A model builds the plan, the reading list and the first-week schedule from documents you already have. The parts Ontario law makes mandatory are not on that list, and they have dates attached.
Key takeaways
Onboarding at a professional firm is two jobs wearing one name. One is a compliance job with fixed content and fixed deadlines. The other is a ramp-up job — getting somebody productive on client work — which is mostly writing, sequencing and answering questions. AI is useful for the second and irrelevant to the first.
Firms that conflate them end up with a beautifully drafted induction pack and a missing statutory document. Separate them on paper before you automate either.
The ministry’s guidance on mandatory information for employees sets these out plainly:
Separately, an employer with 25 or more employees on 1 January must have a written policy on electronic monitoring and provide a copy to its employees within the specified timeframes — which in practice means a new starter needs it too. Treadstone Law’s note on the ESA poster obligation covers the posting-versus-providing distinction that trips up small firms.
A second, separate duty attaches the moment somebody starts, and it applies to almost every firm in this hub regardless of size. Ontario’s accessibility regulation defines an “obligated organization” to include any organization with as few as one employee in the province, and every obligated organization must ensure that employees are trained on the requirements of the accessibility standards and on the Human Rights Code as it pertains to persons with disabilities — training that must happen as soon as practicable after they start, not whenever the firm gets around to it. O. Reg. 191/11, s. 7 is the source; Treadstone Law’s guide to AODA employment obligations covers what a small professional office actually has to do about it. The duty does not end at onboarding either: s. 7(4) requires training on any changes to the organization’s accessibility policies to be provided on an ongoing basis, so a new hire who joins mid-cycle is not the last person this needs revisiting for. The record-keeping duty attached to that training runs only to government, designated public sector bodies and large organizations of 50 or more employees — a small firm must still deliver the training, just without a matching statutory duty to paper it, which is exactly the kind of gap a model can flag when it drafts the onboarding checklist.
The pattern is the same one that runs through this hub: the tool assembles and drafts, a person reviews and owns. It is the same discipline described in our note on building a sign-off workflow so AI drafts never go out unchecked.
Terms of employment are not onboarding content. If the offer, the contract or the compensation changes after somebody has started, that is a contract question with real consequences — amending an employment contract after hiring in Ontario generally requires fresh consideration, and a mid-onboarding “here is the updated version” email is exactly how firms create unenforceable terms.
Nor is a probationary period a free pass. How a probationary period actually works in Ontario is worth ten minutes of a practice manager’s time before the first review meeting is scheduled.
And no model should be anywhere near a hiring decision. Screening, ranking and selection are decisions a person makes and is accountable for; the tool has no role there.
A new hire at an accounting or law practice gets access to client files in the first week. The briefing they get has to be specific: which systems hold client information, what may be copied where, and what may never be pasted into an external tool. A general statement about confidentiality does not survive contact with a deadline. The Privacy Commissioner’s principles for generative AI give the framing — collection, use and disclosure limited to what the identified purpose needs — and the firm turns that into a list of named systems.
A 14-person Ontario practice hires a staff accountant to start on a Monday. The compliance track runs on its own timetable: the written employment information goes out before the first day, the poster is issued in the first week, and the electronic monitoring policy is provided with the handbook. None of it touches an AI tool.
The ramp-up track uses one. The job description, the three procedures the new hire will use most, and the manager’s notes go into a model, which drafts a 30/60/90 plan and a first-week schedule. The manager cuts it by a third — the draft was optimistic about how much a person absorbs in week one — and adds the two client meetings the new hire will sit in on. During the first fortnight the new starter asks procedural questions of the library rather than interrupting a manager mid-file, and every answer names the document it came from.
At day 30 the manager compares the plan to what actually happened. The gap is the next revision of both the plan and the procedures behind it.
Do we have to pay for time spent in training?
As a general rule in Ontario, mandatory training time is work time. Treadstone Law’s answer on paid training time under the ESA sets out the position before you design a programme that assumes otherwise.
Can AI screen candidates for us?
Hiring is a decision a person makes and answers for. Keep the tool on the post-offer side of the line, where it drafts plans and answers questions rather than ranking people.
Do we reissue the poster when a new version comes out?
No. Employers are only required to provide each employee with the poster once, per the ministry’s guidance.
A 30-minute call is enough to tell you whether AI pays for itself here.
