Treadstone Associates
Article · Business development

AI to re-engage dormant clients

The useful work is deciding who to contact and why. CASL decides whether you may, and the two-year clock is the part firms get wrong.

Treadstone Associates · Updated 2026

Key takeaways

  • • Segmentation is the job: which former clients have a live reason to hear from you this quarter, and what that reason is.
  • • Implied consent from an existing business relationship runs two years from the purchase — and only six months from a mere enquiry.
  • • Every commercial electronic message needs a working unsubscribe that stays valid for at least 60 days and is actioned without delay.
  • • Client confidentiality does not lapse because a file closed; a re-engagement campaign must not reveal who your clients are.

The short answer

Use AI for the segmentation and the drafting, and check the consent basis before either. Most firms treat reactivation as a writing problem — what do we say to people who have gone quiet — when it is really a selection problem. A model that can read your practice-management data will tell you which former clients have a live, specific reason to hear from you now: an annual filing that has stopped, a matter type that typically recurs on a cycle, a corporate client whose registered filings suggest a transaction, a file that closed with a next step nobody followed up.

That is a far better foundation for a campaign than a generic ‘we miss you’ email, and it happens to be the version Canadian anti-spam law is most comfortable with, because a message with a specific reason usually sits inside a relationship rather than outside one.

The consent clock most firms get wrong

Canada’s anti-spam legislation is enforced by the CRTC, whose spam and malware pages carry the frequently asked questions, compliance tips and the Act, regulations and guidelines. The provision that matters for reactivation is the definition of an existing business relationship, and it is worth reading in the statute rather than in a summary.

Under Canada's anti-spam legislation, implied consent can arise from an existing business relationship, and subsection 10(10) defines that relationship as arising from, among other things, the purchase or lease of a product, goods, a service, land or an interest in land within the two-year period immediately before the day on which the message was sent; a written contract currently in existence or expired within that same period; or an enquiry or application made within the six-month period immediately before the message was sent.

Read that against a real book of clients and the consequences are immediate. A client whose last engagement completed thirty months ago is not covered by the purchase limb. A prospect who enquired last year and never instructed you is well outside the six-month limb. And a client on a standing written engagement is in a different position from one who bought a discrete piece of work — which is one more reason for the annual engagement letter to exist. Our sister firm’s explainer on implied versus express consent under CASL sets out the distinction in business terms.

The mechanics are prescribed too. Section 11 of the same Act requires the unsubscribe mechanism to let the recipient opt out at no cost using the same electronic means the message came by, to specify an address or web page for doing so, to keep that address or page valid for a minimum of sixty days after the message was sent, and to give effect to an unsubscribe request without delay. A reactivation campaign run out of a spreadsheet rarely satisfies all of that. One run out of a proper platform usually does, which is an argument for the platform rather than for the model.

Where AI genuinely helps

Four jobs. Segmentation, which is the one worth doing first: grouping dormant clients by why they went quiet and whether there is a plausible reason to make contact. Trigger detection: reading your own records for the signals that a former client is likely to need something — a filing cycle, a renewal date, an incorporation anniversary, an expired agreement in your own document set. Drafting: producing a message that refers to the actual matter you handled rather than to nothing in particular. And suppression: cross-checking the list against unsubscribes, conflicts, deceased or wound-up entities, and clients you deliberately disengaged.

That last one is the least glamorous and the most valuable. The characteristic failure of an automated reactivation campaign is not a poor message; it is a good message sent to a client you asked to leave, or to the other side of a matter, or to someone who unsubscribed two years ago in a different system.

Confidentiality does not expire when the file closes

For a law firm in Ontario, rule 3.3-1 of the Rules of Professional Conduct requires a lawyer at all times to hold in strict confidence all information concerning the business and affairs of the client acquired in the course of the professional relationship. ‘At all times’ includes after the retainer ends. That has direct consequences for a campaign: the fact that a person was your client is itself confidential information, so anything that discloses your client list — a shared mailing tool with loose access, a subject line that identifies a matter type, a testimonial request that names the file — is a problem regardless of how well it converts.

The same logic applies to feeding a client list into an external model for segmentation. The Office of the Privacy Commissioner’s principles for responsible, trustworthy and privacy-protective generative AI ask organisations to establish that the use of a generative system is necessary and proportionate, to use de-identified or synthetic data where personal information is not required, and to safeguard information commensurate with its sensitivity. Segmentation can very often be done on de-identified attributes — matter type, close date, value band — and the names reattached inside your own system afterwards.

Worked example: a practice with four hundred closed files

Illustrative. A firm has a long tail of former clients and no systematic contact with any of them. The instinct is to email everyone.

The disciplined version starts by splitting the list on the consent question: clients whose last purchase or written contract falls inside the statutory window, clients whose implied consent has lapsed, and clients who gave express consent at some point that the firm can actually evidence. Only the first and third groups are candidates for an email campaign; the second group is a phone call, a letter, or nothing.

Within the eligible group, segmentation is by reason rather than by recency: clients with an annual obligation the firm used to handle, clients whose closed matter had a natural follow-on, and clients with no identifiable reason at all. The last group gets nothing, because a message with no reason is the one most likely to be marked as spam and least likely to earn work. Drafting is per segment, reviewed by a partner, and sent through a platform that manages unsubscribes properly.

The result is a much smaller campaign than the firm intended, sent to people with a reason to hear from it. That is the trade every time: precision over volume.

Questions we get asked

Does the two-year clock restart if they open our newsletter?
The statutory limbs are tied to purchases, contracts, memberships and enquiries, not to engagement metrics. Read the definition rather than assuming an open counts.

Is a newsletter a commercial electronic message?
Assume it is if it encourages participation in a commercial activity, and build consent and unsubscribe handling accordingly. The CRTC’s own FAQ and compliance-tips pages are the place to check the edge cases for your content.

Can we let AI send the messages automatically?
You can automate delivery. Someone should still approve the list and the copy, because the consequences of an error here are regulatory and reputational rather than merely awkward.

What about clients we let go deliberately?
Keep a permanent suppression flag on them. If disengagement is the live issue, see ending an engagement for cause.

See where AI pays off first in your firm.

A 30-minute call is enough to tell you whether AI pays for itself here.