Every deliverability consultant is selling the same fix right now: more domains, more mailboxes, longer warmup, better spintax. Rotate faster than the filters can catch you.
That's an arms race against the two companies with the best machine-learning teams on the planet, fought on their infrastructure, by vendors reselling $3/month mailboxes. You will lose. Everyone running that playbook eventually loses. The only question is how much pipeline you burn before you do.
There's a category of B2B email that doesn't fight the filters at all. It gets opened, read, and replied to at rates that look fake next to cold benchmarks. It's the intro email — sent by someone the recipient already knows. This piece explains the mechanics of why that's true, and what a deliverability strategy looks like when you build around it instead of against the filter.
The 2026 filtering environment: engagement is the algorithm
Two things changed the game, and most outbound teams still operate as if neither happened.
First, the bulk-sender rules. In February 2024, Google and Yahoo started enforcing hard requirements for high-volume senders: SPF, DKIM, and DMARC authentication, one-click unsubscribe, and — the killer — a spam-complaint ceiling of around 0.3%. Cross it and your domain doesn't get warnings. It gets silently routed to spam, across all recipients, for weeks.
Second, and more important: mailbox providers moved decisively to engagement-based filtering. The strongest inputs into inbox placement are now behavioral. Does this recipient open mail from this sender? Reply to it? Move it out of spam? Add the sender to contacts? Domain reputation still matters, but recipient-level engagement history is the trump card. A sender the recipient has replied to before effectively cannot land in spam. A sender the recipient has never engaged with is one classifier decision away from oblivion.
Now add volume. Gartner predicts that by 2028, AI agents will outnumber human sellers 10-to-1 — and, in the same breath, that fewer than 40% of sellers will report AI improved their productivity (Gartner, Nov 2025). Every AI SDR platform's pitch is "10x the touches per rep." All of that lands in the same inboxes, and buyers respond to the noise the only way they can: 73% of B2B buyers say they actively avoid suppliers who send irrelevant outreach (Gartner, Mar 2026). Avoidance at the buyer level becomes spam complaints and non-engagement at the filter level. Non-engagement becomes suppression. The AI SDR wave isn't just annoying buyers — it's training the filters to kill the entire cold channel.
Cold reply rates have collapsed from roughly 8.5% in 2019 to under 2% today, and in saturated categories like cybersecurity they're approaching 0.5%. That's not a copywriting problem. It's an infrastructure-level rejection of the channel.
The cold-infrastructure treadmill
Look at what a "modern" cold outbound stack actually is in 2026: 20-40 lookalike domains, 100+ warmed mailboxes, warmup networks that fake engagement between bots, spintax to defeat content fingerprinting, rotating IPs, deliverability monitoring dashboards. An entire industry exists to make unsolicited email look solicited.
Read that sentence again. The whole stack is a machine for impersonating the signals of a wanted relationship. Warmup tools fake opens and replies. Spintax fakes human variance. Lookalike domains fake being your company without risking your company's domain — which tells you the operators already know the mail is radioactive.
The treadmill has a cost curve that only goes up: domains burn out in 60-90 days, warmup periods stretch, complaint ceilings tighten, and the filters improve every quarter because they're trained on the entire ecosystem's behavior at once. Meanwhile buyers built their own filter on top: 67% of B2B buyers now prefer an entirely rep-free buying experience (Gartner, Mar 2026). You're paying more every quarter to reach people who have both algorithmic and psychological defenses against you.
Why the warm-intro email is structurally immune
Now trace what happens when a warm intro moves through the same filtering stack.
Your customer champion — someone the buyer worked with for three years — forwards your note to the buyer, or sends a fresh two-paragraph email from their own mailbox. Walk through the filter's checklist:
- Sender reputation: a real person, on a real corporate domain, with years of legitimate one-to-one sending history. No warmup network required — the mailbox is genuinely warm.
- Engagement history: the buyer has opened and replied to this sender before. Recipient-level engagement history is the strongest single input into placement, and the connector already has it.
- Volume pattern: one email, to one person. No burst pattern, no fingerprint shared across a thousand sends, nothing for a bulk classifier to match against.
- Content: written or lightly edited by a human, referencing real shared context. No spintax artifacts, no template skeleton the filters have seen ten thousand times.
- Recipient behavior: the buyer opens it — because the sender's name is the notification — reads it, and usually replies. Every one of those actions feeds back as positive reputation.
The warm intro doesn't beat the spam filter. It never enters the contest. Filters exist to predict "does this recipient want this email?" — and for a warm intro the true answer is usually yes, so every signal the filter reads agrees. This is why intro-based outreach converts to meetings at 3-5x the rate of cold across Boomerang's customer base: deliverability isn't a tax on the motion, it's a property of it.
And there's a second-order effect nobody prices in: the reply. Because warm intros get answered, your domain accumulates genuine positive engagement history with your target accounts. Every warm thread makes your next email to that company more deliverable. Cold volume compounds negative reputation; warm volume compounds positive reputation. Same channel, opposite flywheels.
The forwardable email: deliverability by design
The highest-deliverability artifact in B2B sales is the forwardable email — the note you write for your connector to pass along. It works because it inherits everything the filter trusts:
- It travels inside an existing thread. A forward from a known sender arrives with the full weight of that relationship's engagement history.
- The connector's voice wraps it. The buyer's first read is their trusted contact's framing, not your pitch.
- The favor-cost is one click. Your connector doesn't have to compose anything, which is why the intro actually gets sent instead of dying in their drafts. Most warm intros die not because anyone said no, but because the connector didn't know what to say.
This is exactly the mechanic Boomerang's agent, Rudy, automates: when a signal fires, it finds the strongest path, drafts the forwardable in the connector's voice, and routes it for a one-click approval. Armis ran this motion across 26,000 warm-intro paths and booked a 10x ROI in year one — zero burner domains involved.
The deliverability playbook for warm-led outbound
A warm-led motion still needs email hygiene. The difference is you're maintaining one real domain's reputation, not resuscitating forty fake ones.
Send from real people on your primary domain. No lookalikes. The volume in a warm-led motion is low enough that your actual domain is safe — and every reply strengthens it.
Let the connector send the first touch. The highest-trust mailbox for any given buyer belongs to the super connector who knows them. Your job is to make that send effortless, not to replace it with your own.
Keep authentication clean anyway. SPF, DKIM, DMARC on your primary domain. Table stakes — warm mail from a misconfigured domain still looks suspicious.
Cap follow-up cadence at human scale. Two or three thoughtful follow-ups on a live thread. The moment your sending pattern looks automated, you're re-entering the contest you opted out of.
Strip the surveillance. Open-tracking pixels and per-recipient tracking links are classifier features for bulk mail. On one-to-one warm mail they add risk and no information you can't get from the reply itself.
Measure replies, not opens. The metric that matters for a warm-led motion is warm path velocity — paths surfaced, conversion to meetings, pipeline per rep per week. Open rates were always a vanity metric; in a warm motion they're irrelevant.
Reinvest the savings. The fully loaded cost of a cold stack — domains, mailboxes, warmup tools, monitoring, and the SDR hours babysitting it — typically runs tens of thousands per year for a mid-size team. A warm-led motion redirects that spend into mapping the buying committee and instrumenting the relationship graph, where it compounds instead of depreciating.
What this means for 2026 planning
The strategic read is simple. Engagement-based filtering plus AI-SDR volume is a one-way ratchet: the cold channel gets structurally worse every quarter, and no infrastructure spend reverses it, because the filters are trained on the aggregate behavior of everyone gaming them. The warm channel gets structurally better every quarter, because every intro that lands adds engagement history, and every new customer, hire, and investor adds paths to the graph.
Teams still budgeting for domain farms in 2026 are buying a depreciating asset. Teams instrumenting their relationship graph are buying an appreciating one. Champion tracking alone — knowing when a past buyer lands at a target account — produces intro paths whose deliverability is effectively 100%, because the "sender" is a relationship, not a mailbox.
Outbound isn't dead. Unwanted outbound is dead, and the filters are just the enforcement mechanism for what buyers already decided.
Frequently asked questions
Do warm-intro emails ever land in spam? Rarely, and almost always for mechanical reasons — a misconfigured domain on the sender's side, or an attachment-heavy first send. The relationship signals that drive engagement-based filtering overwhelmingly favor a sender the recipient has previously engaged with. Fix authentication basics and the rate rounds to zero.
Does this mean we should shut down cold email entirely? Not necessarily. Low-volume, high-relevance cold email to accounts with a genuine trigger still works, especially outside saturated verticals. What's dead is volume-first cold — the 10,000-sends-a-month motion. Run cold as a precision complement on your primary domain, at volumes that never threaten your complaint ceiling.
Won't AI SDR vendors just solve deliverability with better rotation? They're trying, and it's a losing trade. Rotation defeats yesterday's filters; engagement-based filtering judges each recipient-sender pair on real behavioral history, which can't be faked at scale. Gartner already predicts over 40% of agentic AI projects will be canceled by end of 2027 on cost and unclear value — deliverability collapse is a big contributor in the SDR category.
What reply rates should a warm-led motion expect? Boomerang customers see 30-50% response on connector-routed intros and 3-5x higher meeting conversion versus cold. The variance is mostly path strength — an intro from a former colleague converts differently than one from a second-degree acquaintance — which is why scoring paths matters more than polishing copy.
How many mailboxes and domains does a warm-led team need? One domain: yours. Real mailboxes for real team members. The entire secondary-domain apparatus exists to absorb the reputational damage of unsolicited volume; a warm-led motion doesn't generate that damage, so the apparatus is dead weight.
How do we find enough connectors to sustain volume? That's the graph problem, and it's why warm-led outbound is an orchestration discipline rather than a tactic. Most companies tap less than 10% of their network. Mapping the four pillars — employees, customers, investors, partners — typically surfaces thousands of viable paths; Armis surfaced 26,000. The constraint is almost never the network. It's the instrumentation.
