Skip to main content
Navigation

Testing AI on sales follow-ups: measure send delay and missed promises

Draft sales follow-ups with AI for a week, count send delay and missed promises, then decide whether to keep going.

Key points

  • AI drafts the follow-up from your notes
  • Sort promises vs. wishes in the draft
  • Decide recording and 4 pre-send checks
  • Count send delay and misses: keep or fix
Draft sales follow-ups with AI for a week, count send delay and missed promises, then decide whether to keep going.
A diagram of this article’s key points.
Contents

AI drafts the follow-up; you decide what to send

A&A perspective

When you test AI on sales follow-ups, give it one job: turning your meeting notes into a draft email. A sales follow-up is the message you send after a meeting to recap what was discussed and the next commitments.

A&A perspective

Whether to send it, and what goes in as a commitment, stays with you as the owner. Once the email reaches the client, it reads as your company’s promise.

From the sources

Morgan Stanley, a global financial firm, announced an AI tool in June 2024 that goes as far as drafting the follow-up email after a client meeting. With client consent, it takes notes during the meeting and surfaces the action items.

Morgan Stanley ↗

From the sources

After the meeting, the tool summarizes the key points and creates an email for the advisor to edit. The announcement says the advisor sends it at their own discretion, and a note is saved to the system that holds client records.

Morgan Stanley ↗

From the sources

Anthropic, a company that builds AI, recommends the same starting point in its guide for people building AI systems: find the simplest solution possible. More complexity is added only when it is needed.

Anthropic ↗

A&A perspective

Automate & Augment (A&A), the company writing this article, thinks this setup is small enough for an owner to test on their own sales meetings for one week. Letting AI send emails without your check is something to consider only after that week of drafts.

A&A perspective

That said, if you already send a short email yourself right after each meeting, fixing an AI draft may take more effort than it saves. In that case, first pin down what is actually going wrong with the way you send follow-ups today.

Sort promises vs. wishes in the draft

A&A perspective

Have AI write the draft so that your promises and the client’s wishes stay apart. A sales conversation mixes promises, wishes, and ideas still under consideration that nobody has agreed to yet.

Hypothetical example

For example, suppose a client says in a meeting, “If you can move our current data over, we’d like to start next month.” You answer, “I’ll confirm by Friday whether we can move it.”

Hypothetical example

Here, your only promise is to tell them by Friday whether the data can be moved. Starting next month is the client’s wish, and moving the data is still an idea under consideration.

Hypothetical example

If the draft says “You can start next month. We will move your data on Friday,” you cannot send it, however well it reads. It turns an unchecked deadline into a company promise.

A&A perspective

So before drafting, have AI list the content in three columns: your promises, the client’s wishes, and ideas under consideration. In the email itself, only items in the promises column are written as promises.

A&A perspective

Sort each item by who said they would do what, and by when. What you took on with a deadline is a promise, and what the client said they would like is a wish. Anything that hangs on an open condition, such as “if you can,” goes under ideas under consideration.

A&A perspective

When the notes give no basis for something, have AI mark it “unconfirmed” instead of filling the gap with smooth wording. That way you can spot what to fix right away.

From the sources

The same Anthropic guide says that when AI processing is split into several steps, you can add checks on the intermediate steps.

Anthropic ↗

A&A perspective

Separating the three-column list from the draft gives you a point in between to check the promises column. A&A sees this as the place where a person should look.

The example meeting, sorted (a hypothetical example; A&A’s proposed design)
What was saidIn the example meetingHow the email treats it
Your promisesTell them by Friday whether the data can be movedWritten as a promise
The client’s wishesStart next monthNot written as a promise
Ideas under considerationMoving the current dataNot written as a promise
Anything the notes give no basis forThe day the data would be movedMarked “unconfirmed”

Decide recording and 4 pre-send checks before you start

A&A perspective

Before the test, decide whether to use recordings and which four items to check before sending. If you decide mid-test, what AI receives changes from meeting to meeting, and the week’s numbers can no longer be compared.

A&A perspective

If you record, ask the client whether you may record and get their agreement. Also decide who can listen to the recording, where it is kept, and when it is deleted.

A&A perspective

You can also skip recording. Right after the meeting, jot down about three lines on your promises and the client’s wishes, and hand them to AI. Short input is enough as long as you can later confirm what you promised.

A&A perspective

If you are unsure which to choose, choose by the checking work each one leaves you. Without a recording you have less information, so you still have to check the promises column yourself. With a recording, you need the client’s agreement and rules for handling the file.

A&A perspective

The four pre-send checks are price, scope, deadline, and owner. Check that these match the meeting before you look at company names or phrasing.

The salesperson offers a polished draft; the owner checks amount, scope, deadline and owner one by one first.

A&A perspective

For price, check that the draft does not state a price the conversation never settled. For scope, check that work the client did not ask for has not slipped into your to-do list. For deadline, check that a “day we will confirm” has not turned into a “day we will do the work.” For owner, check that it names who contacts the client next and who looks into open questions.

A&A perspective

In the earlier example, a draft saying “We will move your data on Friday” gets both scope and deadline wrong for that meeting. You check these four before phrasing because such mismatches are read as company promises the moment the email arrives.

A&A perspective

Keep each sent email in the deal record exactly as it was sent. If it later conflicts with a quote, you can trace where the terms changed.

Count send delay and misses over one week

A&A perspective

Decide whether to keep going only after counting two things over one week of meetings: send delay and missed promises. The quality of the summary is not counted here.

A&A perspective

Send delay is the time from the end of the meeting to when the follow-up email goes out. Missed promises, or misses, are promises made in the meeting that did not make it into the email.

A&A perspective

You can estimate your current send delay by laying last month’s calendar next to your sent emails. Your current number of misses is usually not on record, so start counting in the test week.

Hypothetical example

For example, take a company with six sales meetings a week. For each meeting, log the end time, the send time, the number of promises made, and the number of promises missing from the email. Also log how long you spent fixing the draft.

From the sources

The Morgan Stanley announcement quotes an advisor who says that handing off the note-taking saves him about half an hour per meeting.

Morgan Stanley ↗

A&A perspective

But this is one person’s comment in the company’s own announcement, and it does not say how the time was measured. Check the time you save with your own log.

A&A perspective

After the week, if send delay is shorter than now and no promises were missed, it is worth continuing next month. If fixing takes long and send delay does not shrink, fix how you ask for the draft or how you take notes.

A&A perspective

One week will not tell you whether you are winning more deals. Meetings where you sent the follow-up quickly may simply be ones where the client was already keen. What one week decides is only whether the test is worth continuing.

A&A perspective

When you compare costs, line up the owner time saved, the time spent fixing drafts, and the tool’s monthly fee. Compare against a sales outsourcing or hiring budget only once those numbers are in.

Summary

The salesperson brings a follow-up draft; the owner separates promises from the client's wishes and checks four items.

A&A perspective

For sales follow-ups, AI’s job ends at drafting the email from your meeting notes.

A&A perspective

In the draft, have AI keep your promises, the client’s wishes, and ideas under consideration apart.

A&A perspective

Before testing, decide on recording and the four checks: price, scope, deadline, and owner.

A&A perspective

Over one week of meetings, count send delay and missed promises, then keep going or fix the setup.

Whether to bring AI into sales follow-ups comes down to the numbers from one week in which AI only drafts. Settle recording and the four pre-send checks before you start, and base the decision to continue on send delay and missed promises.

FAQ

Can I use AI drafts for meetings I don’t record?

Yes. Right after the meeting, jot down a few lines on your promises and the client’s wishes, and give them to AI. With less information than a recording, you still have to check the promises column yourself.

Can the same draft also assign work inside the company?

Make the client email and the internal tasks separately, from the same promises. This keeps internal circumstances out of the email to the client. Handing work over inside the company is covered in “From AI meeting notes to next tasks: 4 decisions before you pick a tool.”

Which tool should I start with?

You can start with a chat AI such as ChatGPT: give it your meeting notes and an instruction to list them in the three columns. Compare dedicated tools that go from a meeting recording to a draft only after you have seen the week’s numbers.

Sources & editorial note

Primary pages read for this article. Publication dates below belong to the sources; access dates record our research.

  1. Morgan Stanley Wealth Management Announces Latest Game-Changing Addition to Suite of GenAI Tools

    Morgan Stanley · 2024-06-26

    Accessed 2026-10-23
  2. Building effective agents

    Anthropic · 2024-12-19

    Accessed 2026-10-23

Editorial policy

← All articles