Silence comes in three kinds and only one responds to a reminder — and under CASL the consent you are relying on has an expiry date.
Key takeaways
What you send when a client goes quiet depends entirely on which kind of quiet it is, and firms lose money by sending the same nudge to all three. There is the client who is waiting on you and does not realise it. There is the client stuck on an internal decision. And there is the client who has decided something and has not said so.
Only the second responds to a reminder. The first needs a correction, the third needs a closing message that makes it easy to say no. A generated “just circling back” sent to all three is why follow-up sequences stop working after the second message.
There is also a clock running that most firms have never looked at, because under Canada’s anti-spam legislation the implied consent you rely on to email a quiet contact expires on fixed timelines.
Rung one, day three: the correction. Restate what you are waiting for, in one line, with the consequence. “We are holding the draft until we have the shareholder register — the 14th delivery date moves with it.” No apology, no chasing tone. Most silence ends here because most silence is a misunderstanding about who has the ball.
Rung two, day eight: the cost of delay. Name what the delay is doing — a filing window, a rate, a resourcing slot. Specific consequences get replies; general urgency does not.
Rung three, day fifteen: the decision with a default. Offer two options and state what happens if neither is chosen. “We can proceed on the reduced scope, or hold the file until March. If we do not hear by the 28th we will hold it.” This is the message that converts silence into information.
Rung four: the close. A short, warm note closing the file, with the door open. It generates more replies than rungs two and three combined, because it removes the awkwardness of having gone quiet.
AI is genuinely useful across all four: drafting the variants, holding the schedule, tailoring each rung to what the file actually says. What it must not do is decide which rung a client is on, or decide to write off, terminate or escalate. Those are commercial judgments with consequences, and they belong to a person.
If you are emailing someone who has not responded, ask what your consent rests on. Where it rests on implied consent, the Act is specific. Implied consent arises from an existing business relationship, which subsection 10(10) defines to include the purchase or lease of a product, goods, a service, land or an interest in land within the two-year period immediately before the message was sent; a written contract currently in existence or expired within that same two-year period; and an inquiry or application made by the recipient within the six-month period immediately before the message was sent.
The six-month limb is the one that catches agencies and consultancies. A prospect who enquired in January and never replied is outside that limb by August. The relationship did not feel like it changed; the legal basis for emailing them did.
The CRTC’s summary of the four main requirements still applies to every message you do send: consent, express or implied; identification and contact information; and a working unsubscribe mechanism. Monetary penalties are available, and directors who authorised or participated in a violation can be held liable. Our sister firm sets out implied versus express consent and the limits of the business-to-business exemption, which is narrower than most people assume.
The design consequence for an automated ladder: a sequence that keeps firing for months at a contact who never replied is not only ineffective, it may be running past the basis it started on. Build the expiry into the sequence rather than into someone’s memory.
Drafting the variants. Four rungs across an agency’s live client list is a lot of writing, and the difference between a good rung two and a bad one is specificity to that file. Generation from the file record produces specificity; a template does not.
Detecting a reply that is not an answer. “Thanks, will look at this” is silence wearing a hat. Flagging replies that do not contain a decision or a date is a genuinely useful classification task, and it keeps the ladder running when a human would have marked the thread as handled.
Holding the schedule. The reason ladders fail is that rung three never gets sent. Software does not get busy.
Not deciding. Whether to stop work, whether to charge, whether to close the file — a person decides, and the sequence pauses until they do.
A quiet client with an unpaid invoice is a different problem, and the escalation ladder above is the wrong instrument for it. The commercial and legal steps are separate, and worth understanding before an automated sequence says something that limits your options — our sister firm covers suing a customer over an unpaid invoice.
Practical rule: the moment a thread is about money owed rather than work outstanding, it leaves the automated ladder entirely and goes to a named person.
A digital agency had roughly a dozen threads at any time waiting on client sign-off. Follow-up was whoever remembered, which meant the loudest clients got chased and the quiet ones drifted for six weeks.
The change: every outbound that requires a client action starts a four-rung ladder with the dates above. Drafts are generated from the project record and reviewed in a single daily batch. Replies without a decision or a date are flagged rather than closed. Any thread involving payment is removed from the ladder automatically and assigned to the account director.
Measure the median days to sign-off and the proportion of threads that reach rung four. If a lot of threads reach rung four, that is not a failure of the sequence — it is the sequence telling you which work was never going to happen, six weeks earlier than you would otherwise have found out.
How many follow-ups is too many?
The ladder above ends in four because the fourth message is a close, not a chase. Sequences that keep nudging indefinitely damage the relationship and may outlive the consent basis they started on.
Can I keep emailing a prospect who enquired last year?
Check what your consent rests on. The inquiry limb of implied consent runs for six months from the inquiry; a purchase or written contract limb runs for two years. Express consent is a different matter.
Should the closing message be automated?
Draft it automatically, send it deliberately. It is the message most likely to get a reply and the one most likely to be read as a final impression of the firm.
A 30-minute call is enough to tell you whether AI pays for itself here.
