What AI Should — and Shouldn’t — Automate in a Small Business

I’m James P. Garrity, an entrepreneur based in Lancaster, Pennsylvania.

I’ve spent years building and operating businesses across e-commerce, internet marketing, real estate, and now SaaS.

One of the biggest shifts I see happening in business right now is the move from traditional software to AI-assisted software.

But I think a lot of companies are asking the wrong question.

The question should not be:

“How much can we automate?”

It should be:

“Which parts of this process should software handle, and which parts still require human judgment?”

That distinction matters.

AI can save enormous amounts of time.

But poorly designed automation can also create mistakes, damage customer relationships, and make a business less trustworthy.

I believe the best small-business AI products will be the ones that automate repetitive work without removing people from the decisions that matter.

Start With Repetitive Work

The easiest place to apply AI is repetitive operational work.

Most businesses have dozens of tasks that happen over and over again:

• reviewing information,
• organizing data,
• checking statuses,
• drafting routine communication,
• categorizing requests,
• reminding someone about the next step,
• summarizing activity,
• finding exceptions,
• and prioritizing what deserves attention.

These tasks are important, but they often do not require deep human judgment every single time.

That makes them strong candidates for automation.

If software can save someone thirty minutes a day on one task, that may not sound dramatic.

But multiply that across a year and across several employees, and it becomes meaningful.

AI Is Especially Useful When Information Is Messy

Traditional automation works well when the rules are simple.

If X happens, do Y.

AI becomes more valuable when the input is less structured.

An email from a customer is a good example.

One customer might say:

“I’ll pay Friday.”

Another might say:

“We already mailed the check.”

Another might say:

“We’re disputing this charge.”

Another might say:

“Please stop emailing me.”

All four messages require different responses.

Traditional software can struggle when the wording changes.

AI can help interpret the intent behind the message.

That is where it starts becoming especially useful.

But Interpretation Is Not the Same as Authority

This is where I think companies can get into trouble.

Just because AI can understand a message does not mean it should automatically make every decision that follows.

There is a big difference between:

“This customer appears to be disputing the invoice.”

and:

“Automatically escalate this customer.”

The first is assistance.

The second is authority.

I believe small businesses should be very careful about handing authority to automation in areas involving money, customers, legal issues, or reputation.

AI should often identify and recommend.

A person should often approve.

Customer Relationships Need Context

Business relationships are not identical.

A one-time customer and a customer who has worked with you for ten years should not necessarily be handled the same way.

A small overdue balance and a major disputed account may require very different treatment.

A customer who promised to pay tomorrow should not receive the same follow-up as someone who has ignored repeated communication.

Human beings naturally understand that context.

Good AI software needs to preserve it.

Otherwise, automation becomes blunt.

Where AI Can Be Extremely Powerful

There are several areas where I think AI can create immediate value for small businesses.

Prioritization

Most business owners do not suffer from a lack of tasks.

They suffer from too many tasks.

AI can help rank what actually deserves attention.

Instead of showing a business owner 200 items, good software should help surface the ten that matter most.

Drafting

AI is extremely useful for preparing routine communication.

Emails.

Follow-ups.

Summaries.

Internal notes.

Responses.

The goal should not be to replace the person.

The goal should be to eliminate the blank page.

Classification

AI can help interpret what something means.

Is this customer making a payment promise?

Is this a dispute?

Is this a request for more information?

Is this an unsubscribe request?

That kind of classification can save a large amount of manual review.

Summarization

Small businesses increasingly have information spread across email, accounting systems, CRM platforms, notes, and internal tools.

AI can summarize that history into something a person can understand quickly.

That is a real productivity gain.

Exception Detection

In many workflows, most activity is routine.

The valuable part is identifying what is not routine.

AI can help flag the situations that look unusual or need human review.

That is where I see some of the biggest potential.

What I Would Be Careful Automating

There are also several categories where I would be cautious.

Sensitive Customer Communication

Routine communication can often be drafted or assisted by AI.

But emotionally charged disputes, legal threats, unusual account issues, or valuable customer relationships often need human involvement.

Financial Decisions

AI can analyze financial information.

It can identify patterns.

It can make recommendations.

But important decisions involving money should have appropriate human review.

Legal or Compliance Decisions

AI can help organize information, but businesses should not assume software can replace qualified professional judgment in legal or regulatory matters.

Irreversible Actions

Any action that is difficult to undo deserves extra caution.

Sending a draft for approval is one thing.

Automatically sending a harsh message to a valuable customer is another.

Good systems should distinguish between reversible and irreversible actions.

Why I Think Approval Workflows Matter

One design principle I like is:

automate the preparation, not necessarily the final decision.

That can dramatically reduce work without sacrificing control.

For example:

AI can review an account.

AI can summarize the history.

AI can identify the next likely action.

AI can draft the email.

Then a person approves it.

The person may spend thirty seconds instead of ten minutes.

That is still a major improvement.

And the business retains control.

How This Thinking Influenced Paid360.ai

This philosophy is central to how I think about Paid360.ai.

Paid360 is an AI-powered accounts receivable platform for small businesses using QuickBooks Online.

Accounts receivable is a perfect example of a workflow with both repetitive work and sensitive judgment.

There is a lot AI can help with:

• identifying overdue invoices,
• prioritizing follow-up,
• interpreting customer replies,
• preparing messages,
• tracking promises,
• recognizing disputes,
• and keeping the history organized.

But customer-facing communication still matters.

That is why I did not want to build a system that simply sends whatever it wants.

I wanted a system that helps the business make better decisions and move faster while preserving human control.

Good Automation Needs Safety Rules

AI is only one part of a good automation system.

Rules still matter.

For example:

if an invoice has already been paid, do not keep contacting the customer.

If a customer has asked not to be contacted, respect that.

If there is a dispute, routine follow-up may need to pause.

If someone has promised to pay on a specific date, the system should remember it.

If the accounting data is stale, the software should not pretend it knows something it does not.

Those safeguards are not flashy.

But they matter.

Reliable software is often defined as much by the actions it refuses to take as by the actions it can take.

AI Should Reduce Cognitive Load

I think one of the most underrated benefits of AI is reducing cognitive load.

Small-business owners carry a lot in their heads.

Who needs a response?

What is overdue?

What did that customer say last week?

Which issue is urgent?

What can wait?

AI can help turn that mental clutter into organized information.

That may be more valuable than simply automating clicks.

A business owner does not necessarily need a robot running the company.

They need clarity.

The Best AI Often Feels Boring

The most valuable business AI may not look futuristic.

It may simply:

• save someone an hour,
• prevent something from being forgotten,
• draft a useful response,
• organize information,
• flag a problem,
• or make a decision easier.

That may not look dramatic in a demo.

But operational improvements compound.

If a system saves time every day and reduces errors every week, that becomes very valuable over time.

Do Not Automate Just Because You Can

One of the risks of new technology is that people start looking for reasons to use it everywhere.

That is backwards.

Start with the problem.

Then decide whether AI is the right tool.

Some problems can be solved with a checklist.

Some require traditional software.

Some require automation.

Some benefit from AI.

And some still require a person.

The best technology stack is not the one with the most AI.

It is the one that solves the problem most effectively.

What I Think Small Businesses Should Do Now

I do not think small-business owners need to rebuild their entire company around AI.

I would start smaller.

Look at the repetitive work that happens every week.

Find the processes where people spend time organizing, summarizing, checking, or drafting.

Look for tasks that are annoying but predictable.

Those are often the best opportunities.

Then keep humans involved where judgment, relationships, money, or reputation matter.

That approach is less dramatic than saying AI will run the entire business.

But I think it is much more useful.

Building Practical AI From Lancaster

I’m continuing to build paid360.ai while based in Lancaster, Pennsylvania.

My interest in AI is not really about AI itself.

It is about what business owners can accomplish with better tools.

I have spent enough time operating businesses to know how much energy gets consumed by repetitive administrative work — across my entrepreneurial background and my real estate investing alike.

If AI can reduce that burden while keeping people in control, I think that is a very meaningful improvement.

That is the kind of software I want to build.

Automate the repetitive work. Assist the important work. Keep people in control of the decisions that matter.

That is how I think small businesses should approach AI.

You can read more about the platform on my founder profile.