AI consulting: the firms that sell it and what to ask before you hire one
AI consulting is paid outside help to decide where AI fits in your business and to build it into the systems you already run. Four kinds of firm sell it: large consultancies, specialist AI firms, freelancers and the services arms of software vendors. They charge by the hour, by the project or by the month. Before you hire one, get written answers on where your data goes, who does the work, what you own at the end and how the result will be measured. This guide is for owners, operations leads and IT leads at smaller manufacturers and distributors in Ontario and Quebec.
What AI consulting covers
An AI consultant helps a business put AI to work on its own records, such as quotes, orders, drawings and email. The work falls into four parts, and a firm may sell any of them:
- Strategy. Choosing which jobs AI takes on first. AI strategy and roadmap shows how to do this yourself.
- Implementation. Connecting AI to your data and building the tool your team uses, as set out in AI implementation.
- Training. Teaching your staff to use AI on their own work, which AI training covers.
- Support. Keeping the system running and up to date after launch.
A use case is one specific job that AI does, such as reading a customer’s purchase order and drafting the order in your ERP. ERP stands for enterprise resource planning, the system that holds your orders, inventory and costs.
Statistics Canada asked businesses that had used AI in the previous 12 months how they had changed their training or staffing. In its survey for the second quarter of 2026, 30.2% of those with 100 or more employees had used external consultants or vendors, against 10.7% of those with 1 to 4 employees. To see your own starting point before you call anyone, take the AI readiness assessment, and for a wider view, read AI for small business.
The four kinds of firm that sell AI consulting
Large consultancies include the Big Four audit firms, such as Deloitte, EY and PwC, and firms such as Accenture. Deloitte and Accenture both sell forward deployed engineering, meaning engineers who work inside the client’s business and build on its own systems. Deloitte’s service page describes “pods of specialist professionals” embedded with the client’s team. In June 2026, Accenture launched Accenture Edge for “mid-market companies with annual revenues between $300 million and $3 billion.” ThriveAI vs the Big Four compares what these firms publish.
Specialist AI firms are smaller companies whose whole business is AI. Some give advice and others build and run systems. ThriveAI, which publishes this guide, is a small AI engineering company in Ottawa.
Freelancers are independent developers or advisers who work alone. Make sure the code, the accounts and the documentation end up in your company’s name. If you are weighing a permanent hire instead, read hiring an AI engineer in Canada.
The services arms of software vendors sell consulting on their own platforms. AWS, Amazon’s cloud business, says its Professional Services team helps organizations “identify production ready use cases, build implementation roadmaps, and effectively deploy AI solutions.” Because the work runs on the vendor’s platform, ask what it would take to change the model or the provider later.
| Kind of firm | Who does the work | Fits best when | What to check first |
|---|---|---|---|
| Large consultancy | Teams from the firm, such as Deloitte’s pods | AI is part of a company-wide system rollout, or you need many people at once | Who will be on your floor, and for how many days |
| Specialist AI firm | A small team whose whole business is AI | You want one defined job built on your own data | Named people in the proposal, and support after launch |
| Freelancer | One person | The build is small and has a clear end | Who maintains the system if that person is unavailable |
| Vendor services arm | The vendor’s own team or its partner firms | You have already chosen that vendor’s platform | What it would take to change the model or provider later |
How AI consulting firms charge
Firms charge in three common ways, and one proposal can combine them. This guide gives no amounts, because rates depend on the firm and the scope. What AI costs to run once it is built is covered in the cost of AI for a small business.
| Model | How it works | Who carries the risk of overruns | What to ask for |
|---|---|---|---|
| Hourly or daily rate | You pay for the time worked, plus expenses. This is also called time and materials. | You do. If the work takes longer, the bill grows. | A written estimate, a cap that needs your approval to exceed, and timesheets with each invoice |
| Fixed project price | One price for a defined scope, which can be paid in stages as the work is delivered. | The firm does, within the agreed scope. A change of scope needs a written change order. | A precise scope and the test the work must pass before the final payment |
| Retainer | A set monthly fee for a set amount of time or a list of services, such as support. | The risk is shared. The contract decides what happens to hours you do not use. | What the fee covers, whether unused time carries over, and the notice needed to end it |
Before you sign, ask for a statement of work, the document that lists the tasks, the deliverables and the tests the work must pass. Ask for the first phase to end with something working on your own data, so you can judge the firm before a larger commitment.
Start with the job that costs you the most
Tell Derik which job slows your team down and which systems it touches. He reads every enquiry himself and usually replies within one business day.
Start a conversationWhat to ask before you hire an AI consultant
Put these questions to every firm on your list, including ThriveAI, and ask for the answers in writing.
Where your data goes
An AI system handles your data in two places: where it is stored, and where the model processes each request. The model is the AI program that reads a request and writes the answer, such as the large language models (LLMs) behind ChatGPT and Claude. A prompt is the request, with any records attached to it. Ask each firm:
- Where will our data be stored, by which company and in which country?
- Which model processes our prompts, where does it run, and does its provider keep a copy? Under a zero data retention agreement, the provider keeps no copy.
- Will any of our data be used to train a model?
- Which subcontractors and outside tools will see our data?
The law holds you responsible too. Personal information is information about an identifiable person, such as a customer contact’s email. Under clause 4.1.3 of PIPEDA, the federal private-sector privacy law, an organization “is responsible for personal information in its possession or custody, including information that has been transferred to a third party for processing.” In Quebec, section 18.3 of the private-sector privacy act lets a business share personal information with a contractor without consent only under a written contract that sets out how the contractor will protect it, use it only for the contract and not keep it afterwards. Section 17 requires a privacy impact assessment, a written review of the risks, before personal information is communicated outside Quebec. Private AI for business and data sovereignty in Canada go deeper, and your counsel decides how the rules apply to your contract.
Who does the work
- Who, by name, will do the work, and for how many days?
- Will any of it be subcontracted, and to whom, in which country?
- Who works on site with your team, and who supports the system after launch?
Get names and days in writing, because you can check them against the invoices later.
What you own at the end
Your company should end the work holding everything it needs to run and change the system without the firm: the source code, the prompts and settings, the test cases, the documentation, and the accounts and API keys, all in your company’s name. An API key is the password one program uses to call another.
Canadian law does not give you that by default. Section 13 of the Copyright Act makes the author of a work the first owner of its copyright, and the Act’s definitions count computer programs as literary works. Section 13(3) gives an employer first ownership of what its employees make in the course of their employment. Code from an outside firm or a freelancer passes to you only by an assignment, which section 13(4) says is not valid “unless it is in writing signed by the owner of the right.” If the firm keeps code it reuses across clients, get a written licence to use and change it for as long as you run the system.
How the result will be measured
Agree on the measure before the work starts, and record today’s number as the baseline, such as the minutes it takes to enter an order or the time from a request to a quote. Ask how the firm will test the system on your own data. An eval, short for evaluation, runs real past jobs with known answers through the system and grades each output, and AI evals shows how to build one. Ask who approves the AI’s drafts before they reach a customer, a supplier or your ERP, the step that human in the loop covers.
Every proposal should say what will be running on your own data at the end of the first phase and how it will be measured. If yours does not, ask for that sentence before you sign.
Red flags in an AI consulting proposal
- No separate answer on processing. “Hosted in Canada” can describe storage while the model runs elsewhere. AWS documents that pattern: a model request sent from its Canada (Central) Region can be processed in another region, while stored data “remains exclusively within the Canada (Central) Region.”
- No names. The proposal describes a team but names nobody, or the people you met are missing from it.
- A long paid study first. Months of analysis are planned before anything runs on your data.
- Savings promised in advance. The proposal quotes a result before anyone has looked at your records.
- Accounts in the firm’s name. The firm holds the code, the accounts or the keys, so you cannot run or move the system without it.
- No approval step. The AI sends email, places orders or changes ERP records with nobody approving it.
How ThriveAI works
ThriveAI is an AI engineering company in Ottawa that builds private AI systems for manufacturers and distributors in Ontario and Quebec, on their own data. It works hands on with your team to clear the manual computer work behind your bottleneck. Derik Lawlis, the founder, leads every project and stays close to the build, and working sessions run on site with your team, in French or English.
In stage one, ThriveAI builds a working tool for the one job that costs you the most, at a fixed price, in weeks. It runs on your own systems, and if you stop there, you keep it. You decide whether to start stage two, the full build, after you have seen stage one working. Every project is planned and quoted on its own.
The platform ThriveAI builds on is designed to keep each client’s data on its own server in Canada. You choose the AI model that reads it: one on that server, or a hosted model, one that runs on the provider’s servers, under a written agreement that the provider keeps nothing after answering. A hosted model may process requests outside Canada, so the contract names the model. Nothing is sent or saved in your systems until the person responsible approves it. If you stop working with ThriveAI, you keep your data, and everything the system recorded exports in full in a documented format.
Rios-Karim Mercier of Belmont Capital wrote that Derik “proposed solutions that were effective to implement and everyone on our team was impressed and happy to work with him.” About ThriveAI covers the company, ThriveAI vs alternatives sets ThriveAI beside other ways to get this work done, and the guides go deeper on each topic.