The most dangerous email address in B2B sales is [email protected]. Since August 18, anyone can send, forward, or cc it on a thread, and Perplexity Computer completes the task inside the email: it reads the thread context and replies to the verified sender, wielding that person's connectors, permissions, and Memory. The launch examples are cheerful productivity fare, financial models and contract summaries. The uncheerful implication for anyone who sells: Perplexity email tasks put an analyst inside your buyer's inbox, one forward away from every pitch you send.
What Shipped, Precisely
The mechanics reward attention. The agent answers only the verified sender, in the same thread, so it behaves like a trusted colleague rather than an open endpoint. It arrives with the sender's context: their connected tools and permissions, plus the agent Memory of everything that worked before. Enterprise reply-all support is on the roadmap, which will put the agent into group threads where deals actually get discussed. And it's available to every Computer user, so this is a default capability now, and a preview of where every mail provider is headed.
Perplexity's own suggested uses say plenty: build a five-year DCF from a thread and send back the Excel model, or summarize a contract and draft the reply. Delegated knowledge work, initiated from the place B2B decisions already live.
Your Cold Email Now Has a Second Reader
Here's the scenario that should reorganize your outbound playbook. Your carefully sequenced email lands in a buyer's inbox. The buyer doesn't reply or click, and may not even finish reading. They forward it: "Is this legit? How do they compare to what we use?"
The agent researches you with the full weight of its retrieval, whatever the engines currently believe about your brand, your reviews, your docs, your competitors, and replies into the thread. Its verdict sits directly under your pitch, in the same scroll. You were never in that conversation, and it was the most important conversation your email started.
Between you and me, this was already happening with buyers pasting emails into chatbots. The email-native version removes the last bit of friction, and friction was doing more work for vendors than anyone admitted.
26 content formats. Published on your domain. Matched to your brand.
Guides, comparisons, listicles, case studies, and more. RankControl generates content that gets cited by ChatGPT, Perplexity, Claude, Gemini, Grok, and Google AI Mode.
Discovery Moved Into the Thread
The forward-my-pitch case is defensive. The offensive case is bigger: buyers initiating research from threads that mention a problem. Someone complains about churn tooling in a thread, cc's the agent, and asks for options. The shortlist comes back into the thread, assembled from whatever sources the engines trust, shaped by what the agent's memory already banked about tools that worked before.
No search box was opened. No landing page recorded a visit worth attributing. The agentic work wave keeps producing versions of this pattern, and the inbox version is the most invisible yet: research that looks, from your analytics, like nothing at all, followed weeks later by a demo request from a company you've apparently never touched.
Writing for the Machine That Reads Over Shoulders
So what changes in the actual writing? Four things, none of them exotic.
- Make claims the agent can verify. "Cuts onboarding time 40% for mid-market teams, methodology here" survives a vetting pass. "Revolutionary AI-powered platform" gets summarized into nothing or flagged as fluff. Every claim should have a public page behind it.
- Keep your email and your website in agreement. The agent follows your links and cross-checks. A pitch that promises what the pricing page contradicts earns a written verdict saying so, delivered right under your signature.
- Front-load the factual skeleton. Machine summarization keeps concrete nouns and numbers and discards atmosphere. Read your last sequence and ask what survives if only facts remain. That surviving residue is your actual pitch now.
- Forward your own emails to an agent before sending the sequence. Honestly, this is the cheapest QA step in modern outbound: read the verdict a buyer would read, then fix what it got wrong or doubted.
And notice the loop underneath all four: the agent's verdict draws on what AI systems already believe about you. The email is just the trigger. The evaluation runs on your citations, your reviews, your docs, your entity presence, which is why tracking how engines describe your brand has become outbound infrastructure, strange as that sentence would have sounded in 2024. RankControl's agents watch that description layer across engines weekly; the outbound side of your house should be reading those reports as closely as the content side.
Your competitors are getting cited by AI. You're not.
Every day without citation tracking is a day your competitors pull ahead in ChatGPT, Perplexity, and Claude.
A New Objection Class Is Coming to Your Replies
Sales teams should brief for a reply type that barely existed last quarter: "I ran this by our AI and it said..." The objection arrives pre-researched, citing whatever the agent found, and arguing with it blind is a losing posture. The rep who can say "that comparison misses our new tier, here's the page" wins the exchange; the rep who bluffs against a machine-compiled brief loses credibility twice.
Two operational habits make this manageable. First, keep a living document of what agents currently get wrong about you, sourced from your own weekly forwarding tests, so reps walk in knowing the machine's talking points. Second, when an agent verdict contains an actual error, treat it as a visibility bug: find which stale page or wrong record feeds it, fix the source, and re-test. Arguing with the output does nothing. Correcting the input fixes every future thread at once.
That's the quiet discipline shift: outbound teams inherit a maintenance loop that used to belong entirely to SEO.
The Inbox Is the New SERP, Sort Of
One thing I should've mentioned earlier: email marketing survives this fine. What changes is that email stops being a private channel between your copywriter and a human skimmer. Every message now potentially enters the same evaluation machinery as your website, your reviews, and your docs, judged for consistency against all of them.
The vendors that thrive will treat that as an audit they can pass on purpose: verifiable claims and consistent facts, with a public footprint that backs the pitch. The ones that struggle will keep writing for the skimmer alone and wonder why sequences that used to book meetings now produce polite silence. The machine reader is already cc'd.
200+ SaaS teams already track their AI citations.
They know exactly when ChatGPT mentions their brand, and when it stops. Do you?
