Ravi Shankar

Building ColdMail and AI native growth systems for B2B SaaS. Sharing operator notes on GTM, AI agents, outbound, founder led sales and SaaS growth.

[Ravi Shankar @ruhhvee]

i am a technical founder building AI-native growth systems for b2b companies.

i write about GTM, AI agents, outbound, founder sales, product decsions, and building global saas.

ColdStack Labs is where i test these ideas in the real world.

[Ravi Shankar @ruhhvee]

@tdinh_me keyboards are a hard place to win on features alone. it's the one app that sees everything you type, so the real switching cost is trust, not price.

[Ravi Shankar @ruhhvee]

@cbwritescopy the ramp still matters though. an inbox going from zero to full volume on day 1 is exactly the pattern filters watch, whatever it did before you got it.

[Ravi Shankar @ruhhvee]

@coldlytics B for me. replies over emails sent punishes extra steps: add a follow up and the rate drops even when more people replied. per person contacted is the only one you can compare across sequences.

[Ravi Shankar @ruhhvee]

@RCIllingworth the slight drop is the hard one. every single day still looks normal, so nothing alerts. comparing this week against the last four, split by provider, shows it well before any one day does.

[Ravi Shankar @ruhhvee]

@arvidkahl same gap in sending tools. if the agent running a campaign can't read how much of today's per domain limit is left, it can't pace itself. it just sends until something starts bouncing.

[Ravi Shankar @ruhhvee]

@MichLieben closed lost is the useful half. most icp docs only describe who bought, so the list still fills up with companies that look right on paper and always say no.

[Ravi Shankar @ruhhvee]

most email monitoring alerts on errors.

silent failures don't throw errors. every check passes, nothing is technically broken, the result is just wrong.

alert on the thing that should always be true. replies by provider, placement on a seed, bounces staying flat. that's where the scary bugs show up first.

[Ravi Shankar @ruhhvee]

worth keeping in mind next time someone says their copy stopped working. per this validity data, gmail inbox placement slid from about 91% to 82% this year. some of that bad copy is mail nobody ever saw.

[Letterland LIVE @LetterlandLIVE]

Today in "things you hate to see..."

Validity's data showed Gmail inbox placement is tanking.

And they've got receipts...

Q1 2026: Gmail inbox placement averaged 90.97%.

Q3 2026: 81.57%.

Spam placement jumped from 5.4% to 9.5% during that time.

Write a B2B newsletter? It's worse. @GoogleWorkspace inbox placement dropped from 88.9% to 74.3%. And that's before corporate spam filters get their turn.

Validity's theory is a perfect storm: the 2024 bulk sender rules are now fully enforced, the relevance-sorted Promotions tab, and AI summaries in the inbox.

Validity's advice: send less and only to people who actually open.

@JPMorinChase shared this and 11 other things you need to know about the newsletter industry in today's Letterland Brief.

You can join for free at https://t.co/T0IWU672Tn

[Ravi Shankar @ruhhvee]

@KyleMau knowing the customer shows up in deliverability too, not just replies. a tight list where people actually open keeps the domain healthy, so a 1 in 500 rate tends to hold longer than a blast would.

[Ravi Shankar @ruhhvee]

@RCIllingworth the ramp gets skipped more than the dns. warmup finishes, volume jumps the same day, and nothing in the sequencer shows placement sliding until replies dry up.

[Ravi Shankar @ruhhvee]

@jasonlk funny part is that's also when attribution breaks. leads show up from places you never tracked, and the first instinct is to hunt for the channel instead of asking what changed in the product.

[Ravi Shankar @ruhhvee]

@marclou @jessethanley email is a good one to learn this way because the docs only cover setup. what actually decides placement is how recipients react over months, and that part you only learn by watching your own numbers move.

[Ravi Shankar @ruhhvee]

ai can write email 1 from a profile.

email 3 is the hard one. it has to know what happened after 1 and 2.

most sequences still send step 3 on day 7 no matter what. that's not follow up, that's a timer.

[Ravi Shankar @ruhhvee]

agents writing to fewer, better fit accounts also means less volume per domain.

so part of a jump like that is the copy, and part is just more of it landing in the inbox. worth knowing which before you copy the playbook.

[Jason ✨👾SaaStr.Ai✨ Lemkin @jasonlk]

New!! Databricks CRO Ron Gabrisko: 5 Learnings on Selling at a $190B AI Company

Flat SDR headcount. Reply rates up from 1-2% to 6-10%. And free seats unlocked consumption.

AI did not replace sellers. It raised the bar for them.

https://t.co/Zg1ANbhLdk

[Ravi Shankar @ruhhvee]

@yongfook building 3000 got cheap. getting the first 10 customers for each of them didn't. distribution is still one product at a time.

[Ravi Shankar @ruhhvee]

@MichLieben the micro list point is the one i'd underline. a 100k blast doesn't just reply worse, it teaches the providers to file mail from those domains lower, so the good list you send next week pays for it too.

[Ravi Shankar @ruhhvee]

@NickAbraham12 agree at that volume. the cost people skip is reputation: every extra step is another send per domain. drop steps 3 and 4 and the same inboxes can carry more first emails without looking any heavier to the filters.

[Ravi Shankar @ruhhvee]

"get back to me in q4" is worth chasing only if they told you why.

a contract ending, a budget reset, a hire starting. that's a real follow up.

no reason usually meant a polite no.

[Ravi Shankar @ruhhvee]

outbound tools are heading the same way. if a campaign setting only exists as a button in the dashboard, the agent running the campaign can't touch it, and that becomes the thing people work around.

[Arvid Kahl @arvidkahl]

Getting more and more of that from my customers, directly or indirectly (by asking for feature parity in the API and MCP).

SaaS products of the future MUST present all affordances equally to humans, agents, and machines. UI, MCP, and API have to be fully interchangeable.

[Ravi Shankar @ruhhvee]

@peeplaja same split in cold email. ai made it cheap to reach out with an actual idea, and just as cheap to send something that clearly never looked at you. the reader can tell which one by the first line.