The short version
- Moving from spreadsheet prospecting to an AI co-worker means writing down what the status column quietly did. Here is who drafts, who approves, and who owns the trail.
Contents
- 1What does an approval handoff actually define?
- 2Which steps does the AI co-worker draft?
- 3Who approves outbound, and how fast?
- 4Who should own the approval handoff trail?
- 5What should the handoff say when the approver is absent?
- 6Does the approval handoff survive a reorganisation?
- 7How do other tools handle the same handoff?
- 8What to write down first
A good approval handoff for spreadsheet-run prospecting says three things. The AI co-worker drafts research, lists and message copy. A named approver role accepts or rejects every outbound send. RevOps owns the logged trail. The detail beneath those three lines decides whether the approval handoff still works when someone is on leave or a manager changes.
What does an approval handoff actually define?
A handoff is a mechanism, not an attitude. It has a trigger, an acceptance condition, an owner on each side and a rejection path. In a spreadsheet, most of those were implicit. A row turned yellow, an SDR pinged a lead, someone hit send. The status column was the entire protocol, and nobody wrote it down because it never failed loudly.
Migrating means writing down what that column was quietly doing. The trigger is a draft marked ready for review. The acceptance condition is that the approver can see the account, the research behind it, the exact message and the recipient list, and says yes. The rejection path is the part most documents skip: a rejected item returns to the drafter with a reason code, gets revised and is resubmitted. Without that path, rejections pile up in a corner and approvers learn to avoid the queue. That is a design gap, not a motivation problem.
Which steps does the AI co-worker draft?
Drafting is cheap to hand over. Sending is where the document earns its keep. A workable split looks like this:
| Step | Drafted by the AI co-worker | Approver role | Rejection returns to |
|---|---|---|---|
| Account research and list | Yes | RevOps | Drafter, with reason code |
| Message copy | Yes | Sales manager | Drafter, with edit notes |
| Sequence send | Prepared, not released | Sales manager | Drafter, held unsent |
| CRM field writes | Yes | RevOps | Drafter, change reverted |
| Paid enrichment spend | Requested | RevOps | Request closed |
Prefer removing a step over adding a check. If two roles currently review the same message, pick one. A second reviewer who sees the same draft adds delay without adding information.
Who approves outbound, and how fast?
One role, named by function, accepts every send. In Coco, customer-facing actions such as sending an outbound sequence sit behind a hard approval floor with no bypass. So the design question is not whether a human approves but which role and on what clock.
Put numbers in the document. As an illustrative assumption: a team that drafts 200 messages a week and clears the queue in one daily 20-minute batch can set time to acceptance at under one business day. Track the rejection rate. If it stays above roughly 30% for two weeks, the problem usually sits in the brief given to the drafter, not in the approver's taste. Fix the brief.
If the queue clears inside a day and rejections are low, leave it alone. A working handoff does not need another checkpoint to feel safe.
Who should own the approval handoff trail?
RevOps should own the log. It is the one role that sits across both sides of the handoff, so it can answer a question from sales management or from legal without being a party to the decision. Sales management owns the decisions recorded inside the trail. Legal and privacy get read access and a defined export, not custody.
Retention is set by the longest obligation that applies and written once, in the handoff document, as a number of months. The log must answer one question on demand: who approved which exact version of which message to which recipient, and when. Coco's strategies run with a logged execution trail, and each autonomous step is permission-gated and confidence-scored. Treat that as control and auditability. It makes the trail checkable. It does not prove the underlying work was right, which is why the approver still reads the draft.
What should the handoff say when the approver is absent?
Name a delegate role, a time limit and an escalation. A workable version: the sales manager role approves; if no decision arrives within one business day, the item moves to a named delegate role; after two days it goes to the RevOps lead. The default for silence is that nothing sends. An unreviewed message waiting costs a day. An unreviewed message sent costs a recipient.
Write roles, never people. The delegate list is a line in the document, reviewed whenever the team changes, so that a holiday does not become a policy decision made by whoever is online.
Does the approval handoff survive a reorganisation?
Run one test. Suppose both the sales manager and the RevOps lead are replaced next quarter. Can the new people operate the queue from the document alone? If the approval handoff depends on one manager's habit of checking the queue at nine, it fails. If it depends on roles, a trigger, an acceptance condition and a log that sits outside either manager's inbox, it survives.
Skills and strategies in Coco are versioned with rollback, which matters here. When a new manager changes a playbook, the change is a recorded version with a way back, not a silent edit to a shared sheet.
How do other tools handle the same handoff?
Vendors describe their permission models in different terms, and the handoff should be written against the behaviour you need, not a product. According to its changelog, Lindy Apps sit behind one login with per-tool permissions and support version restore. Per its pricing page, Unify holds CRM write-back for its custom, annually billed business tier, which affects who can approve writes at all on lower tiers. Viktor states it puts approval buttons on anything irreversible.
For a team whose handoff needs a hard floor on outbound, Coco fits the document directly: approval gates on sends, CRM writes and spend, a logged trail, and per-user isolated agent runtimes. It is a private beta, so there are no customer figures to cite, and no claim here that it is safer than the tools above. The case rests on the match between the mechanism and the handoff described in this post.
What to write down first
Start with the one step that carries the most volume, usually the sequence send. Write its trigger, its acceptance condition, its rejection path, its delegate and its log owner on a single page. Add the next step only when the first has run clean for a few weeks. The sheet it replaces survived on unwritten rules, and the replacement should not.
Sources
- Lindy changelog: Lindy Apps launch — Lindy, 2026-10-28
- Unify pricing — Unify, 2026-11-15