A one-hour AI audit of your own business: find the task worth automating first

Before buying any tool, a short audit of your own operations will tell you which task is the right first automation candidate — and exactly what success looks like.

A one-hour AI audit of your own business has four steps: list every task your team does more than once a week; apply a simple filter to find which of those tasks is rule-based and high-volume; pick the single task that clears the filter and costs the most cumulative time across your team; and write down, in one paragraph, what the correct output looks like. The task that survives those four steps is your first automation candidate. The audit does not require an outside consultant or a software subscription — a notepad and one hour of honest attention to your own operations is enough to produce a specific answer. That answer is more useful than any general AI strategy document, because it tells you where to start and what success looks like before any tool is selected or any dollar is spent.

The filter: what makes a task worth automating first

Not every manual task is a good automation candidate. The ones worth prioritizing share four characteristics, and a task is a stronger candidate the more of these it satisfies.

• Frequency: the task happens multiple times per week, ideally multiple times per day. A task that takes two hours but happens twice a year is a poor automation target. A ten-minute task that happens forty times a week is an excellent one. Multiply the time per instance by the number of instances per month — anything above twenty hours of cumulative monthly effort across your team is worth examining closely.
• Rule-based: a competent new hire could execute the task correctly on their first day by following a written checklist, without applying judgment about context, relationship, or history. If the right output depends on something that is not written down anywhere, the task is not yet ready to automate.
• Involves data re-entry or copying: the task requires moving information from one place to another — a web form into a spreadsheet, a voicemail into a booking system, an email into a CRM field. Manual re-entry is the clearest signal of an automation opportunity because the input and the output are both already defined.
• Recoverable when wrong: a failure on any single instance is annoying but fixable, not catastrophic. The first automation project should not be the task where a single error causes a compliance breach, a safety incident, or an unrecoverable customer situation. Start somewhere that a mistake surfaces quickly and is easy to correct.

The one-hour audit, step by step

• Set a timer for 20 minutes and list every task your team does more than once a week. Do not edit or filter as you go — capture everything: booking confirmations, follow-up calls, estimate requests, invoice reminders, scheduling, data entry, status updates, answering the same recurring questions. If other staff are available, spend five minutes asking them to add to the list from their own day.
• Apply the filter. For each item on the list, ask: does this happen at least weekly, could a new hire do it from a written checklist, and does it involve copying or re-entering information from one place to another? Items that answer yes to all three move to the short list.
• Estimate cumulative monthly time. For each item on the short list, write the time per instance and the number of times it happens per month, then multiply. Sort by total monthly minutes. The item at the top is your priority candidate.
• Write the definition of done. In one paragraph, describe what the correct output looks like for the top candidate — not the current process, just the output. What does a good result look like? If you cannot write this clearly and specifically, the task needs more definition before it can be automated. That is useful information before you spend anything.

What this looks like in a Northwestern Ontario business

The tasks that clear this filter vary by sector but follow the same pattern. A trades or construction business in Thunder Bay often finds it in job confirmations and appointment reminders — the same message sent to a customer before every scheduled visit, currently written by hand or dictated by phone. A tourism operator or lodge typically finds it in booking inquiries that all ask the same three or four questions before a reservation can be made. A healthcare clinic frequently identifies it in appointment follow-up and recall messages — the same text or call, triggered by the same schedule. A professional services firm — accounting, insurance, legal — often finds it in document request sequences: the same list of items, requested the same way, for every new client file. In each case, the task is repetitive, the correct output is already known, and the person doing it manually could write it down in a paragraph. That describability is the signal. The ISED SME AI Adoption Blueprint notes that successful AI projects tend to begin with tightly defined problems that align with specific business priorities. The one-hour audit is the mechanism for finding and defining that problem before any tool is selected.

Frequently asked questions

• What if our whole business feels like one big manual process? Start smaller. Pick one department, one service line, or one week of one person's work. The audit works at any scope — the goal is to narrow down to a single specific task with a defined output, not to map everything at once. One specific candidate is the result.
• What if the task we identify involves a phone call or a paper form, not a software process? That is fine — the task does not need to be digital already. A phone call that always follows the same script and always ends with the same action is automatable. A paper form that gets transcribed into a spreadsheet is automatable. The audit is about finding the task, not pre-filtering for tasks that are already partly digital.
• How do we know whether AI is the right tool or whether basic automation is enough? If the correct output is identical every time — the same message, the same document, the same data entry — basic automation is likely sufficient and is simpler to implement and maintain. AI earns its place when the output needs to vary based on what came in: a reply that changes based on the customer's specific situation, a document that incorporates variable context, a summary that adapts to the contents of a call. Match the tool to the actual variability in the output, not to the most sophisticated option available.

Sources: ISED Canada — The SME AI Adoption Blueprint, including guidance that successful AI projects begin with tightly defined problems aligned with business priorities: ised-isde.canada.ca/site/ised/en/sme-ai-adoption-blueprint | BDC LIFT program (announced April 24, 2026; $500 million allocated; loans from $25,000 to $5 million; principal deferral up to two years; priority to Canadian-developed AI solutions; advisory support available): bdc.ca/en/about/mediaroom/news-releases/bdc-launches-lift-getting-canadian-smes-off-the-ai-sidelines | G7 SME AI Adoption Blueprint — Ministerial Statement: g7.utoronto.ca/ict/2025-sme-ai-adoption-blueprint.html. This is general information and does not constitute advice specific to any business or project. Confirm program eligibility and current terms with BDC directly before applying.