Attach actions to your timeline steps, write execution instructions an agent can prepare against, and route each action to the owner, function or role that should hold it.
Inside a timeline step, each action asks for three decisions: what the work is (its instructions), who it routes to (an owner, a function, or a role), and how autonomous it is, meaning whether the agent completes it or prepares it and waits for approval. The due date follows from the step's position on the timeline.
When the play is live and an account reaches that step, the action becomes a real item in someone's Action Center, carrying the account it belongs to, the instructions you wrote here, the due date and the standard controls. Completing it there feeds straight back into the play's progress.
Hold that moment in mind while you write, because it tells you who you are writing for. Your instructions are read by a person with a queue of other work, on a morning when they are already behind. They have to say what this is about, what has been prepared, and what decision is being asked, without sending anyone to four other tabs to find out. That is the test to write against.
The execution instructions are doing two jobs at once. They tell the agent what to assemble, and they tell the person what to decide. Written well, the two fit together: the agent arrives with the pilot results, the stakeholder map and a draft that references both, and the person decides whether the framing is right for this account.
The failure mode is writing an instruction that describes only the physical act. "Send a follow-up email" gives an agent nothing to prepare against, there is no stated purpose, no material to draw on, and no way to tell a good draft from a bad one. "Draft outreach to the economic buyer that leads with the pilot outcome and asks for a scoping conversation" gives it all three.
Route actions to a function or role wherever you can, rather than to a named individual. People move between teams, go on leave, leave the company; roles do not. A play routed to individuals is correct on the day it is built and quietly wrong three months later, with actions landing in the queue of someone who stopped owning the account in April.
Each action also carries an autonomy level, and one rule holds regardless of what you set: external touches never fire on their own. So the question is never whether a person is involved in outbound. It is how much is ready before they are. Setting that deliberately is the whole of the next lesson.
Before: "Send a follow-up email to the account." Everything is missing — the purpose, the material, the decision being asked of the human who approves it.
After: "Draft outreach to the economic buyer re-opening the conversation. Lead with the pilot's measured outcome (in the account plan's pilot summary). Reference the success criteria they set at kickoff. Ask for a 30-minute scoping conversation, not a demo. Approver: check the framing fits what you know of this account's current priorities, and edit the emphasis rather than the structure."
Read the after version as the agent: it names the audience, the opening argument, the source material, and the specific ask, so there is enough to prepare a real draft. Read it as the approver: the last sentence tells you what your judgment is actually needed for. That double audience is the standard for every instruction you write.
Do this in the product
You should be able to answer each of these from memory before opening it. Recalling the answer is what makes it stick; recognizing it when you read it does not.
People move teams, take leave and leave the company; roles persist. A play routed to individuals is correct the day it is built and quietly wrong months later, with actions landing on someone who no longer owns the account.
It describes only the physical act. There is no stated purpose, no material to draw on, and no way to tell a good draft from a bad one. Name what the step achieves and which material the draft should lean on.
Once a play is live, its actions appear in the Action Center as assigned work, routed to the owner, function or role the action names, carrying its instructions, due date and the account it belongs to.
The instructions tell the agent what to assemble and the person what to decide. "Draft the outreach referencing the pilot outcome" prepares work; "send an email" does not.
Each action spells out what the agent prepares, what a human reviews and approves, and what happens after. Nothing leaves the building without a decision.
See agents carry the repeatable work of GTM across sales, marketing, customer success, and RevOps. Every action prepared, reviewed, and recorded. Fictional data, real product.
Explore the demo