"Where should we use AI?"
It's the right question. Most people just ask it too broadly. You can't answer "where should we use AI in our business" any more than you can answer "where should we use electricity in our building." The answer is "it depends on what you're doing and where the friction is."
An AI opportunity scan is a structured way to find those friction points. Here's how it works, and how you could do a rough version yourself.
Step One: Inventory Your Workflows
Before you think about AI, you think about work. What does your team actually do all day?
This sounds obvious, but most business owners can't describe their workflows in detail. They know the big picture: "we get leads, we do sales calls, we send proposals, we do the work, we invoice." But the connective tissue between those stages, the admin, the coordination, the data handling, tends to be invisible.
Spend a day just watching. Or better, ask each person on your team to track their tasks for a week. Not in a surveillance way. In a "help me understand where your time goes" way. Use something simple, a shared spreadsheet, a time-tracking app, or even sticky notes.
What you're looking for is the actual work, not the org chart version of the work. The real version always has more steps, more handoffs, and more small tasks than you expect.
Step Two: Find the Time Sinks
Once you have an inventory, look for three things.
Repetitive tasks. Anything someone does the same way multiple times a week. Data entry, email responses that follow a pattern, report generation, status updates.
Bottleneck tasks. Things that hold up other work because they're waiting for one person to finish something manually. Approvals that require someone to read and summarize information. Handoffs that require reformatting data from one system to another.
High-volume, low-complexity tasks. Things that aren't hard but there are a lot of them. Processing applications, categorizing inquiries, updating records.
These three categories are where AI tends to have the highest impact. Not because they're the most important work, but because they're the most compressible. A task that takes 20 minutes and happens 10 times a week is 3+ hours of recoverable time if you can automate it.
Step Three: Score by Impact and Risk
Not every opportunity is worth pursuing. Some tasks are automatable in theory but risky in practice. Some save time but not enough to justify the setup cost.
I use a simple scoring approach. For each candidate task, rate two things on a scale of 1 to 5.
Impact: how much time or friction does this task create? A task that eats five hours a week across your team scores higher than one that takes 15 minutes.
Risk: what happens if the automation gets it wrong? A misrouted internal email is low risk. A wrong number on a client invoice is high risk.
High impact, low risk tasks are your starting points. They're the easy wins. High impact, high risk tasks are worth pursuing but need more safeguards, like human review steps or approval gates. Low impact tasks of any risk level go to the bottom of the list.
What a Scan Typically Finds
Having done this for a lot of small businesses, the patterns are remarkably consistent.
Email is always on the list. Drafting, responding, following up, sorting. Every business spends more time on email than they realize, and a meaningful chunk of it is automatable.
Data moves between systems manually. Information comes in through one channel (email, a form, a phone call) and someone manually enters it into another system (CRM, project management tool, spreadsheet). This is integration work, and it's almost always worth automating.
Scheduling eats more time than expected. The back-and-forth of coordinating meetings, appointments, and deadlines is pure friction that compounds across team members.
Reporting is manual. Someone spends Friday afternoon pulling numbers from three different tools into a spreadsheet to create the weekly report. AI and automation can assemble those reports automatically.
Follow-ups fall through cracks. Proposals sent without follow-up, tasks assigned without check-ins, client requests acknowledged but not tracked. Not because people are careless, but because the volume is high and the system relies on memory.
The DIY Version
You can do a rough version of this scan yourself. Here's the condensed approach.
Pick your three most time-consuming workflows. For each one, map out every step, including the ones that feel too small to mention. Time each step, or at least estimate honestly. Identify which steps are repetitive, manual, and low-judgment. Those are your AI candidates.
It won't be as thorough as having someone external do it. You'll have blind spots, things that feel normal because you've always done them that way. But it'll get you 60-70% of the way there, which is plenty to start.
The Professional Version
When we do an AI opportunity scan as part of a Flow Check, we bring the outside perspective that internal scans miss. We interview team members individually (people are more honest about where their time goes when they're talking to someone outside the org), map workflows in detail, and come back with a scored list of opportunities ranked by impact, risk, and implementation effort.
The output isn't a strategy document. It's a list that says "here are the five things worth automating, here's the estimated time savings for each, and here's what it would take to set them up."
Whether you do it yourself or have someone help, the scan is the right starting point. It turns "we should probably be using AI" into "here's specifically where AI saves us the most time."
The three automations post covers the most common findings. And the admin tasks post gets specific about tools and approaches for each category.