Copilot
Copilot open full view →
Ask Assay anything.
Straight answers about Assay, grounded in our own Truth Graph. Not a generic model's best guess.
Straight answers about Assay, grounded in our own Truth Graph. Not a generic model's best guess.
Everyone promises pipeline without headcount. You've seen how it ends: a burned domain and claims nobody approved.
The control layer for your outbound. Every channel, every tool, even the other agents you run: whatever writes the message, it clears one gate before it goes. Sent from your own seats, learning claim by claim what wins, where, with whom.
In preview with design partners · runs plays through Instantly, Smartlead, and HeyReach today
Where it sits
It sits between your team's plays and your sending tools. Nothing goes out without clearing it. Every channel. Every tool. Even the other agents you run.
A check that didn't run is a block. A claim without an approved source never auto-clears.
We mapped 18 ways B2B teams send outbound today. In 12 of them, the approver is nobody.
Our GTM motion analysisIt runs plays through Instantly, Smartlead, and HeyReach today. It's designed to sit above enterprise sequencers like Outreach and Salesloft as the governed decision layer.
The demo
Three fears keep AI outbound switched off. Here's the answer to each one, as it runs.
Fear 1 — "The AI will say something wrong."
Every touch clears one gate before it goes. Every cited claim grounds against your approved truth and has to appear in the body word for word, as approved. No approved source, no auto-clear. A check that didn't run is a block. And every verdict lands in the audit chain with the reason attached.
The gate runs at send time, against your truth and your consent state as of the send, not the draft. The step-six touch your play drafted two weeks ago clears against today's story.
It says only what you've approved. If the truth changed yesterday, today's touch knows.
Fear 2 — "It'll burn our domain."
It has no sender of its own. No mailbox, no domain, no sending engine. The cleared touch goes into the tool you already warmed up and out from your own seat. Your warmup, your mailboxes, and your domain reputation stay inside the tools that already protect them.
Fear 3 — "It'll run away with itself."
Autonomy is earned, never toggled. A new play runs as approve-each until it proves itself, send by cleared send, on evidence you can see: measured precision, reversal rate, sample size. The checks never loosen as autonomy grows. Claims about pricing, proof, or policy always force review. And one switch stops everything, org-wide.
Running as approve-each until this play earns auto · 42 / 100 cleared sends
The compounding part
The winner lands back in your truth, and every play inherits it. Wins are measured on real replies and conversions, not predictions.
An asset-level A/B tool can't produce this, structurally: it doesn't know what claims its emails contain.
This agent runs the plays and attributes the outcomes. The AI CMO promotes the winning positioning with the evidence.
Fits your team
Author, owner, runner, approver. Four settings on every play, each assignable to a seat, a team, or the agent. The agent never approves its own work. And every play has a named human owner, even at full auto.
A play marketing wrote, running from the rep's own seat, with RevOps owning it and the rep approving each touch. Try describing that in a sequencer.
The receipts
Legal, brand, and security ask the same four questions. These are the answers, already written down.
One audited suppression ledger every send checks, across every tool and channel. An opt-out on LinkedIn is an opt-out in email, wherever we know it's the same person. Plus account-level holds.
Every verdict, every reason, every dispatch, hash-bound in order. The day someone asks what went out and why, the answer takes a query.
Every play has a human owner of record, even at full auto. Accountability never transfers to the software.
An org-wide kill switch. Every play, every channel, every tool, stopped in one action.
The category, plainly
Not an AI SDR. The control layer for your outbound.
It hires no personas and brings no sender. It sits above whatever already writes and sends for you, and it decides what clears. That includes the AI SDR you already run: it can pull its messaging from your governed graph, measured and approved. What it sends on its own stays its own.
The honest questions
No. An AI SDR is another seller with its own sender. This is the control layer above every seller you already have, human or AI. It composes governed plays, clears every touch through one gate, and dispatches through your own tools and seats. Nothing about your sending identity changes.
It can feed it. Your AI SDR can pull its messaging from the governed graph, measured and approved. What it sends on its own stays its own.
No. It has no mailbox, no domain, and no sending engine. Cleared touches go out through Instantly, Smartlead, or HeyReach, from seats your org owns.
Instantly, Smartlead, and HeyReach, in preview with design partners. It's designed to sit above enterprise sequencers like Outreach and Salesloft as the governed decision layer.
The touch holds, with the exact sentence and the reason attached. The gate is fail-closed: a check that didn't run is a block, and a claim with no approved source never auto-clears. Nothing goes out on a shrug.
As autonomous as each play has proven itself, and no more. New plays run approve-each. Auto is earned on visible evidence and revoked the same way. Pricing, proof, and policy claims always force review, and one org-wide switch stops everything.
Not yet. Today it composes, clears, and dispatches outbound. Replies stay with your team and your tools. Reply drafting will clear the same gate when it ships.
Who says yes
Preview · design partners
We're onboarding a small number of design partners. You bring the tools you already send with. We bring the gate, the graph, and the receipts.
Preview means preview. We'll tell you what's built, what's shipping, and what's not there yet.
Where this agent stops
This agent is the control layer for outbound. It composes, clears, and dispatches. It doesn't set the positioning it runs, the AI CMO does. It doesn't hold the veto, the AI Legal agent does. One elite team, one governed brain.