Article
Beyond the chatbot demos — three specific AI workflows we've seen quietly earn their keep in real Canadian SMBs, with honest numbers and the gotchas nobody mentions in the keynote.
This is a placeholder post — replace the body with the real article before launch. Outline below so you can see where it’s going.
Power Automate + AI Builder, sorting an inbound shared mailbox by intent (quote request / support / spam / billing). Quick to build, runs on existing licences for most M365 plans, and saves a real chunk of someone’s day.
The gotcha: training data. Without 50–100 real examples per category, the model picks up your last week’s noise instead of the underlying pattern. Most teams have the data; nobody’s labelled it.
Internal-use Azure OpenAI deployment, fed with your past proposals as grounding context, drafting a first-cut response from a project brief. Three to five times faster on first draft. Reviewer still has to do the work of being right.
The gotcha: the legal/insurance language. The model will happily make up clauses that read fluent and aren’t. Treat this as a first-draft tool, not a publishing tool.
Custom Power Automate flow with AI Builder reading uploaded SharePoint documents and applying metadata — client name, project, document type — automatically. Removes the “tagging tax” most SMBs skip until they need to find something.
The gotcha: PIPEDA awareness. If you’re tagging documents that contain personal information, the model needs to be a deployment that doesn’t train on your data. We default to Azure OpenAI with the no-training switch on; consumer ChatGPT is not the right tool for this.
— Continue reading
Copilot is real, useful, and — at roughly $40 CAD/user/month on top of M365 — easy to overspend on. A practical look at where it earns its keep and where you should hold off.
How Microsoft 365 plans line up against the workflows of a typical Canadian SMB, what they cost in Canadian dollars, and the licence mistakes most small businesses make.
How to configure SharePoint Online so a Canadian small business can responsibly handle personal information under PIPEDA, without enterprise tooling or enterprise pricing.
