Client communications: voice, pre-reads, and difficult news

The weekly status note is four lines: done, next, risks, asks. Not a narrative, not a mood - a gauge the client can read in thirty seconds. The steering pre-read is the b5 pyramid in miniature: the answer in the first sentence, support beneath, the ask explicit - sent early enough that the meeting argues with the recommendation instead of hearing it for the first time. A pre-read that buries the ask wastes the committee. The meeting follow-up lands within 24 hours and contains exactly three things: decisions made, owners named, dates attached. It is the version of the meeting everyone works from afterward.

AI's role differs per text, and the division is the same one this whole course teaches:

Notice the right column never changes character: judgment, politics, accountability. Feed AI your actual bullets and it produces a strong skeleton in seconds. Ask it to invent the content and it produces confident filler - which a client can smell by week 2.

A status note that arrives every Monday at the same hour builds a kind of trust no single brilliant email can: the client stops wondering what is happening, because the gauge always reads. Miss two Mondays and every silence becomes a signal. The pre-read's clock matters for the opposite reason - sent two days ahead, it converts meeting time from information transfer to decision making; sent the night before, it converts the committee into an audience. And the follow-up's 24-hour window exists because memory of a meeting decays fastest in the first day - the first written version becomes the truth. Put the three clocks in your calendar as recurring blocks; the texts then write themselves on schedule.

You know the AI default email voice on sight: "I hope this email finds you well... I wanted to circle back and touch base regarding our previous discussion..." It is recognizable slop - and the recipient's own AI probably wrote half the mail in their inbox in the same register, which is precisely why it signals nothing. The b5 scorer would price it instantly: no answer in line one, zero numbers, weasels everywhere, and not a single client anchor.

Voice, defined operationally: your firm's register (how direct, how formal, how warm Northbridge sounds) plus the client's vocabulary. Cadence's team says "seats" and "pilots" - so you write in their nouns, not in "user licenses" and "trial deployments". Writing in the client's own vocabulary is the email version of the scorer's client-anchor measure: it proves a specific human wrote this for a specific reader.

The fix is the same division as b6: feed your actual bullets in, let AI draft the structure, edit the draft into your register, and never send unread. The edit pass is not polish - it is where the voice enters. A sent email is a signed deliverable with your name at the bottom.

The cheapest voice check ever devised: read the draft aloud, and if you would not say the sentence to the client's face, delete it. "I hope this email finds you well" fails instantly - nobody has ever said that to a face. "Per my previous correspondence" fails. "The $30 tier is the answer, and here are the three numbers" passes, because you would say exactly that across a table. The test works because your speaking voice is the register the client actually knows you by; email that matches it reads as you. Run it on every AI draft before the send button - it takes thirty seconds and catches what no grammar check can.

The discipline: bad news never travels alone. It travels with (1) the evidence - the retention math laid out so the reader can check it, not a verdict to take on faith; (2) the alternative - the $30 tier, so the email closes a door and opens one in the same breath; and (3) the respect of directness - weasel cushioning like "some challenges have emerged around the proposed tier" is not kindness, it is fog, and the client pays you specifically to see through fog. "The $15 tier loses money at current churn" is more respectful, because it treats the reader as someone who can hear it.

AI's proper role here is rehearsal partner, never sender. Ask for three versions - too soft, too blunt, and right - and read all three: the spread teaches you where the line sits. Then role-play the reply: "respond as a CEO who has championed this tier for a year." You will meet Thursday's objections on Tuesday. The final draft is yours, sent under your name, ideally trailed by a phone call. One confidentiality note: the rehearsal prompt uses anonymized numbers and no client names - session b8 makes that reflex non-negotiable.