Sotirios Gianniris

Sotirios GiannirisSep 9, 2026

This summer, cold email in the EU changed. Τt's now harder to stay safe & compliant the recent EU AI act, article 50, kicked in on 2 August 2026. and yeah the internet spent the last month telling outbound operators to slap "written by AI" on every send, but from what i've seen that advice is wrong. Together with our deliverability guru Simon Kostorrizos we investigated what matters & what the act covers: > deepfakes have to be disclosed (ok of course duh) > chatbots have to tell users they're chatting with an AI (realistically applies only to AI SDRs, who cares) > AI providers (OpenAI, Anthropic, etc.) have to machine-readable-mark their generative outputs (this is quite interesting since there's already a market of tools for removing those marks) > AI-gen text has to be labeled when it's published to inform the public on matters of public interest (journalism / news, not cold (sales) emails) SO: the exemption is clear: If the content went through substantive human editorial review, and a real person holds editorial responsibility, it's out of scope. Aka: if you reviewed the copy, it's compliant. Nothing new for cold email specifically. The actual news this summer was different & everyone was too busy writing compliance hot takes to notice. The boring stuff is the important stuff: > Gmail turned on LLM-grade spam filters (RETVec + Gemini Nano). The classic cold-email evasion tricks (like spintax) stopped working. > Microsoft moved to SMTP-level rejection on Outlook + Hotmail. Aka: before, the worst case was "spam folder". Now it's "never delivered." Both of those hit at the individual mailbox level, not the campaign level. Aggregate reply rate + bounce rate at the top of your dashboard lags a problem that already happened on one mailbox two weeks ago. due to these, I've been running together with Simon Kostorrizos from the team, per-mailbox checks on Smartlead's incredible insights view for a few weeks. ive seen 3 things it does, that actually matter: > it actually catches per-mailbox reply-rate drops the day they happen. If a mailbox dies on a Monday I know Tuesday, not 3 weeks later once it drags the campaign average down. > volume-weighted bounce alerts, so one mailbox drifting past the threshold doesn't hide behind 20 clean ones (matters more now that Microsoft rejects at SMTP instead of just spam-foldering). > bulk reconnect across the whole fleet in one flow. When a Gmail OAuth token dies or a workspace password rolls over, you're not sitting there re-authing 15 inboxes one at a time. Running an AI native lean team means i realistically can't afford to waste my time on admin work like that, so smartlead has been doing some real heavy lifting here

Sotirios GiannirisSep 14, 2026

The single best skillfor a GTM Engineer in 2026 is TAM (Total Addressable Market) modeling. and here's why: I recently made a post about why companies are willing to pay GTM engineers 250k/year to build assets that keep serving the company even AFTER they leave. (and got 70,000 views in under 48 hours) Clay just raised another 115million, but at the same time more and more GTME's are saying "goodbye" to it Why? Its no longer necessary... and quite frankly, companies are not into "clay operators" anymore. So what is a TAM model? Let me explain what its not: it's not a random pull of 11-50, Software Development, in X, Y and Z country based companies, filtered perhaps by a quick AI agent at best it's not a spreadsheet of 10,000 companies with emails attached, where nobody can explain why each account belongs or which ones sales should prioritize it's not a fresh list for every campaign, built with slightly different filters each time, making it impossible to tell whether the specific segment performed better or you just changed who was in it It is though: a map of the companies you can realistically sell to, WHY they fit, HOW they differ from each other, and who matters inside each account, all calculated by AI agents that operate on very specific rules that always stay the same. We assessed 8,732 companies for our MSSP market map. Only 1,544 qualified, the rest would've been the cause of a failed outbound campaign, just noise I've used Deepline here and it makes the enrichment part easier, giving your coding agent access to multiple data providers. But the value is in the processing: agents running on your rubrics turn that data into consistent answers: does this company qualify, and WHY? WHAT does it sell, and to whom? WHO should we approach? (buying committee) Sales can compare segments and find the best, instead of rebuilding lists every month and comparing lists that compare as well as oranges compare to apples. Even job titles need context. A Head of IT can be your champion at a 15-person firm, but the wrong contact at a 600-person company with a CISO and SOC leadership. recorded a short loom on how to do this - comment "tam" below and ill shoot it over"

Sotirios GiannirisSep 10, 2026

A few notes on overblown sales teams like Chris Balestras points out here; but from a strictly GTM POV: 1. the role of the GTME is fundamentally built around efficiency & if you're looking to join a company, if they're aggressively hiring sales workforce isn't always a particularly good signal they'll be doing well over the next few years 2. If you haven't proven demand yet, a GTME that's more outbound focused with SDR experience is going to have 10x the ROI of 5 SDRs. Let them run controlled outbound experiments + 1 cold caller SDR on their side to warm call: - SUPER high fit prospects & hunt them - and also positive replies from cold email / linkedin outbound 3. VC capital subsidizing most of this wave + hype coming from companies like Clay (and rightfully so) but over time, just like stocks, only the sustainable sales motions will survive, not the overblown sales teams