How to Identify Automation Opportunities in Your Company

Everyone can name one annoying task. Few companies know which of their hundred small tasks is worth automating first. This guide is a practical audit you can run in about two weeks with a spreadsheet: inventory the work, score it on frequency, duration, error cost, data location and approval needs, and come out with a ranked list that a developer or a no-code builder can act on.

The business problem

Automation projects fail at the selection stage more often than at the build stage. Teams pick the task that most irritates the loudest person, or the one that looks impressive in a demo, instead of the one that returns the most time for the least risk. The result is a clever automation nobody uses while the real time sink stays manual.

The fix is a method, not a tool: a short, structured audit that treats every repetitive task the same way and lets the numbers decide the order.

Why the current process costs time and causes errors

Manual work hides in plain sight because it is split into small pieces. Nobody spends a day on data entry; forty people spend fifteen minutes on it. Common signals that a process is costing more than it appears:

• The same information is typed into more than one system.
• A spreadsheet is emailed around and reconciled by hand.
• Status questions (where is my order, has this been approved) arrive by email or phone.
• Work waits on one person who is the only one who knows how to do it.
• Month-end, quarter-end or audit periods bring overtime for the same tasks every time.
• Errors are found by the customer, the auditor or the bank rather than internally.

What can realistically be automated

Not every annoying task is a good candidate. The traits that make a task automatable, in rough order of importance:

1. Repetition: it happens weekly or more often, in roughly the same form.
2. Digital inputs: the information arrives as email, forms, files or system records, not as a conversation in the hallway.
3. Explainable rules: a competent new hire could do it correctly from a written procedure, even if the procedure has branches.
4. Tolerable error cost: a wrong result is caught by a later check, or is cheap to fix.
5. Reachable systems: the source and destination have an API, an export, or at least a stable file format.

Tasks that need judgment on incomplete information, negotiation, or physical handling are poor candidates, though AI can often prepare the work (summarize, draft, pre-fill) so the person does only the judgment part.

Existing tools that may solve it

Before scoring anything, check what your current software already does. Most teams pay for automation they have not switched on.

• Google Workspace and Microsoft 365: rules, templates, forms, shared inboxes, and built-in AI assistants (Gemini and Microsoft Copilot) for drafting and summarizing.
• Your CRM: HubSpot and its competitors include workflow automation for follow-ups, task creation and record updates.
• Your accounting software: QuickBooks Online and Xero both include receipt and bill capture, recurring invoices and payment reminders.
• Integration platforms: Zapier, Make, Microsoft Power Automate and n8n connect systems and include AI steps for classification and extraction; n8n can also be self-hosted on your own infrastructure.
• General-purpose assistants: ChatGPT, Claude and Gemini are useful for one-off drafting and analysis, but they are not a workflow until something triggers them, logs the result and writes it somewhere.

Capture the tool landscape as part of the inventory: knowing a task already lives in HubSpot changes its score.

When custom software is necessary

Custom software earns its cost when the audit reveals a cluster of high-scoring tasks that share a data problem: a system without an API, data split across a legacy database and spreadsheets, or rules too specific for a generic tool. It is also the right answer when volume is high enough that per-task pricing becomes material, when personal or financial information must stay within a specific jurisdiction, or when a regulator will want an audit trail you control.

The audit output tells you which: if your top five candidates all touch the same unloved system, that integration is the project, and each task becomes a feature of it.

An example workflow

Here is the audit itself, as we run it.

Week one: inventory

Give every team a simple form (Google Forms or Microsoft Forms works) with six fields: task name, who does it, how often, how long each time, what systems it touches, and what happens when it goes wrong. Ask for every task done more than once a month. Expect a long list; that is the point. Interview the two or three people who handle the most exceptions, because the form will understate their work.

Week two: scoring

Transfer the inventory to a sheet and score each task on five factors from one to five, then multiply by the weight.

Add the weighted scores. The maximum is forty-five. Sort descending. The thresholds in the table are starting points; adjust them to your scale, but keep them fixed once you start scoring so tasks are comparable.

Reading the result

• High total with a high tool-fit score: switch on the feature this month.
• High hours and error cost but low data readiness: fix the data first (a form instead of email, a system instead of a spreadsheet), then automate.
• A cluster around one system: scope a custom integration.
• Low frequency but high error cost: do not automate; write a checklist.

Implementation steps

1. Appoint one owner for the audit and give it a deadline of two to three weeks.
2. Run the inventory form across all teams; follow up with short interviews.
3. Normalize the list: merge duplicates, split tasks that are really three tasks.
4. Score with the matrix in a single sitting, two people scoring together to reduce bias.
5. Validate the top ten by timing each task for real over one week.
6. Map each finalist to a route: switch on a feature, build a no-code flow, or scope custom work.
7. Start the first automation within a month of finishing the audit, while the energy is there.
8. Repeat the audit annually; the list changes as the business does.

Security and quality considerations

• The inventory itself will contain sensitive detail about who does what and where the weaknesses are. Keep it to a small group.
• Note for every task whether it touches personal information. Under PIPEDA and the provincial private-sector privacy laws, moving that data into a new tool is a new use or disclosure that needs a reason and, often, a review of the vendor’s terms.
• Record where data resides today and where each candidate tool would store it, especially for health, financial or public-sector data where residency commitments may exist.
• Score honestly on error cost. Underestimating it is how a harmless-looking automation ends up sending wrong amounts to customers.
• Prefer automations with a reversible output (a draft, a flagged record) for anything in the top error-cost band.

Effort, costs and maintenance

The audit costs mostly internal time: a few hours per team for the inventory, a day or two for scoring and validation, and a short write-up. If you bring in outside help, ask for the scored spreadsheet as the deliverable, not a slide deck. The routes the audit produces have very different cost shapes.

How to measure whether it worked

• Audit coverage: the share of teams that submitted an inventory and the number of tasks captured.
• Time to first automation: days from audit completion to the first change live.
• Time saved: timed before and after for each automated task, summed monthly.
• Exception rate and error rate for each automation, tracked from day one.
• Backlog health: how many scored candidates remain and whether the top of the list keeps moving.

Re-score the automated tasks after three months. If their hours did not fall, find out why before touching the next item.

Talk to us about your process

If you would like a second pair of eyes on your audit, or help running it, Data Function offers a free 30-minute consultation at /book. Bring your inventory, even a rough one, and we will help you read it: which items a tool you already own can handle, and which justify a build. We work with companies across Canada and the United States from Toronto.