Applied AI North

A sample AI Roadmap.

This is what you get at the end of the two-week AI Roadmap, written for a fictional property manager in Mississauga. Real method, real Ontario rules, illustrative numbers.

Sample. Tarnwood Residential is a fictional company. Every figure is illustrative, built from published benchmarks and the assumptions listed at the end. In a real Roadmap we measure these numbers in your business instead. This sample covers more opportunities than a typical Roadmap, so you can see every kind of verdict.

Prepared for

Tarnwood Residential Ltd.

Owner-run since 1979 · Mississauga, Ontario

Homes
1,450 apartments in 16 buildings, Mississauga and Brampton
People
42: 16 in the office, 26 on site
Tools
Yardi Breeze for rent and maintenance, Microsoft 365 for email
Their question
“Our office is buried in maintenance requests and email. Where would AI actually help, and what should we leave alone?”

The answer, on one page.

Do two things now, try one next, fix two without AI, and keep AI away from two decisions that carry legal risk.

Do now

Do next

Not yet

Don’t

Together, the “do now” items cost CA$17,000 one time (including this Roadmap), plus about CA$125 a month in AI usage fees. They give the office back roughly 10 to 18 hours a week, and pay back in about 7 months (13 if the cautious numbers hold).

Where the office’s time goes today.

About 65 hours a week of office time goes to five kinds of repetitive work. That’s where we looked first.

Estimated office hours per week, by task
TaskHours per week
Sorting maintenance requests and updating tenants17.9 h
Answering routine tenant email16.7 h
Looking for documents and building information15.0 h
Replying to people asking about vacant units8.3 h
Chasing late rent6.7 h

Illustrative. Built from an industry benchmark of 0.4 maintenance requests per apartment per month and our time-per-task assumptions (see the end). In a real Roadmap, your team counts these for two weeks.

Each recommendation, and why.

Hours are a range: cautious to expected. Costs use our published prices.

Team workshop Do now, first

Today: the 12 office staff who write the most use AI unevenly, and some paste tenant emails into personal ChatGPT accounts.

What changes: a half-day session on their real tasks (notices to all tenants, summarising long threads, plain-language replies, Excel) and on what never goes into an AI tool. Staff move to Microsoft 365 Copilot Chat, which comes with Tarnwood’s Microsoft 365 plan at no extra cost and keeps company data protected.

Hours back
3–6 a week
One-time
CA$1,500
Running cost
None (no new licence)
Risk
Low. It reduces the risk of tenant data in personal accounts.

Why first: fastest payback, lowest risk, and every later step goes better when staff are comfortable with the tools.

Drafted maintenance replies and tenant email Do now

Today: about 580 maintenance requests a month arrive by portal, email and phone. For each one, someone reads it, chases missing details, sets the priority, assigns a superintendent or vendor, writes the 24-hour entry notice and updates the tenant. Meanwhile about 50 routine emails a day (parking, fobs, receipts, move-out steps) wait for an administrator.

A normal Tuesday, after

A tenant reports a leaking tap at 9:12. By 9:13 the maintenance coordinator sees a ready draft: category “plumbing, routine”; assign “Building 7 superintendent”; an entry notice for Thursday between 10 and 2 (sent by email, because this tenant agreed in writing to email notices); and a short reply to the tenant. The coordinator moves the time to the afternoon, clicks approve, and updates the request in Yardi Breeze. Anything that mentions gas, flooding, no heat or fire skips the AI and goes straight to a person.

Hours back
7–12 a week (maintenance 4.5–7, email 3–5)
One-time
CA$13,000: two separate fixed-price builds, one after the other (maintenance CA$7,500, then email CA$5,500)
Running cost
About CA$125/month in AI usage, billed to Tarnwood’s own provider account at cost
Risk
Medium-low. Nothing reaches a tenant without a person approving it. Some topics are never drafted: arrears, notices, eviction, accommodation requests, harassment, anything from a lawyer. The AI provider may process text outside Canada, so Tarnwood’s privacy notice says so. An off switch returns everything to today’s process.

Why: it’s the biggest block of repetitive writing, the drafts are easy to check, and the volume grows with every building Tarnwood adds.

Fast first replies on vacant units Do next

Today: about 360 enquiries a month reach two leasing agents. Each takes about 6 minutes: answer the basics, book a showing, send a reminder.

What changes: AI answers unit facts from the live listing and offers showing times. An agent does every showing and every decision. Replies never ask about family status, receipt of public assistance or anything else the Human Rights Code protects.

Hours back
2.5–4 a week, plus fewer empty days between tenants (not counted)
One-time
CA$10,000 (a customer-facing assistant on one channel)
Running cost
About CA$40/month in AI usage
Risk
Medium: fair-housing wording, and contact details in a new system.

Why next, not now: prove the approve-every-draft pattern on maintenance first, then decide in week 12 with real numbers.

Searching company documents Not yet

Today: staff hunt through paper binders, scanned PDFs and inboxes for boiler models, warranties, vendor rates and past work on a unit.

Instead, first: one administrator spends 3–4 hours a week for 8 weeks scanning the binders, filing each building’s documents in one consistent folder structure, and keeping a single vendor list. Much of the hunting disappears without AI.

If built today
3–5 h/week back for about CA$9,000, plus about CA$145/month in AI usage. That pays back in about 16 to 33 months, and slower while the documents stay messy.
Look again
In 6 months. With clean documents it costs less and works better.

Late-rent reminders Not yet

Today: about 116 accounts are late each month, and the accounts clerk emails and calls each one.

Instead: a sign-up campaign for pre-authorized debit, plus the standard payment emails already in Yardi Breeze. The message is the same every time, so AI adds risk (tone, wrong amounts) and little value.

Hours back
1.3–2.7 a week, from the process fix alone
Cost
Nothing from us

Eviction notices (N4, N5) Don’t

These notices start the process of ending someone’s tenancy. The rules just changed: for N4s given on or after September 21, 2026, the minimum notice period fell from 14 to 7 days, and the Landlord and Tenant Board replaced several forms. A wrong date or an outdated form can make a notice invalid.

Keep it human: the property manager decides and fills in the current form from the Board’s website, and a second person checks the dates. Automation may produce a weekly list of who is behind, for a person to review. It never decides or drafts.

Ranking rental applicants Don’t

Ontario’s Human Rights Code limits how landlords may screen tenants, and tribunals have found rigid income rules discriminatory. A ranking model can quietly encode exactly that, and can’t explain itself. Applications also hold the most sensitive data the company has: credit reports, pay stubs, ID.

What’s fine: a simple checklist of which documents are missing, and the standard credit report, read by a person.

What it costs, and when it pays back.

Payback is shown as a range, from expected to cautious. The workshop pays back within a few months. The build pays back in under a year on the expected numbers, and in about 15 months on the cautious ones.

Payback in months, expected to cautious
ItemOne-timePer monthPayback in months, expected to cautious
Team workshopCA$1,500none2–4 months
Maintenance + email buildCA$13,000~CA$1258–15 months
Everything above + this RoadmapCA$17,000~CA$1257–13 months

Hours back are capacity, not cash. They turn into money only if Tarnwood takes on more buildings without hiring, avoids overtime or a planned hire (an extra administrator costs roughly CA$70,000 a year), or answers tenants faster and keeps them longer. That last effect isn’t counted here.

We value staff time at Job Bank’s Toronto median wages plus 25% for payroll costs: about CA$33–36 an hour.

The first 90 days.

What we do, what the owner decides, and what the team does. Nothing goes live until it has run in the background and been checked.

  1. Weeks 1–2

    We
    Run the two-week Roadmap and deliver this plan with a walkthrough.
    Owner
    Approves the budget, names an internal lead, and confirms in writing that eviction notices and applicant decisions stay with people. Asks whoever handles privacy to update the privacy notice before anything goes live.
    Team
    Counts requests and emails for two weeks to set the baseline.

  2. Weeks 3–7

    We
    Run the workshop in week 3 (it ends with a one-page AI use rule), then build the maintenance drafts: weeks 4–5 in the background (drafts only, staff compare), weeks 6–7 live with approval. We test the emergency bypass and the off switch.
    Owner
    Approves the answers AI may draft from, and the topics it must never touch. 30 minutes a week.
    Team
    Attends the workshop and switches to Copilot Chat. The coordinator spends about 2 hours a week reviewing drafts. Document clean-up starts.

  3. Weeks 8–12

    We
    Build the tenant-email drafts the same way (background first, then live), send a short weekly report (drafts accepted, edited, rejected; hours against baseline), and close with a 90-day review and a go/no-go on leasing replies.
    Owner
    At 90 days, keeps, adjusts or switches off; decides on leasing replies; decides whether AI Office Hours (CA$750 a month) is worth it.
    Team
    Keeps approving, and repeats the two-week count so the change is measured, not guessed. The email drafts have only just gone live, so the team counts them again a month later.

Funding that may help.

We don’t administer grants. Here’s what exists and how it fits this plan; check with the program and your accountant.

Ontario Job Grant
Not for this workshop. The 2026 rules only fund training delivered by colleges, career colleges, union training centres, sector councils and similar providers.
BDC LIFT
A loan, not a grant, from CA$25,000, and it requires a BDC digital plan first. This plan costs less than the smallest loan, so it doesn’t fit. Worth a call if bigger AI spending is planned.
Productivity Super-Deduction
Now law. Ask your accountant whether any part of the build counts as a capital asset; much of it may simply be a deductible expense.
FedDev Ontario AI adoption
A repayable contribution of CA$125,000 to CA$5 million, covering up to half the cost, so it suits projects of roughly CA$250,000 and up. Not this one.

How we got these numbers.

Every figure is either sourced or a stated assumption. In a real Roadmap, the assumptions are what we measure.

Assumptions
  • Maintenance volume: 0.4 requests per apartment per month (industry benchmark) × 1,450 apartments ≈ 580 a month.
  • Time per task (not sourced; measured in a real Roadmap): 8 minutes per maintenance request, 4 minutes per routine email, 6 minutes per leasing enquiry, 15 minutes per late account, 1.25 hours a week spent looking for information by each of the 12 office staff who write the most.
  • Workshop: each of the 12 staff saves 15–30 minutes a week. Document clean-up: one administrator, 3–4 hours a week for 8 weeks.
  • Volumes: about 50 routine tenant emails per business day, 10% of tenants moving out a year, 30 enquiries per vacancy, 8% of accounts late in a month.
  • Share of the work AI drafting removes, with a person approving: 25–40% for maintenance, 17–30% of inbox time. This sits between two published studies (15% and 40%).
  • Staff cost: Job Bank Toronto median wages (2023–24 data) × 1.25 for payroll costs: about CA$33 an hour for an administrative assistant and CA$36 for a property administrator (CA$35 where the work spans both); 48 working weeks a year.
  • Payback in months: one-time cost ÷ (hours back a week × that hourly cost × 48 − a year of AI usage) × 12.
  • AI usage: current published prices for a mid-size model at US-region rates, doubled as a safety margin, at CA$1.42 per US dollar (October 2026). Software is included in our prices; AI usage is billed to the client’s own provider account at cost.
Sources

All accessed October 25, 2026.

Want one for your business?

The AI Roadmap takes two weeks and costs CA$2,500, fixed. You keep the plan, whether you build it with us, someone else, or no one.

Book a free intro call About the AI Roadmap

A free 30-minute call. If it’s a fit, you get a scope and a fixed price by the next morning.