Module 2: finding your AI opportunities
In Module 1, you built your AI reality check. You named your current state, the work that feels heavy, the risks you need to respect, and what a useful result would look like. Now we turn that into a target.
This module helps you find the parts of your business where AI can create practical value first. Not because a tool vendor says it can. Not because a trend makes it sound exciting. Because the work itself has the right ingredients: it repeats, it follows a pattern, it takes meaningful time, and it can be improved without handing over judgment too early.
By the end of this module, you’ll have a completed AI opportunity matrix. That matrix becomes your decision filter for the rest of the course. It tells you what to try first, what to handle with AI assistance, what to leave human, and what needs cleanup before AI touches it.
This is one of the places where small businesses can get ahead quickly. Most people adopt AI by wandering from tool to tool. You’re going to do it by studying the work.
Lesson 1: map the work before you automate
The fastest way to waste time with AI is to start with a tool and then go looking for a reason to use it.
That pattern feels productive at first. You open a new AI app, test a few prompts, get a few interesting answers, and imagine all the things it might do. Then a week later, nothing in the business has actually changed.
That doesn’t happen because the tool is useless. It happens because the tool was never attached to a real workflow.
Before you can decide where AI belongs, you need to see where your work actually goes. That means building a task inventory: a plain record of recurring work across the business.
Think of this as an X-ray of your operations. You’re not trying to judge anything yet. You’re just trying to see the work clearly enough to make better decisions.
Start with real work, not ideal work
Most business owners describe their workload based on the version of the business they want to have. The cleaner version. The organized version. The version where every process is intentional and every task has a clear owner.
The real business usually looks messier. There are emails you answer because nobody else knows what to say. Proposals you rewrite from old files. Customer questions that repeat every week. Content ideas scattered across notes, meetings, and memory. Reports that require copying information from three different places. Follow-ups that depend on someone remembering at the right time. That’s the work you’re looking for.
Don’t map your business from the org chart. Map it from last week. Look at your calendar, inbox, CRM, task manager, accounting system, customer messages, support tickets, project updates, and notes. The evidence is usually sitting right in front of you.
Capture the task, trigger, input, and output
The draft version of a task inventory usually lists tasks too broadly. “Admin.” “Marketing.” “Customer service.” “Sales follow-up.” Those aren’t tasks. They’re buckets.
A useful task inventory names the action clearly enough that someone else could understand what happens. Instead of “marketing,” write “turn one blog post into three LinkedIn posts.” Instead of “customer service,” write “reply to delivery status questions.” Instead of “sales,” write “send a follow-up email after a discovery call.”
For every recurring task, capture four things first: what triggers it, what information goes into it, what output gets produced, and who reviews or uses the result.
The trigger tells you when the task starts. A customer fills out a form. A meeting ends. An invoice becomes overdue. A new lead enters the CRM. A weekly report is due.
The input tells you what the task depends on. That may be form answers, meeting notes, a customer profile, a quote template, a product catalog, support history, or raw data from a spreadsheet.
The output tells you what gets created. That may be an email, summary, invoice, report, task assignment, document, social post, proposal, or decision.
The owner tells you who does it now and who needs to approve it later. This is where the AI opportunity starts to show itself. A task with a clear trigger, clear input, and clear output is much easier to support with AI than a task that only exists in someone’s head.
Walk through the main areas of the business
A small business usually hides AI opportunities in ordinary places. Customer communication is often the first one: inquiries, follow-ups, support replies, appointment reminders, quote requests, thank-you emails, and status updates.
Marketing is another. AI can help with idea generation, drafting, repurposing, editing, research summaries, campaign briefs, content calendars, and turning customer questions into useful content.
Operations tends to be the quiet source of value. Meeting notes, SOPs, internal updates, document summaries, task handoffs, project status reports, checklists, vendor comparisons, and recurring admin work can all become easier when the process is clear.
Sales has its own opportunities: lead qualification, call summaries, proposal drafts, CRM notes, objection tracking, follow-up reminders, and quote preparation. You still need human judgment, especially around pricing and relationship management, but AI can reduce the manual work around the decision.
Finance and HR require more caution. Invoicing, expense categorization, payroll support, resume screening, onboarding materials, and policy drafts may be useful candidates, but privacy, accuracy, and fairness matter more here. These are usually AI-assisted workflows with review, not set-and-forget automation.
If you have a team, don’t build the inventory alone. Ask each person to list the recurring tasks they do every week, especially the tasks that feel too small to mention. Those are often the ones eating the most time because they happen constantly.
Look for friction, not just volume
Time matters, but it isn’t the only signal. Some tasks don’t take many hours, but they create friction every time they happen.
Maybe a five-minute customer reply interrupts focused work ten times a day. Maybe a weekly report only takes an hour, but it always requires chasing three people for updates. Maybe proposal writing takes less time than you expected, but the quality varies depending on who wrote the last one.
AI opportunity isn’t only about raw hours. It’s also about reducing rework, delays, inconsistency, interruptions, and avoidable decisions.
As you build your inventory, mark the tasks that feel heavier than their time estimate suggests. Those are worth watching. They may not score highest on time saved, but they might improve the business in ways your calendar doesn’t show.
Separate tasks from decisions
A task is the work required to prepare, organize, draft, send, file, update, summarize, compare, or document something. A decision is the judgment about what should happen. AI is usually better at helping with the task than owning the decision.
For example, AI can summarize a customer complaint, pull out the timeline, draft a response, and list possible resolution options. But deciding whether to refund, replace, escalate, or make an exception may still belong to a human.
AI can draft a proposal using your notes, prior examples, and service descriptions. But deciding the price, terms, and strategic positioning may still need you.
AI can summarize job applications against role criteria. But deciding who gets interviewed needs care, fairness, and accountability.
When you inventory your work, split the preparation from the decision whenever possible. That one move creates safer, more useful AI workflows.
Lesson 1 takeaway
You can’t find strong AI opportunities by asking, “Where can we use AI?” You find them by asking, “Where does the business repeat work, lose time, create rework, or depend too much on one person’s memory?”
Your task inventory doesn’t need to be perfect. It needs to be honest. Once the work is visible, the best opportunities become much easier to rank.
Lesson 2: score your AI opportunities
Now that you have a task inventory, you need a way to choose what comes first. This matters because not every AI opportunity deserves your attention. Some tasks are easy to automate but barely worth the setup. Some tasks save time but carry too much risk. Some tasks are annoying but not valuable. Some tasks look complicated until you split them into smaller pieces.
The AI opportunity matrix gives you a practical scoring system. It isn’t meant to replace your judgment. It’s meant to organize it.
The three factors that matter most
For each task, you’ll score three things: repetition and pattern clarity, judgment fit, and business impact.
Repetition and pattern clarity asks how often the task follows a similar structure. Appointment reminders, intake summaries, invoice reminders, and standard FAQs usually score high. Strategy work, complex negotiations, and unusual client issues score lower.
Judgment fit asks how safe it is for AI to help. A high score means the task has clear rules, a human can review the output easily, and the cost of a mistake is manageable. A low score means the work depends heavily on context, trust, legal sensitivity, financial accuracy, health or safety, hiring fairness, or relationship judgment.
Business impact asks whether improving the task would meaningfully help the business. Time saved counts, but so do faster response times, fewer errors, better follow-up, more consistent content, stronger customer experience, and more owner capacity.
These three factors keep you from chasing novelty. A task isn’t a strong AI candidate just because AI can technically do it. It needs a clear pattern, a safe review path, and a business reason.
Use the scoring rubric in the visual components to score each task from 1 to 5.
Add two gates before you decide
After the three-factor score, add two gates: input readiness and risk level.
Input readiness asks whether the task has usable information going into it. If the task depends on messy notes, inconsistent forms, unlabeled files, missing CRM fields, or knowledge trapped in someone’s head, AI may still help later. But first you need to clean up the input.
Risk level asks what happens if the output is wrong. Low-risk tasks can usually be tested quickly. Medium-risk tasks need review. High-risk tasks may still benefit from AI assistance, but they shouldn’t be fully automated without clear rules, testing, and accountability.
These gates protect you from a common mistake: scoring a task high because it repeats often, then ignoring the fact that the inputs are messy or the downside is serious.
If the risk is high, human review stays in the workflow. If the inputs aren’t ready, fix the input before you automate.
Choose the right category
Once you score each task, place it into one of four categories. Automate is for tasks that are repetitive, low-risk, clear, and meaningful enough to justify setup. These tasks have predictable inputs and outputs. Appointment reminders, standard status updates, form-to-CRM entry, simple internal notifications, and basic document routing often fit here.
AI-assisted is where most small business value lives. These tasks repeat enough for AI to help, but they still need a human review step. Customer replies, proposal drafts, blog outlines, call summaries, sales follow-ups, SOP drafts, and internal reports often belong here.
Selectively assist is for tasks where AI helps with one piece, not the whole workflow. For example, AI might summarize research for a strategy session, compare vendor options before a human chooses, or draft interview questions while a manager makes the hiring decision.
Keep human is for work that’s relationship-sensitive, unusual, high-stakes, low-volume, poorly understood, or not worth the setup. This category isn’t a failure. It’s discipline.
The goal isn’t to automate everything. The goal is to choose the right level of AI involvement for the task.
A few examples
Imagine a local service business that receives estimate requests through a website form. The form has clear fields: name, service needed, location, timeline, and project notes. The owner currently reads each submission, writes a short reply, and adds the lead to a spreadsheet.
That task likely scores well. It repeats, the input is clear, the output is predictable, and response speed affects revenue. The right first workflow might be AI-assisted: summarize the request, draft a reply, tag the service type, and let a human review before sending.
Now imagine the same business has occasional angry customers with complex service issues. AI can still help summarize the complaint and prepare options, but the final response should stay human. That’s selectively assist, not automate.
Or imagine an agency writing custom strategy proposals. AI can help turn discovery notes into a first draft, pull in standard service descriptions, and check whether the proposal answers the client’s stated goals. But pricing, scope, tradeoffs, and positioning need human judgment. That’s AI-assisted.
The matrix helps you avoid false choices. You don’t have to pick between “AI does everything” and “AI does nothing.” Most useful workflows sit between those extremes.
Watch out for false quick wins
Some tasks look like perfect AI candidates until you inspect them more closely.
The first type is the task with inconsistent inputs. If customers send requests in wildly different formats, or files arrive with no structure, AI may need too much cleanup to be reliable at first. Standardize the intake before building the workflow.
The second type is the task with a high error cost. Financial calculations, legal language, health or safety guidance, hiring decisions, sensitive customer issues, and compliance-related work need a stricter review path. AI may prepare, but it shouldn’t decide.
The third type is the task nobody understands. If the team can’t explain how the work should be done, AI won’t magically make it clear. Document the process first.
The fourth type is the task that saves almost no time. A five-minute monthly task may be easy to automate, but setup may take longer than the next year’s worth of manual work.
This is why the matrix includes business impact. Automation is only worth building when the payoff justifies the effort.
How to handle ties
You may end up with several tasks that score similarly. That’s normal. When scores are tied, choose the task with the clearest input, the easiest review step, and the strongest connection to revenue, response speed, customer experience, or owner capacity. If all else is equal, choose the workflow your team will actually use.
Adoption beats elegance. A slightly smaller workflow that gets used every week is more valuable than a perfect workflow nobody trusts.
Lesson 2 takeaway
The best AI opportunities aren’t always the flashiest ones. They’re the tasks where repetition, clear inputs, safe review, and business value meet.
Your matrix turns vague interest into a ranked list. Once you have that list, tool selection becomes much easier because you’re no longer asking, “Which AI tool should I buy?” You’re asking, “Which tool helps with this specific workflow?”
Lesson 3: estimate value before you build
You now have a scored list of AI opportunities. Before you choose tools or write prompts, you need one more thing: a baseline. Without a baseline, AI success becomes a feeling. It seems faster. It seems easier. It seems useful. But three months later, you can’t say what changed.
A baseline gives you a before picture. It helps you see whether the workflow is saving time, improving quality, reducing delays, or making the business easier to run.
Start with your top candidates
Don’t estimate every task in the business. Start with your top five to ten candidates from the matrix.
For each one, estimate the current weekly time cost. Take the time per instance and multiply it by frequency. If the task happens daily, use a normal workweek. If it happens monthly, convert it into a weekly average.
Then estimate the AI-assisted version. Be conservative. AI may reduce drafting time, research time, formatting time, or routing time, but review still takes time. Setup also takes time at the beginning.
For early planning, think in ranges:
Tasks that are safe to automate may eventually remove most of the manual work, but they still need setup, monitoring, and exception handling.
AI-assisted tasks often produce meaningful savings because the blank-page or first-pass work disappears, but the human still reviews, edits, and approves.
Selectively assisted tasks usually save less time, but they can improve thinking, quality, or speed on specific steps.
The exact number matters less than the habit. You’re training yourself to connect AI work to business results instead of vibes.
Use the quick-win calculator in the visual components to make this simple.
Time saved is capacity, not value
This is one of the most useful distinctions in the course: when AI saves two hours, it has created capacity. It hasn’t automatically created value.
Value depends on what you do with that capacity. If those two hours go into sales follow-up, customer retention, content that attracts leads, strategic planning, or faster delivery, the business benefits. If those two hours disappear into more email, more scrolling, or more scattered work, the workflow may feel nice without changing much.
This is why each quick win needs a reinvestment plan. If AI saves you three hours per week on content drafting, maybe those hours go into publishing consistently, improving offers, or contacting past customers. If AI reduces proposal prep time, maybe you use the capacity to follow up faster. If AI shortens internal reporting, maybe managers use that time to fix blockers earlier.
The point isn’t to squeeze every minute. The point is to decide where the recovered attention should go before it gets absorbed by the day.
Measure quality, not just speed
Time savings are easy to understand, but they aren’t the only outcome. Some AI workflows are valuable because they improve consistency. Customer replies become more complete. Proposals follow a better structure. Meeting summaries capture decisions more reliably. SOPs become easier to produce. Content gets published on schedule instead of staying half-finished.
For each top candidate, choose one quality signal.
For customer communication, that might be response time, number of unresolved follow-ups, customer satisfaction, or how often messages need rewriting.
For content, it might be publishing consistency, revision rounds, topic coverage, or the time between idea and finished draft.
For operations, it might be fewer missed steps, faster handoffs, fewer repeated questions, or cleaner documentation.
For sales, it might be follow-up speed, proposal turnaround time, CRM completeness, or fewer leads slipping through.
You don’t need a complicated dashboard. You need enough of a baseline to know whether the workflow is worth keeping.
Pick one or two first workflows
This is where discipline pays off: after scoring and estimating, choose one or two workflows to build first. Not five. Not ten. One or two.
The first workflow teaches you how your business handles AI in practice. You learn what context the tool needs, what review looks like, where people hesitate, what breaks, and what documentation is missing.
That learning carries into the next workflow. If you try to implement too many at once, every problem gets harder to diagnose.
Your first workflow should have four traits: it repeats often, it has clear inputs, the risk is manageable with review, and the benefit is visible within a few weeks.
That last trait matters. Early wins build trust. If your team sees AI making one real workflow better, the next workflow becomes easier to adopt.
Know when a self-assessment isn’t enough
The matrix gives you a strong starting point, but there are times when a deeper audit is worth it. If the workflow touches sensitive data, multiple systems, legal or financial risk, complex customer handoffs, or several team members, it may need more than a spreadsheet score. You may need process mapping, tool review, privacy review, automation design, or outside implementation support.
That doesn’t mean you can’t start. It means you shouldn’t pretend a high score removes complexity.
Use this module to identify the opportunities. Then use the risk gates to decide how much structure each one needs before you build.
Lesson 3 takeaway
AI work should create a measurable change: faster response, cleaner documentation, better follow-up, more consistent content, less rework, or more owner capacity.
Before you choose tools, define the baseline. Before you celebrate time saved, decide where that capacity goes. Before you automate, choose the level of review the task deserves. That’s how AI starts becoming part of your operating system instead of another experiment.
Module 2 deliverable: your AI opportunity matrix
Your deliverable for this module is a completed AI opportunity matrix. Start with the task inventory from Lesson 1. Score each recurring task using the rubric from Lesson 2. Add input readiness and risk level. Then estimate the value of your strongest candidates using the quick-win calculator from Lesson 3.
By the time you finish, you should have:
- A list of recurring tasks across the business.
- Scores for repetition and pattern clarity, judgment fit, and business impact.
- Input readiness and risk level for each task.
- A category for each task: automate, AI-assisted, selectively assist, or keep human.
- Your top one to three quick wins.
- A simple plan for where the saved time or attention will go.
Save this matrix. You’ll use it in Module 3 when you evaluate tools, in Modules 4 through 7 when you build prompts and workflows, and when you review your progress.
You’re no longer asking whether AI is useful in general. You now have a way to decide where it’s useful for your business.
Module complete
Module 2 complete
In Module 3, you’ll use your opportunity matrix to choose tools without getting pulled into tool overload. You’ll learn how to evaluate an AI tool against a real workflow, what red flags to watch for, and when a paid tool is worth it.
You have the target now. Next, we choose the right tools for it.