Room for Judgement: How Trellis Keeps Every Send Human
Where a person decides in Trellis: which accounts we work, what each email says, and whether a campaign launches, with do-not-contact above every approval.
Trellis is the workspace every Lead Line Partners engagement runs in, and it's built so that a person decides before anything reaches a buyer. No campaign email goes out without approval from your executive or our team. A reply your seller types in Inbox is theirs to send.
AI is part of the work, and our team drafts with it outside Trellis. But inside Trellis the AI can't send, approve or dial.
Sales needs to be an undeniably human function, and we built Trellis to keep it that way across more accounts than a team could research by hand. The judgement sits at three points: the accounts we work, the words each person reads, and the moment a campaign launches. And a few decisions outrank all three. I'd rather have a slower send with a name on it than a fast one nobody can account for. Announcing Trellis covers the rest of the product.
Where the AI stops
Our team writes every email, with AI, and the drafts come into Trellis for you to read. Inside Trellis the AI has a short list of jobs. It summarizes accounts and people. It answers your questions in Ask, from what's in your workspace, and names the records it read. And it prepares a brief before a seller's call and gives that seller cues during it.
That's the list.
But it can't send, approve or dial. It can't change who's in a sequence, and it can't edit the Brain (the written record of your business that our team works from: who you sell to, why they buy and how you sound). An answer from Ask stays with the person who asked until they choose to keep it, by pinning it to a summary, saving it as a note or adding it to a call brief.
The research and preparation repeat across many accounts, so we do them the same way each time. Every decision about what goes out, and to whom, stays with a person, which is why every approval in Trellis has a name on it.
Which accounts we work
But judgement starts before any email exists.
We find the accounts, and you choose which ones we work. Each proposed account arrives on Radar with why we picked it, the tier it earned and the dated signals behind both. Nothing arrives pre-approved. You approve, hold or decline, and a decline carries a reason.
Write the reason down, because before each new search we read what you decided and why. And when we send research again, your decisions and edits stay as you left them.
The dates matter too. A hiring pattern from last week and one from last spring tell you different things, and you can weigh each against what you already know about your own market. When our team approves an account on your behalf, Trellis records who did it and when.
And you can add accounts yourself, too. They arrive without our reasons, since we didn't pick them, and they wait for a decision like any other. And full research runs only on the accounts you accept, which keeps the deep work pointed where you've chosen.
What each person is told
Copy Review is where you read the words. It shows each person's whole sequence word for word, branches included, so you read what that person will read, in the order they'll read it. Each email shows who it comes from and any seats copied on it.
Who writes matters as much as what's written, because seniority writes to seniority: your BDR writes to the managers while your Sales Director writes to their vice president. And if the seat on an email looks wrong for the person reading it, you'll see that here, before it goes.

Every decision on this screen belongs to the reader: what to fix, whether to hold, and one approval for one person. The company, people and email text are illustrative.
The example is one person's sequence, a VP of data and analytics, and the map at the top lays out their path. One seller opens. A second seat is copied on that first email, and in this example one branch brings the second seat in to write if the emails are opened on at least two different days, while another covers a reply. Below the map sits every email in full, each with its sender and its copied seat. At the bottom there's one approval block, for that person only.
Fix anything in place. Change an email after approving it and the sequence goes back to awaiting review. One approval covers one person and their sequence. Once an email has gone out, you can't rewrite it.
For a buying committee, one person at a time is the right grain, since a technical evaluator and an economic buyer at the same account read different emails, built on the same research.
Launch asks twice
Approving people doesn't start a campaign. Launch takes one more yes, from your executive or from our team, and Trellis asks for it twice.
That yes also approves any email nobody has decided on yet. Trellis asks twice because it covers emails nobody approved one by one, so read the undecided ones first in Copy Review. If your executive wants every email read before launch, read the undecided ones first, then the rest.
Every approval records who gave it, in what role, and when, and launch records what it covered. When someone asks later who approved an email, the answer is on file.
Your executive can also pause a running campaign. You don't have to be in Trellis to know a decision is waiting on you, because it can reach you by email.
Who is left alone
Some decisions outrank every approval. A do-not-contact entry beats every approval, your executive's included, and every person is checked against the list again before anything goes out.
From a person's drawer, Stop lets you pause them with a reason, or set do not contact for the person, the address or, for your executive, the whole company. That entry can't be undone from the drawer.
The same menu lets you flag a person as the wrong fit, as someone who has left the company, or as holding a different title. Left the company pauses them, and you can resume them later.
A title from a lookup is not proof someone still works there. Before anyone at an accepted account is written to, we check on our side that they still work there. And a send waits until that check is on file and current.
And the mechanical stops sit underneath all of it. An unsubscribe or a hard bounce blocks the address, a soft bounce is only stored, and a do-not-contact entry stops every seat on every channel.
After the reply
A reply hands the conversation to a person. One reply pauses that person for every seat, and so does a call that connects, whatever the outcome, because a seller now has the conversation. And they stay paused in every other campaign for your company until someone resumes them by hand. An out-of-office doesn't stop anything. (If one is missed, it can be marked so the person goes back in.)
Replies land in Inbox, one thread per person, and each goes back to the seller it belongs to. Our team can't reply in your name. When your seller answers, the reply goes from the address the buyer wrote to. The buyer sees one conversation with one sender.
You also say what kind of reply it was: Interested, Meeting booked, Referral, Not now, Not interested, Neutral or Wrong person. A reply from an address we can't place lands as an unmatched reply, with the original message beside it. Nothing runs on it automatically. You choose whether to stop the sequence or ignore it.
And calls stay with people too. Outbound goes out under your real sellers' names, through each seller's digital twin, the identity their outbound carries. But one of your sellers places every call. None by a twin. And nothing records until the seller presses Record, and only the person on that number can start or stop it.
LinkedIn steps are tasks for the seller who works the account, and the seller marks each one done once it's sent. A connect request waits until the seller marks it accepted. And when your team's connections match a target account, nothing is sent from the match. Whether to pursue it is yours.
The limits
We don't promise meetings. But Trellis shows you what happened, whatever it was.
Copy Review takes a person's time. Reading each sequence word for word is slower than skimming a campaign, and no screen in Trellis makes it faster. And that's the price of a name on every approval.
Human in the loop here means a person decides, and our team drafts with AI.
Trellis suits B2B companies that sell to buyers in the United States and want a say in which accounts we work. And if you want your sellers' names on the outbound and your own people deciding what goes out under them, this is where that happens. If that describes your team, bring your seats and the accounts you'd want worked first.
Frequently asked questions
Does anything send in Trellis without a person approving it?
No campaign email goes out without approval from your executive or the Lead Line Partners team. Launch takes one more yes, and Trellis asks for it twice, because that yes also approves any email nobody has decided on yet, so read those first. A do-not-contact entry beats every approval. But a reply your seller types in Inbox is the exception: it's theirs to send.
Who can approve a campaign email in Trellis?
Your executive or our team. Each person's sequence is approved in Copy Review, with one approval covering one person and their sequence, and launch takes one more yes from either side. Every approval records who gave it, in what role and when, and launch records what it covered, so when someone asks later who approved an email, the answer is on file.
What stops when a buyer replies?
One reply pauses that person for every seat, and so does a call that connects. They stay paused in every other campaign for your company until someone resumes them by hand. But an out-of-office doesn't stop anything. An unsubscribe or a hard bounce blocks the address, and a do-not-contact entry stops every seat on every channel. The reply itself goes back to the seller it belongs to.
Can the AI in Trellis send or approve on its own?
No. Our team drafts with AI outside Trellis. Inside it, the AI summarizes accounts and people, answers questions in Ask, prepares a brief before a call and gives the seller cues during it. It can't send, approve or dial, change who's in a sequence, or edit the Brain, and every approval in Trellis has a person's name on it.
Who decides which accounts Lead Line Partners works?
You do. We find the accounts, and each one arrives on Radar with why we picked it, the tier it earned and the dated signals behind both. You approve, hold or decline, and a decline carries a reason. Nothing arrives pre-approved. When our team approves an account on your behalf, Trellis records who did it and when.
Supporting
What You Should Be Able to Open When You Hire an ABM Agency
An ABM agency writes to your buyers under your name. Seven things you should be able to open while the work can still change, from the account list to the bill.
What Should a Reply Stop? A Decision Table for Outbound From Several Sellers
When several sellers write to one account, a reply has to stop the person, not one sequence. Eight events, with who stops, for how long, who restarts, and where each rule comes from.
Score Every Account, Contact Only a Beachhead
Scoring a company from a domain cost 1.99 records in a 200-company test. That makes scoring a whole market cheap enough to test. The limit moves to review, sending and the employment check.