Questions an AI consultant should ask you, and what a missing one tells you

The questions an AI consultant should ask you fall into five groups: what the repeated work is and who does it, which systems hold it, how many people you employ in Ontario, where personal information goes, and what would make you stop the project. A competent consultant asks most of them before naming a single tool. If you leave a first meeting having answered almost none of them, whatever you were quoted describes somebody else’s business.

This is the other half of a first meeting. What you bring to it is covered in how to prepare for an AI consultant. This page is about what you should hear across the table, and how to tell a diagnosis from a pitch by the questions alone.

Why the questions are the test

You cannot judge a consultant’s technical skill from one meeting, and you do not need to. You can judge whether they are curious about the right things. A recommendation is only as good as what it is based on, and the questions show you what that is.

There is a second reason. Most small businesses have not started. In Statistics Canada’s survey for the second quarter of 2026, 19.2% of businesses reported using AI to produce goods or deliver services, so many of the owners a consultant meets have no way yet to check a technical answer. The questions are the part they can check.

The work itself

These come first, and they should take most of the meeting.

• “Walk me through the last time this job happened.” Not how it is supposed to work. The last real time, with the workaround in it. A consultant who asks for the manual version will design for a process nobody runs.
• “Who does it, and what do they do when something does not fit?” The exceptions decide whether a job can be automated cleanly or needs a person approving part of it.
• “How often, and how long each time?” A rough count is fine. A consultant who does not ask has no way to tell you whether a build pays back.
• “What goes wrong, and who finds out?” A job that fails quietly and gets caught by a customer is worth more to fix than one that is merely slow.

If you wrote a one-page brief before the meeting, several of these are already answered on it. Watch whether the consultant reads it and asks the next question down, or asks you to repeat it.

The systems

• “What does the information live in, at each step?” Software, a spreadsheet, an inbox, a notebook in a truck.
• “Who holds the admin login for each one?” This is a practical question. An automation needs somebody on your side who can grant access and take it back.
• “Which of these are you willing to change, and which are fixed?” Some businesses can switch a scheduling tool. A firm tied to an industry system may not be able to, and a good design works around that instead of assuming it away.

The answer to the first question shapes the cost. A project in workflow automation in Ontario runs $2,500 to $12,000 here, and where a quote lands depends mostly on how many of those systems must exchange data.

Your people, and the Ontario rules that come with them

In my judgment this is the group most likely to be skipped, and it is the one that shows whether the consultant has worked with Ontario employers before.

• “How many people do you employ in Ontario?” Headcount is not small talk. Under the Employment Standards Act, an employer with 25 or more employees in Ontario on January 1 of any year must have a written policy on the electronic monitoring of employees in place before March 1 of that year. The province counts individual people, not full-time equivalents, so part-time staff each count as one.
• “Will this project watch, record or measure anybody’s work?” Ontario’s guide says “electronic monitoring” includes all forms of employee monitoring done electronically, and gives tracking the websites staff visit during working hours as one example. A tool that logs how fast staff process orders, or records their screens to capture a process, is worth checking against your policy. If you are over the threshold and have no policy, that is a finding in its own right.
• “Will any of this touch hiring?” If an automation screens or ranks applicants, the Ontario AI job posting disclosure applies to employers of the same size, and the posting has to say so.
• “Who on your staff will use it, and who will resist it?” The people doing the work today decide whether a build gets used or quietly abandoned.

None of this makes a consultant a lawyer, and a good one will say so. The point is that they raise it before you have to.

Personal information

• “Does customer or staff personal information pass through this job?” Names, addresses, health details, account numbers.
• “Which tools are you already allowed to put it into?” Some businesses have a client agreement or an industry rule that limits where information can go. That limit decides which tools are eligible before anyone compares features.
• “Do you have a written rule on what staff may paste into a chatbot?” If the answer is no, the consultant should say whether that comes first.

The federal rules behind these questions are set out in PIPEDA AI compliance in Canada. You do not need to know them before the meeting. You do need to hear the consultant ask.

The questions about stopping

These are the easiest to leave out, and the most telling.

• “What would make this not worth doing?” A consultant who asks this is prepared to tell you to leave a job alone.
• “If the build stopped working on a Friday night, who would notice, and what would it cost by Monday?” That answer decides how much checking and alerting a build needs.
• “If you wanted to change supplier in a year, what would you need from me?” A consultant who asks this plans to leave you with accounts in your name and a written handover.

Questions a consultant should not ask yet

Some questions are fine later and wrong at the start:

• For a password, an export of your customer list, or access to a live inbox. None of that is needed to understand the work, and access needs paperwork first. The AI consultant NDA basics post explains what to sign, and when.
• “Which AI tool were you thinking of?” Asked first, it hands you the diagnosis. The tool follows from the job.
• “What is your budget?” Asked before anything about the work, it tells you the recommendation will be sized to the money rather than to the problem.

If the opening questions are about tools and budget, read signs an AI consultant is overselling before the second meeting.

Keep a tally when you talk to several

The quotes you collect will not line up, and the questions are the easiest way to compare the people behind them. After each meeting, tick what was asked:

Group	Asked?	Notes
The last real run of the job, with exceptions
Volume and time per occurrence
Every system, and who holds the admin login
Ontario headcount and electronic monitoring
Hiring, if the build touches applicants
Where personal information goes
What would make it not worth doing
What you would need to change supplier

A consultant who ticked six or more rows has heard enough to be wrong in a useful way. One who ticked two has quoted a guess.

If you are hiring in Toronto

For firms across the Greater Toronto Area, from a downtown accounting practice to a distribution office in Mississauga, add one question to the list: “Which client information may never leave the system it lives in?” A professional firm may have engagement terms or regulator rules that answer that before any tool is chosen. An AI consultant for Toronto firms like these should ask it in the first meeting, because the answer can rule out the cheaper options early.

Headcount matters here too. A Toronto firm that has grown past 25 people can cross the electronic monitoring threshold without anyone noticing, and a project to measure document turnaround is exactly the kind that the policy has to describe.

Where the assessment fits

Asking these questions properly takes longer than one meeting, which is why the diagnosis is worth buying on its own. The fixed-fee AI business assessment for Ontario firms on this site is $999 over ten business days. Its 45 minute kickoff call asks about headcount, your software and the hour when the business is most stretched, and the result is a written report you walk through with the person who signs.

Whether you buy that or not, use the table. The consultant worth hiring is the one who asked you the most about your business before telling you anything about AI.