The typing was never the slow part. Generating the email from a structured meeting note is what gets it sent on the day of the call.
Key takeaways
You write follow-up emails faster by generating them from a structured record of the call rather than from memory. The typing was never the slow part. The slow part is reconstructing, three hours and two meetings later, what was actually agreed, what you owe, and what you are waiting for — which is why the follow-up gets postponed to the evening and then to never.
Which means the fix is upstream. If the meeting produced a structured note with decisions, commitments and open items, the follow-up email is a formatting exercise. If it did not, no email generator will save you, because the model does not know what happened either.
Five elements, in this order, and almost nothing else.
What we agreed. Stated as decisions, not as a recap of the discussion. What I owe you, and when. Named deliverable, named date. What I need from you, and when. Same. What is still open. The questions nobody answered, in writing, so they cannot later be described as never raised. What happens next. The next contact, and who initiates it.
The fourth element is the one firms drop and the one that earns its keep. An open item recorded in an email you sent is a very different artefact from an open item nobody wrote down.
The practical workflow is two steps. The meeting note is produced first, in a fixed structure. The follow-up email is then generated from that note with a short instruction — write to the client in the second person, use only what is in the note, keep it under 200 words, end with the single thing you need from them.
Two constraints matter. Use only what is in the note, or the model will smooth over gaps with plausible-sounding commitments you never made. And keep it short, because the follow-up that gets read is the one that fits on a phone screen.
Nothing here is exotic. The gain is that the email exists on the day of the call, with the deadlines still correct, rather than four days later with the dates quietly rounded.
The CRTC sets out the four main requirements for commercial electronic messages: prior consent, express or implied; identification and contact information; and a working unsubscribe mechanism. A follow-up email after a business call is frequently a commercial electronic message, and the fact that the recipient asked you to send it does not remove every obligation.
The statute is more precise than the summary. Under section 6 of the Act, sending a commercial electronic message is prohibited unless the recipient has consented and the message complies with the content requirements — identification of the sender, contact information, and an unsubscribe mechanism. Subsection 6(6) then removes the consent requirement for a message that solely provides a quote or estimate that the recipient requested, or that solely facilitates, completes or confirms a commercial transaction the recipient previously agreed to enter into.
Read that carefully, because it is routinely misread. The exception lifts the consent requirement. It does not lift the content requirements. A requested proposal still needs sender identification, contact information and an unsubscribe mechanism, and the contact information must stay valid for at least 60 days after the message is sent. An unsubscribe request must be given effect without delay and in any event no later than 10 business days.
The words “solely” is where automation goes wrong. A follow-up that provides the requested quote is one thing; the same email with a paragraph about your three other service lines is another, and a template that adds a promotional footer to every message has changed the character of every message. Our sister firm covers the unsubscribe mechanism requirements and email marketing compliance in more depth.
Anything containing a number you have not checked. Anything committing the firm to a date. Anything characterising a client’s position, their counterparty, or their prospects. Anything that would be the first written record of bad news.
The reliable division is that AI drafts and a person sends. It sounds like a small distinction and it is not: a queue of drafts reviewed in one sitting is fast, and an outbound stream nobody reads is a liability that scales.
A boutique consultancy ran six to eight discovery calls a week. Follow-ups went out within a day for about half of them and within a week for the rest, which correlated uncomfortably well with which prospects went quiet.
The change: the call produces a structured note. A generated follow-up lands in the consultant’s drafts within ten minutes, containing the five elements above and nothing promotional. The consultant edits and sends before leaving the call block — typically two minutes. Anything with a fee figure in it is checked against the rate card by a second person before sending.
Measure the share of calls followed up within four hours, and the share of follow-ups that get a reply. The second is the one that tells you whether the emails are actually any good. Faster bad emails are still bad emails.
Does a follow-up after a meeting need an unsubscribe link?
Assess the message, not the habit. The consent exceptions in subsection 6(6) do not remove the content requirements in subsection 6(2), so identification, contact information and an unsubscribe mechanism should be your default in any message with a commercial purpose.
Can AI send the follow-up automatically?
It can, and for a professional firm it should not. Drafting is the time saving; sending is the risk. Keep the human between them.
How do I stop the drafts sounding generic?
Feed it the structured note and forbid it from adding anything not in the note. Generic output is almost always a symptom of a thin input, not a weak model.
A 30-minute call is enough to tell you whether AI pays for itself here.
