In short
- Meeting notes, email summaries and first drafts save real hours today, at roughly $30 to $36 per person a month.
- Everything else depends on whether your records and your ways of working are already tidy. Clean those first.
- Write a one-page rule about what staff may paste into an AI tool before you hand any of it out.
Every software company has added AI to its product name and every conference promises a transformation. Most of it is loud and disconnected from what happens in an office.
Here is what actually saves time in a Canadian business of 20 to 50 people, what it costs, and what to leave alone this year.
Every section below gives you the same three things: what the tool does in an ordinary office, what it costs per person a month, and what has to be true before it works. Where something is worth waiting on, that is said plainly. The groundwork it all rests on, and who keeps it running afterwards, is on the managed IT page.
One thing to understand before you buy anything
These tools speed up what you already have. Messy records produce messy answers faster. An unclear way of working becomes an unclear way of working at speed.
Two pieces of preparation pay for themselves before you spend a dollar on AI:
- Clean the contact list. Sales and marketing features are only as good as the records underneath them. A list full of duplicates gives you confident, wrong answers.
- Write down how the work is done. These tools help carry out a process. They cannot invent one. The businesses that get the most from them already knew how their own work flows.
Writing and reading documents
What works: Claude, ChatGPT, and Microsoft Copilot inside Word and Outlook.
Genuinely useful for first drafts of letters, proposals and reports. Boiling a long contract or tender down to the points that matter. Rewriting something so it reads more clearly. Turning rough meeting notes into a list of who does what.
Leave alone for anything that needs real expertise or original research, for writing that has to sound like your business, for legal, financial or medical advice, and for anything urgent you have not read through yourself.
Realistic gain: hours back every week. In a firm where partners spend six to eight hours a week producing documents, the first draft is the part that gets shorter.
Privacy in Canada: what your staff type into these tools may be processed and, on some plans, used to improve the model. For client-sensitive material, use a business subscription with proper terms. Microsoft Copilot for Microsoft 365 comes with an agreement that covers PIPEDA. Consumer ChatGPT does not.
Cost: Microsoft Copilot runs $30 to $36 USD per person a month. Claude for business runs $25 USD per person a month.
What we usually recommend: Copilot if you are already on Microsoft 365, and Claude alongside it for heavy analysis work.
Microsoft 365 Copilot, feature by feature
This deserves its own section, because it is the easiest path for most Canadian businesses already paying for Microsoft.
Teams. Meeting summaries and action items are the strongest feature in the whole product. If your team meets a lot, it removes most of the write-up that somebody used to do by hand after every meeting.
Outlook. Summarising a 50-message thread into the decisions takes half a minute. Drafting a reply from three bullet points takes two minutes instead of fifteen.
Excel. Good for people who need an answer out of a spreadsheet and struggle with the formula. A genuine gain for office administrators and operations staff.
Word. Good for a first draft from an outline and for summarising. Weaker when the wording has to be exactly right.
SharePoint. Asking questions of your own document library is promising, and it depends entirely on your library being organised. A messy SharePoint becomes a faster messy SharePoint.
What you need first: one of the business Microsoft 365 plans. Copilot costs extra on top, and Microsoft asks for a minimum of five people.
AI on your website and in your inbox
What works: AI chat on the website, suggested replies in the support inbox, and tone detection that flags an unhappy customer.
Genuinely useful for answering common questions at 9 pm, sending an enquiry to the right person, drafting support replies for a human to check, and flagging a frustrated customer for a call.
Leave alone for complicated sales conversations, for sensitive client messages where a wrong answer puts you on the hook, and for any part of winning a business client where the relationship is the point.
Realistic gain for a 20-person firm: faster first replies, and far fewer enquiries that sit unanswered overnight.
If you are regulated: a clinic or a law firm putting AI chat on the website has to be careful about what it says and what it collects. A chat window that collects health information carries the same duties as any other system that holds it. Check your obligations before it goes live.
Cost: roughly $1 USD for every conversation the chat actually settles on the common platforms. A simple chat window built out of off-the-shelf parts runs $200 to $500 a month.
Sales and marketing
What works: email personalisation and content suggestions inside your CRM, prospecting tools, and subject line testing.
Genuinely useful for drafting a run of follow-up emails, ranking enquiries by how interested they look, testing which subject line gets opened, and turning one long article into a month of social posts.
Leave alone for judging whether an enquiry is actually a fit, for building real relationships, and for anything that needs original research. People can tell the difference between an email a person wrote and one a machine wrote.
Canadian rule you cannot skip: Canada’s anti-spam law still applies to email a machine wrote. Sending without consent is still spam, and penalties reach $10 million per violation for an organisation. Make sure whatever you use manages consent properly.
Numbers and reporting
What works: Power BI with Copilot, the AI features in Tableau, and purpose-built setups for businesses that live in spreadsheets.
Genuinely useful for asking your own data a question in plain language, spotting anomalies in financial or operational figures, forecasting demand, and writing the commentary on a weekly report.
What it needs first: clean, consistent numbers in systems that talk to each other. These tools are impressive when the data is tidy. When the same figures sit in ten different shapes across five systems, that has to be sorted out before AI helps at all.
Cost: Power BI Pro is $10 USD per person a month and is included in some Microsoft plans. Tableau costs considerably more.
Four things to leave alone this year
AI screening of job applicants. Automated hiring decisions are under increasing scrutiny under provincial human rights codes. The legal exposure outweighs the time saved.
Software that makes business decisions on its own. The technology exists and it is not accurate enough for decisions that carry consequences. Let these tools advise, and keep the decision with a person.
AI-written code nobody reviews. A development team using an AI assistant with proper review is reasonable. A vendor selling you software that no human has read is a risk you are taking on their behalf.
AI answering for you in a regulated field. When a machine gives incorrect medical or legal information, the liability is yours. In regulated work, AI helps a person; it does not replace one.
A realistic first year
For a 20 to 50 person Canadian business:
Months 1 to 3. Tidy Microsoft 365 and SharePoint. Give Teams Copilot to leadership and the people who live in meetings. Write a one-page rule about what staff may and may not paste into an AI tool. A practice holding patient records has a firmer version of that rule to write, which is covered in what does an IT company do for a dental office.
Months 4 to 6. Extend Copilot to Outlook for document-heavy roles. Put AI chat on the website covering your common questions. Connect email drafting to the CRM for the sales team.
Months 7 to 9. Pick three to five pieces of work where AI can handle routing or a first pass. Build the reporting dashboards. Train staff on asking these tools better questions.
Months 10 to 12. Turn AI analysis on your operations and finance numbers. Widen the customer-facing side based on what the earlier months showed. Review what it actually returned and plan year two.
Budget: $15,000 to $35,000 for the first year of setup and training beyond your existing Microsoft licences, and $8,000 to $20,000 a year in ongoing tool costs after that.
How we run this, starting with an assessment before any purchase, is on the automation page. Businesses on the west coast can see what we cover locally on the Vancouver page.
Ask us what is ready and what is not
Leave your name and number on the contact page and a real person calls you back the same business day, usually within an hour. We will tell you plainly which of these is ready for your business and which needs groundwork first. The free infrastructure audit covers the data and process side that decides the answer.
itopsi builds automation and AI workflows for Canadian businesses from a Toronto base.
