# AI Consulting for Small Business > Practical AI consulting for small businesses in Ontario: we find the repetitive work in your week, automate the right parts and train your team to run it. Source page: https://mnpsystems.ca/ai/ MNP Systems is an engineering consultancy based in Brampton, Ontario, Canada, that takes work from anywhere: remotely, or on site where a job needs hands on the equipment. Contact: https://mnpsystems.ca/contact/ or contact@mnpsystems.ca. **Give your team its hours back.** We find the repetitive work in your week, hand the right parts to AI, and train your team to run it, with a person approving anything that leaves the building. Works with the tools you have: Microsoft 365 Copilot, Google Workspace with Gemini, ChatGPT and Claude business plans, Zapier, Make and n8n. ## Where the hours go - The same questions, answered again every day. An assistant answers from your own policies, and hands anything unusual to a person. - Copying the same details between three apps. The apps pass the details along themselves, and flag what does not match. - Quotes and proposals that eat an evening. AI drafts from your price sheet; you check it and send it. - Everyone using ChatGPT their own way, customer data included. One set of tools, a short policy and some training, so it is used well and safely. ## What we do Most businesses start with the audit. The rest follows from what it finds. ### AI opportunity audit We map where your team's week actually goes, then rank what to hand to AI, what to automate, and what to leave alone. Example: You get a ranked list of your tasks, the hours each one takes, and a plan for what to do first. ### Workflow automation The copying and pasting between your inbox, calendar, spreadsheets and accounting software, done by a system that flags anything it is unsure of. Example: A web enquiry is filed, answered and booked into the calendar without anyone retyping it. ### Assistants that know your business An assistant that answers from your own price list, policies and procedures, for your staff, your customers or both. Example: "Do you deliver to my postal code?" is answered from your real delivery zones, and anything unusual goes to a person. ### Team training & AI policy Hands-on sessions, a playbook for each role, and a one-page policy that says which information may go into which tool. Example: Your front desk learns to draft replies with AI, and everyone knows what never goes into a chat window. ## Beyond the office Four kinds of AI work that need an engineering firm, not just a subscription. ### Assistants on the shop floor A technician asks "what does alarm E-217 mean on line 3?" and gets the answer from that machine's manuals and your own repair log, on a small computer beside the line that keeps working when the internet doesn't. Why an engineering firm: We build embedded and industrial systems, so the assistant is one more device on the line rather than a subscription bolted on. ### Getting found by AI People now ask ChatGPT, Gemini and other assistants for a plumber nearby or a caterer for Friday. Assistants can only recommend what they can read, so we make your site easy for them to read and to trust. Why an engineering firm: This site is built that way, and so are the client sites we make: every page has a plain copy that AI assistants read directly. ### AI you can test Before your chatbot talks to a customer, it passes a set of real questions: prices it must not invent, details it must not ask for, topics it must hand to a person. The same set runs again whenever it changes. Why an engineering firm: We test firmware this way. The assistant on this site comes with its own set of test questions. ### Ready for your customers' AI Customers are starting to ask their own AI assistants to book a table or reorder supplies. We connect your menu, stock or calendar so those assistants can see it and book, within the rules you set. Why an engineering firm: It is the same work as an online store's back end, with a new kind of visitor. ## What it looks like An example, not a case study: a quote request, before and after. **Before.** Four of the five steps wait for someone to be free. 1. The request arrives from your website (Automated) 2. It waits in the inbox until someone is free (Person) 3. The details are retyped into a spreadsheet (Person) 4. The quote is written that evening (Person) 5. It goes out the next day (Person) **After.** A person does the one step that needs judgement. 1. The request arrives from your website (Automated) 2. Its details are read and filed (Automated) 3. A quote is drafted from your price sheet (AI draft) 4. You check it and approve it on your phone (Person) 5. It is sent, and a copy is filed (Automated) ## How an engagement runs 1. **Audit.** We sit with your team, map where the week goes, and rank what to change first. 2. **Pilot.** One workflow, built and running on your real work, so you can judge it on what it does. 3. **Roll out & train.** What works spreads to the rest of the team, with training and a playbook for each role. 4. **Support.** We keep it running as the tools change, and review with you each month what to try next. ## How we keep it safe - **Your data stays yours:** Business plans with training switched off, or a model that runs on your own computer when the information must not leave the building. - **A person approves what goes out:** AI drafts the quote, the reply or the post. Someone on your team decides whether it is sent. - **Every automation has an off switch:** Each one keeps a log of what it did, and anyone on your team can pause it. - **You keep the know-how:** Playbooks, training and documentation stay with you, so you are never tied to us. ## We use it ourselves - **Our articles read themselves aloud:** Every post on our blog has a Listen button. The voice is an AI model running on our own computer, not a cloud service, pinned to exact versions and checked by tests, so an article cannot be published saying something its audio does not. [Hear it on the blog](https://mnpsystems.ca/blog/) - **Websites that AI assistants can read:** Every page of this site has a plain-text copy that ChatGPT, Claude and other assistants can read directly, and we build the same into the websites we make for clients. [See it in a client site](https://mnpsystems.ca/websites/airbloom/) - **This site has an assistant:** The "Ask MNP" button on every page opens an AI assistant that answers only from this site's pages, sends you to a person for anything else, and comes with a set of test questions it must handle correctly, such as never quoting a price. ## Frequently asked questions ### Will AI replace my staff? That is not the goal, and it is not how we work. We hand AI the repetitive parts of a job, such as retyping, first drafts and reminders, so your people spend their time on customers and judgement. Anything that goes out still goes past a person. ### Is it safe to put customer data into ChatGPT? It depends on the plan and its settings. On many free and personal plans, what you type can be used to improve the provider's models unless you switch that off; business plans keep your data out of training. Every engagement includes a one-page policy saying which information may go into which tool, and when information must not leave the building at all, we can run a model on your own computer instead. ### Which tools do you work with? Mostly the ones you already pay for: Microsoft 365 Copilot if your business runs on Outlook and Teams, Google Workspace with Gemini if it runs on Gmail, and the business plans of ChatGPT and Claude. To connect your apps to each other we use Zapier, Make or n8n. We recommend what fits the tools you have. ### Do we have to buy new software? Usually not. Most of what we build runs on software you already pay for. When something new would help, the audit says what it costs and what it replaces before you decide. ### What if we don't know what to automate? That is what the audit is for. Knowing which parts of the week you dread is enough to start. We sit with your team, see where the hours go, and bring back a ranked list. ### Can I try an AI assistant before we buy one? Yes. The "Ask MNP" button on every page of this site opens one. It answers from our own pages, says when it does not know, and hands you to a person for anything it should not handle. Yours would work the same way, from your own information. ### How soon would we notice a difference? We choose the pilot so its effect shows up during the pilot itself: one workflow, running on your real work. If it does not earn its keep there, we say so before anything else is built.