A deep read of Instantly's The New Way of Cold Emailing in 2026 (14:41), hosted by Austin Schneider. Instantly is the cold email tool category leader for unlimited-inbox outbound. This video reflects how Instantly's top users are running campaigns now that AI spam filters have rewritten the rules.
Why this analysis exists
This is one of the source layers behind the BuildOS cold-email-engagement-first-outreach skill. It supplies the 2026-specific platform layer — what filters now reward, the deliverability floor required to even reach the inbox, the under-50-recipient reply-rate math, and the two-touch rule that replaced the seven-touch sequence.
Core thesis
The old playbook (10,000 emails/month, light personalization, seven-touch cadence) was built for 2022–2023 inboxes that scanned keywords. In 2024, Gmail and Yahoo rolled out AI-driven spam filters that detect mass campaign patterns — sender behavior, engagement rates, campaign structure. Microsoft followed in 2025. Result: average cold email reply rates dropped from 1% to 5% across the industry, and inbox placement for bulk senders fell 10–27% from Q1 2024 to Q1 2025.
The replacement is engagement-first outreach. Filters and recipients now reward (in this order):
- Domain reputation — domains that act like legitimate senders, with warmed inboxes and spam complaint rates under 0.1%.
- Engagement quality — positive replies, clicks, and reads. Low engagement is a future deliverability tax.
- Relevance — buyers who can tell the email was meant for them respond. Buyers who recognize the spray-and-pray template delete.
The three-pillar system that operationalizes this: deliverability infrastructure, micro-targeted value campaigns, and the two-touch rule.
Pillar 1: Deliverability infrastructure
Gmail's 2024 requirements made this non-negotiable: SPF, DKIM, DMARC authentication on every sending domain, and a spam complaint rate under 0.1%. Skipping setup tanks deliverability; emails land in spam or never reach the inbox.
The arithmetic floor for safe sending:
- 5 sending accounts per domain.
- 30–50 emails per account per day (Schneider recommends starting at 30).
- ~250 emails per day per domain, max.
- Two weeks of automated warm-up before the inboxes are usable. Health score should hit 100% before campaigns ship.
To go beyond ~250/day, add more domains, not more emails per account. The math is intentionally conservative — overshooting the per-account ceiling is what gets a sender flagged.
The deliverability discipline Schneider names explicitly:
- Warm up before sending. Sender reputation has to be built; it cannot be skipped.
- Use inbox rotation so no single account carries the full campaign load.
- Spam complaint rate under 0.1% means the recipient list quality and the offer quality both matter — complaints aren't only about deliverability tools.
This pillar exists because nothing else in the system matters if the email doesn't reach the inbox.
Pillar 2: Micro-targeted value campaigns
The headline data point:
"Campaigns under 50 recipients get a 5.8% reply rate versus 2.1% for campaigns with over 1,000 recipients. That's almost a three times difference. Not because of personalization, but because of relevance."
The mechanism is relevance, not personalization. A 50-person list naturally tightens the persona, which tightens the offer, which tightens the language. A 10,000-CEOs-of-agencies list cannot be specific enough about anything to feel meant for the reader.
The 2026 way to keep volume and keep relevance is to slice campaigns into micro-segments by signal:
- Job title × industry × company size
- Funding stage / recent funding
- Open job listings (and the specific job)
- Industry news / triggering events
- Geography / time zone
Each segment gets its own campaign, its own opener, and its own value offer.
The AI-enrichment step is what makes it scale. Use the prospect's LinkedIn headline, LinkedIn summary, company website, and industry as input columns. Feed those into a prompt (Schneider uses Claude inside Instantly's AI column tool) with a templatized email and clear direction. The output is a personalized line per recipient that pulls from real public context — not "saw your post" fake personalization.
The other half of relevance is what's actually in the email. Schneider's clearest rule:
"Booking a call is not valuable. Sending a Loom video is not valuable in 2026. What we want to do is we want to solve the problem with an actual action."
The opener offers a real free deliverable:
- SEO agency → free Google Business Profile optimization
- Cold email agency → 100 verified leads and a sample sequence
- Branding agency → a free positioning teardown
The value has to be specific enough that the recipient can either say "yes, send me that" or self-disqualify. Loom videos and calendar links don't qualify.
Pillar 3: The two-touch rule
The most counterintuitive shift. The 7-touch / 14-touch / 20-touch sequence is dead.
The data Schneider cites:
| Touch | Effect on reply rate |
|---|---|
| Email 1 (initial) | Baseline |
| Email 2 (first follow-up) | +49% replies |
| Email 3 | −20% replies vs. 2023 baseline (was +9% lift in 2023) |
| Email 4+ | −55% replies; trains filters to flag the sender as bulk |
The rule: one initial + one follow-up, then stop. Two reasons:
- More emails train spam filters that the sender is a bulk operator. The deliverability cost outweighs the marginal reply.
- If they didn't reply by email 2, they're not interested in this offer — they may be interested in a different angle.
The recovery move is non-responder recycling: pull the list of non-responders, build a new campaign with a different opener and a different value offer, and try again. Same domain reputation, fresh angle. This is the structural replacement for follow-ups #3–#7.
What "engagement-first" means in operating terms
Schneider keeps coming back to the same loop:
- Filters reward engagement (replies, clicks, no complaints).
- Engagement comes from relevance plus a real value offer.
- Relevance comes from micro-segmentation × AI enrichment.
- Real value offers come from solving a small piece of the recipient's problem for free.
- Deliverability is the floor that lets any of this be visible to the recipient.
This stacks: infrastructure → list segmentation → enriched opener with real value → two-touch cadence → recycle non-responders. Each step is a pre-condition for the next.
Where Schneider's system diverges from the rest of cold-email-internet
- Inbox-placement math is explicit. Most cold-email content treats deliverability as a black box. Schneider gives a number per account, per domain, and a warm-up window.
- Reply rate has an arithmetic ceiling. The under-50-recipient figure is the most useful data point — it explains why over-personalization arguments and over-volume arguments both miss the real driver, which is relevance.
- Follow-ups are not the meeting engine. This is the most direct disagreement with the SDR-school consensus (e.g., Connor Murray's "70–80% of meetings come from follow-ups"). Schneider's data says past touch #2 you're harming deliverability faster than you're harvesting replies.
- Loom videos are out. A common 2023–2024 personalization tactic. He calls it dead in 2026.
- Recycle, don't perpetuate. Non-responders are a list to re-campaign, not a list to keep emailing.
A reconciliation with the SDR-school view
The Murray and Schneider playbooks look like they disagree on follow-up cadence — Murray runs four touches, Schneider runs two. The reconciliation is list scale and offer type:
- Murray is running strategic-account or enterprise volume with a single sender identity, where each prospect is high-value and the follow-up cost per prospect is acceptable. His follow-up sequence is short and assumptive (three notes that all redirect to the original), not a 14-touch carousel.
- Schneider is running agency-scale volume with rotated inboxes, where every extra touch is amortized across hundreds of thousands of recipients and any filter-flag is catastrophic.
The skill should let the agent ask which mode the campaign is in. For most BuildOS users running founder-led outreach, Schneider's 2-touch with recycled non-responders is the safer default. For strategic-account or high-trust founder relationships, Murray's 4-touch is more appropriate.
What this layer contributes to the BuildOS skill
The Schneider analysis supplies the deliverability floor, the relevance-via-segmentation math, the "value is a deliverable not a meeting" rule, and the two-touch cadence of the skill. It is the platform-realist layer that prevents the agent from recommending tactics that work in theory but get the sender's domain torched in practice.