AI Sales Replies That Sound Like Your Team, Not Like Everyone Else
You run a sales team of two to ten SDRs and AEs. You are measured on pipeline, and on how long a new rep takes to start producing. Your sequencer works. The touches go out, the volume is there, the dashboard is green.
The threads still stall. Not the ones your best closer is in. Theirs keep moving, and they have never been able to explain what they do differently. Everyone else approximates it badly, and the AE who started this quarter is still guessing at a method nobody on the team has written down.
That is a voice problem. Sending more of the same messages makes it louder, not better.
The problem is sameness, not volume#
Here is a count you can run this afternoon without buying anything. Open the last twenty first-touch replies your team sent. Count how many open on the prospect's specific problem, and how many open on a paragraph about you. Count how many end with a question a prospect can answer in one line, and how many end with "let me know."
That ratio predicts your reply rate better than your send volume does, and it is the one number your sequencer cannot move. Volume tooling was built for a bottleneck you no longer have.
The tooling most teams reach for next makes the ratio worse. A general chat tab writes competently in nobody's voice. The rep pastes the prospect's message in, pastes the draft back out, then edits the register back toward the way the team actually writes, on every single message. Meanwhile the gap between your best rep and the other five is not effort. It is that your best rep's method lives in their head and has never left it.
Encode the voice once, then let every seat draft from it#
Ezly's answer to that is a persona, which is a writing identity rather than a tone dropdown. A persona carries a professional role and level of experience, a communication style, and guardrails in two explicit lists: what to always mention, and what to never do.
You then upload your own material to it, in PDF, DOCX, DOC, TXT, or MD. Ezly writes the drafts from your case studies, your offers, and the messages that already closed, rather than from a model's general sense of how a sales email sounds.
On Ezly Team, that persona is shared. The whole team drafts from one pool, and every seat also keeps one private persona that only that seat can see. Whoever owns the way your team writes defines it once, and the AE who started this quarter drafts from it in their first week rather than reverse-engineering it over three months.
Underneath the persona sits a sales method that loads into every generation, so the rep never has to learn how to prompt. Replies come out already shaped: a hook on the prospect's specific problem, then proof, then either a qualifying question or a push toward a call. Never "let me know." The encoded principles are the ones your best closer already follows without naming them. Listen first. Qualify fast. Value before credentials. Brevity, and no exaggeration. Ezly treats initial inquiries, follow-ups, and objections on price, timeline, and trust as separate situations.
Tone stays separate from the persona and lives at the message level: four built-in tones, plus named custom tones you write the directive for yourself.
What happens inside the live thread#
Drafting happens where the conversation already is. The Chrome extension opens inside the platform's own chat on LinkedIn, Upwork, Gmail, Fiverr, and Discord, and the web app is never required to use it. When the panel opens it has already picked up the client's name and the context of the thread, and it offers one to three suggested next replies read off the live conversation, switchable by tone.
Before it writes, it reads. Ezly classifies the conversation two ways: whether the counterpart is a real lead, and where the deal stands across five stages, cold outreach, discovery, proposal, negotiation, and closing. The suggestions follow that read, because a reply to somebody in negotiation has a different job than a reply to somebody who has not answered yet.
The web app composer keeps two things apart on purpose. The prospect's message goes in one field, your private instructions in the other, and the two never mix. Next to them sit one-click refinements: Shorter, Longer, More casual, More formal, Push for a call, Qualify. You can dictate into the composer by voice. You can use it as a guest with no account, which is the cheapest way to test this against a real thread before you involve the team.
You can lift any conversation into the web app on demand, and after that both surfaces stay live. Threads are keyed to the conversation URL, so reopening one months later reconnects its full history and its persona.
Two more checks run without anyone asking. Ezly compares every draft against marketplace rules before you send, so off-platform solicitation and unsupported guarantee language get caught. And it writes every generation to sound human by default, with a one-click pass on any reply when you want more of it.
A person sends every message#
Ezly is a copilot. Drafts insert where the rep types, and nothing sends itself. Nothing runs against the platform either: the extension reads the visible text of the recent conversation, executes no script on the page, and automates nothing.
That is a stance, and it costs us something. Ezly answers nobody while you sleep, and it will never appear in a comparison table under unattended outreach. The market is drifting toward tools that act in the customer's name, and we are not following it, because a message that goes out under a rep's name should have been read by that rep before it went.
The rule also protects the asset you are buying. A shared voice stays a voice for exactly as long as a person is still checking that the draft sounds like the firm.
Where the limits are#
The deal-stage read is a manual trigger. You ask for it when you want it. It is not a watcher sitting behind every thread refreshing itself.
A conversation that has not been synced to the web app lives on the client side, and clearing the browser's local storage can lose it. Sync the threads that matter.
The quick-reply library, curated hooks and rebuttals and boundary scripts tagged by situation, alongside replies your team writes itself, is a library page today. You copy a reply to the clipboard from there. There is no picker inside the composer yet.
Platform rate limits still apply and always did. Exceed a platform's messaging limits and accounts get banned with or without an extension, because that is the platform's rule and not an Ezly risk.
Security answers here are mechanism-level, not certificate-level. Sign in with Google or with email and password, see and revoke active sessions and devices, and read the extension's access model above. The product runs in English, French, and German.
Ezly is shipped#
Ezly is shipped, and that is a plain statement about what exists. The web app is live, the Chrome extension is live, and sign-up works today. Ezly has 400 users across the web app and the Chrome extension, and 28 paying customers.
Silverthread Labs built Ezly and ran its own multi-account pipeline through it. The shared persona is not a feature that came out of a roadmap workshop. It came from watching an account manager hold a warm reply for a day, because the one person who knew how that client should be answered was busy being that one person.
Teams buy it as Ezly Team, and they buy it in the app: create the org, invite members, and manage seats and billing centrally, without talking to us first.
If you run a small sales team and your reps read like a template, there are two places to go next. The Ezly product page covers what it does and who it is for. Ezly's own site is where you sign up, and where the pricing lives.
