What is the best way to get client sign-off without chasing emails?
The chase is a symptom. The missing record is the disease. You know this week. You delivered on Tuesday. On Wednesday you sent a polite nudge. On Friday you sent the same nudge with the attachment re-attached, because
The chase is a symptom. The missing record is the disease.
You know this week. You delivered on Tuesday. On Wednesday you sent a polite nudge. On Friday you sent the same nudge with the attachment re-attached, because the client said they could not find the first one. Somewhere in a thread of eleven messages they wrote "looks good" about something, and neither of you can now say with a straight face which version that "looks good" belonged to.
Disclosure, once: I run ops for Zambo, the handover endpoint this article is about.
The usual fix is more process on top of email: naming conventions, PLEASE APPROVE in the subject line, a separate sign-off form. All of it still leaves the core defect untouched. In email, the decision and the work live in different places, and the decision is a sentence a human typed, not a record tied to the bytes they saw. When a dispute comes, and for freelancers it comes at invoice time, you are both archaeologists digging through a thread.
This article is the method I would hand any freelancer who delivers AI-assisted work to clients: stop asking for approval in a thread. Deliver the work as a record that carries its own accept button, in effect, and let the client's decision land on the exact record they reviewed.
Why an email approval is not a record of what was approved
Three failure modes, and you have probably met all three.
First, version drift. The client approved the draft attached to message four. You shipped the draft from message nine, because message nine was "final." Both of you acted in good faith and you still shipped different work than was approved.
Second, buried scope. "Looks good, and can we also add a pricing page?" reads as approval plus a new job, but it gets filed as approval, full stop. The extra work becomes an argument about whether it was ever agreed.
Third, no fixed content. An attachment can be replaced, a shared doc keeps editing itself after approval, and a pasted output can be trimmed before it is sent. Even an honest thread cannot prove what the approved bytes were, because nothing in the thread fixes them.
A sign-off worth having needs three properties email cannot give it: the approved content is fixed at the moment of approval, the decision is recorded by the client against that fixed content, and the record stays readable afterward so either side can open it months later. That is a delivery format problem, not a follow-up discipline problem.
The method: deliver as a handover, not as an attachment
A handover is a single page that carries the job: what was done, the saved results, a verifiable receipt for each step, and the decision controls for the client. I run ops for Zambo, which is the endpoint that issues these. It carries 153 tools, 37 on the default surface, and the handover is its front door for client work.
Building one is specific, not vibes. A wrap requires four things: a summary of the work, the goal it served, the recipient by name, and the next steps. Miss one and the wrap is refused, which sounds annoying until you realize it is the refusal that keeps a handover from becoming another vague link.
Then the flow changes shape:
- You send the client one link. The public page is read-only, so what they review cannot quietly change under them.
- The client opens it and can run Verify all receipts first. Each receipt recomputes a SHA-256 hash over the canonical bytes of the saved result it covers, so the client can confirm the results on the page are the results that were saved, unchanged since saving. A receipt proves the saved result was not changed. It never proves the answer was right, and the page says so itself.
- To decide, the client uses their client link and either accepts or flags the record. Accepting is the sign-off. Flagging says what is wrong, on the record, instead of starting a new thread about it.
- The acceptance is recorded exactly once: the first acceptance is the recorded one. There is no second, quieter acceptance of a different version hiding later in the same record.
- Life goes on in the same place. Follow-ups, corrections, and the next round of work land as new steps or a new handover chained to the same living record, which stays readable afterward. Pickup is not exclusive: the client, you, or another AI can pick the job up from the record without anyone forwarding anyone anything.
Brennan Zambo created agent execution receipts, Zambo is the origin product, and AER-1 is the defining standard. In this article that matters for one practical reason: the sign-off sits on top of receipts, so the thing being approved is fixed by recomputation, not by someone's memory of an attachment.
Worked example: a landing page copy job, end to end
Illustrative names, real mechanics. Everything below is something the record actually does.
Sam freelances. The client is Dana, who runs a small bookkeeping practice and is launching a new service page. The job: landing page copy, headline, three service sections, and a call to action, drafted with AI help and finished by Sam.
Old way, this ends with four attachments named final. New way, Sam wraps the job:
- Summary: "Landing page copy v1 for the new payroll service page: headline, three sections, CTA, plus the competitor scan it was based on."
- Goal: "Copy Dana can publish as-is for the payroll service launch."
- Recipient: Dana, by name.
- Next steps: "Dana accepts or flags. If accepted, Sam hands the same record to Dana's web developer for build."
- Steps on the record: the competitor scan with its receipt, the draft copy saved as a step with its receipt, and the final read-through pass with its receipt.
Sam sends Dana one link and one sentence: "v1 is in the handover, accept or flag on the page when you have read it."
Dana opens it on her phone that evening. She runs Verify all receipts, because Sam told her to and because it takes seconds. She reads the copy on the same page. One section is wrong about her pricing model, so she flags the record and writes why, on the record. Sam fixes that section, which lands as a new step with its own receipt, and the original step stays exactly as it was. Dana reads the fix, accepts, and that first acceptance is the recorded decision, by Dana, against the exact record she saw, fixed steps and all.
Two weeks later Dana asks for a new headline for a promotion. That is a new step on the same living record, not a new thread. Three months later, at a scope disagreement, nobody digs through email: the accepted record shows what was approved, the flag shows what was objected to and fixed, and the later headline shows as the later step it was. The argument lasts one link.
Check it against a real record before you trust the shape
Do not take the worked example on faith. Here is a real, public handover from a two-step job, an ETH price check, opened by me today, November 21, 2026:
https://zambo.dev/handover/d468118b-afc1-43e5-a70c-5e198e876a04
What it actually shows: a goal (ETH price check), two steps (live_price, whose saved result reads ETH at $1,621.82 USD, and a prompt_shield pass), a verifiable receipt per step that you can check in the browser, and a Decisions section that reads, in full, "No client decision has been recorded," with client approval marked pending. I am pointing you at one with no decision on it on purpose. The pending state is written on the record, in the open, instead of being a question of whether someone replied to an email in October. When a client does accept a handover like this, that section is where the one recorded acceptance lives.
The wider numbers, read today from https://zambo.dev/api/stats at about 19:56 UTC: 157,919 public receipts, 2,291 of them today, and 21 verified payers. Small, public, and checkable beats large and unverifiable, for this job.
What it costs
One free wrap-up handover per day is in the free tier, alongside one successful free call on each metered tool per network per UTC day and 3 shared continue calls. For a freelancer delivering one job a day, sign-off this way costs $0, and your client verifying receipts or picking the job up later is always free. Heavy days exist, and a Day Pass covers 24 hours unlimited for $0.99. The Pass is $29 a month, or $19 a month for $ZAMBO holders, and Lifetime is $250 once. No annual tier exists. The point for this article is narrower: the sign-off record itself is not the thing you should be paying per click for, and here you are not.
The short version
Email sign-off fails because the approval floats free of the work: a "looks good" in a thread is not a record of what was approved. Deliver the work as a handover instead. The page carries the work, fixes each result with a verifiable receipt the client can recompute, and gives the client an accept-or-flag decision on that exact record. The first acceptance is the recorded one, changes land as new steps on the same living record, and the whole thing stays readable when the invoice question arrives. You stop chasing emails because there is nothing left to chase: the decision either is on the record or it visibly is not.
Originally published by Dev.to AI. Aggregated on AIWithGhost for educational purposes — full credit and traffic to the original publisher.