Quebec SMBs: the 3 processes to automate with AI before anything else
Dipesh Walia
5/3/2026
The problem with "doing AI"
In 2026, most SMB owners in Quebec are asking the same question: "where do we start with AI?" The answers they get are usually bad — either horizontal tools (ChatGPT for everyone), or 6-month transformation projects that ship nothing concrete.
The truth is simpler. Before automating anything, a process has to pass four filters:
- Frequency. At least 20 times per month.
- Structure. Same input shape, same output shape.
- Low risk tolerance. An error is catchable by a human before it causes damage.
- Visible current cost. You can name how many hours it takes this week.
Anything that passes the 4 filters is a candidate. Anything that fails one, is not.
The 3 processes that pass the filter 80% of the time
1. First-level triage and response on inbound email
In almost every SMB, a generic inbox (info@, contact@, sales@) absorbs 30-80 messages a day. A human sorts them, answers the easy ones, and routes the real ones.
An AI agent does that well. It reads, classifies (product question / quote request / support / spam / other), answers the 3-4 most common cases with a template, and forwards the rest with a summary. Typical saving: 10-15 hours per week.
2. Proposal / quote preparation from a client brief
Whether it is a construction bid, a marketing campaign quote, or a legal mandate plan, the pattern is the same: a sales person or partner gets a fuzzy brief, and 2-4 hours later produces a structured proposal.
An LLM with access to your last 200 proposals does 80% of the work in 10 minutes. The sales person reviews, adjusts the numbers, personalizes the tone. Time per proposal: 30 minutes instead of 3 hours. And quality goes up, because AI does not skip sections.
3. Meeting summary and action routing
SMBs run far more meetings than they admit. Almost none come out with structured notes and clear actions.
An AI tool that listens to the meeting (with consent), produces a structured summary (decisions, actions, owners, deadlines), and sends it to participants changes operational culture in 2-3 months. It sounds light. In practice, it is the process with the biggest effect on overall execution.
How to know which of the three to start with
The rule we use with clients: start with the one you can quantify fastest. If you can say "we lose X hours/week on this, and it costs us $Y", you have your pilot. If you cannot, spend 2 weeks measuring before automating.
What not to automate first
- Strategic decisions. AI is bad at choosing between two directions — it is excellent at executing once the direction is set.
- High-empathy interactions. Negotiating with an unhappy customer, hiring interview, delivering bad news. Leave that to humans.
- Processes you are in the middle of changing. If your workflow is still moving, automation will crystallize the wrong version.
The honest question to ask before starting
"If I freed 200 hours per month inside my team, what would I do with them?" If the answer is not clear (more clients, new product, advisory, quality of life), automation will not give you ROI. The problem is not technical — it is what you will do with the output.
If the answer is clear, then 2 weeks is all it takes to ship a first pilot.
To clarify your case
The AI Process Audit Sprint identifies your 3 most profitable processes to automate, quantifies the ROI, and ships the first pilot live. Flat fee, 2 weeks.