for 2026#

Based on current provider behavior and data from Mailreach, Mailivery, and SMTP.com, here's what a healthy ramp looks like for a new sending domain:

A few things worth calling out about this timeline.

Week one is entirely about establishing a baseline. Send to people who will actually open and reply. If your agent handles inbound requests, replying to real incoming messages is the fastest reputation builder. Each genuine reply tells Gmail and Microsoft that this domain participates in real conversations, not just blasts.

The cold outreach question comes up constantly. Mailivery's data shows that light cold sending (5–10 per day) can start around day 15, and by week four you can match cold email volume to warmup volume at a 1:1 ratio. Pushing cold sends before that two-week mark is gambling with your domain.

One more thing: high-volume senders targeting 500+ emails per day should plan for 6–8 weeks of gradual ramp rather than four. The table above gets you to moderate volume. Aggressive scaling takes more patience, not less.

why agents blow through this schedule#

Humans intuitively understand pacing. You wouldn't send 500 handwritten letters on your first day at a new address. Agents don't share that instinct. They optimize for throughput, and throughput without pacing is how you burn a domain in 48 hours.

The most common pattern: an agent gets email credentials and immediately starts draining its task queue. If that queue holds 200 outbound messages, the agent sends all 200. From a domain that has existed for 12 hours. Every major inbox provider flags it within the same afternoon.

Most agent frameworks also don't surface bounce rates or spam complaints back to the agent in real time. The agent sees 200 successful API responses and assumes everything worked. But "accepted for delivery" and "delivered to inbox" are very different outcomes. We covered several of these failure modes in our post on agent email setup mistakes that tank deliverability.

Then there's the burst pattern. Agents tend to process work in batches: fire off 50 emails in three minutes, go idle for an hour, fire off another 50. Inbox providers notice this. It looks mechanical (because it is), and mechanical patterns trigger extra scrutiny on domains that haven't built any trust yet.

Finally, agents pulling contacts from a CRM or database will send to whatever addresses they find, including outdated ones that hard-bounce. No list validation, no hygiene check. Just raw throughput into a deliverability wall.

building warmup into your agent's workflow#

The fix isn't complicated, but it requires intentional design rather than letting the agent wing it.

Start with hard daily send caps that increase on a schedule. Match the limits to the warmup table above and enforce them at the infrastructure level, not through agent self-discipline. If the agent physically cannot exceed 20 sends on day one, it won't.

LobsterMail's tiered structure does this by design. Free accounts can only receive email, which gives the domain time to exist and accumulate legitimate inbound mail before any outbound traffic starts. When you move to the Builder plan ($9/month), you get up to 500 emails per day with rate controls built in. The domain has already been warming passively through inbound activity by the time outbound sending begins.

Make replies the priority. If your agent handles both inbound and outbound email, it should always process replies before starting new conversations. Replying to real threads builds reputation much faster than cold outreach, so structure the task queue to put messages with an inReplyTo header at the front of the line.

Build an automatic feedback loop around bounce data. After each batch of sends, check delivery status. If bounces spike above 2–3%, the agent should reduce its own volume and flag the issue for human review. This kind of self-regulation during warmup is the difference between a healthy domain and one that needs months of rehabilitation.

Validate every email address before the agent sends to it. Run your contact list through a verification service ahead of any outreach. Removing invalid addresses before they bounce is the highest-impact deliverability move you can make in month one.

Space sends throughout the day instead of firing them in bursts. Add random delays between 30 seconds and 5 minutes between individual messages. This looks like human sending behavior, which is exactly what inbox providers reward from new domains. If your agent hits a 550 denied by policy error during the warmup period, you've already pushed too hard.

four weeks of patience saves months of recovery#

A burned domain takes weeks to rehabilitate, if it recovers at all. In many cases, registering a fresh domain and restarting the warmup process from scratch is faster than trying to dig out of a spam blacklist from Gmail or Microsoft.

For agents that depend on email for their core function (client communication, outreach, support tickets), a blacklisted domain means the agent is effectively offline. It can send messages all day. Nobody will see them.

Build the warmup schedule into your agent's configuration from day one. Four weeks of controlled ramp at the start prevents months of deliverability problems later. If you want the infrastructure to enforce those limits for you, get your agent set up with LobsterMail and let the pacing happen automatically.