AI for Automation: How Businesses Are Using AI to Save Time and Work Smarter
One of my favorite examples from our recent AI automation panel involved Santa Claus.
A few years ago, panelist Aaron Watkins built an automation where a parent filled out a form with information about their child: what they wanted for Christmas, what they had accomplished in school, their siblings, and other personal details.
AI took that information and wrote a personalized letter from Santa. Then the automation sent the letter to a service with a robotic arm that physically wrote it with a pen on Christmas stationery and mailed it to the child.
The project went on to win a Zapier AI innovation award in 2024.
It was a fun example, but it also demonstrated something much bigger.
AI automation isn't simply about getting ChatGPT to write something for you. It is about connecting multiple steps, systems, information sources, and decisions into a workflow that can perform work with little or no human intervention.
That was the focus of our recent AI for Automation panel discussion at Roseville Venture Lab.
I moderated the discussion with three people who approach automation from very different perspectives:
Linda Gilbert, VP of Product and Engineering at Huloop Automation, who works primarily with enterprise organizations including banks and credit unions.
Nicholas Haystings, co-founder of Sweet Fleet Virtual Assistance, Square Root Academy, and Sac Tech Week, who works with small businesses and nonprofits.
Aaron Watkins, Director of Automation AI at NetBrain and founder of Canopy Automated, where he helps small businesses automate repetitive work.
Despite those very different perspectives, there was surprising agreement about what businesses should actually be doing with AI.
Here are some of my biggest takeaways.
1. AI Automation Isn't New. AI Just Made Automation Much More Powerful.
Aaron made an important distinction early in the conversation.
Businesses have been automating processes for years.
Traditional automation looks something like this:
If A happens → do B.
If B happens → check C.
If C is true → do D.
Maybe someone fills out a form on your website. Their information goes into your CRM, an email gets sent, and a salesperson receives a notification.
None of that requires artificial intelligence.
Where AI changes things is in the amount of judgment and personalization that can happen inside that workflow.
Instead of sending everyone the exact same canned email, AI can look at what the customer submitted and write a different response for every person.
Aaron described it well: traditional automation connects systems such as your website, email, CRM, and accounting software. AI lets those systems begin responding intelligently to the information moving between them.
That's an important distinction.
Automation moves the work. AI can help interpret the work.
Put the two together and things get interesting.
2. Don't Automate a Bad Process
This may have been my favorite lesson from the discussion.
When I asked Nicholas about the automation projects he builds for clients, he almost immediately backed up the conversation.
Before automating anything, he said, you have to define what you're actually doing.
That means having a Standard Operating Procedure, or SOP.
What starts the process?
What is the outcome?
What happens in between?
Who is responsible for each step?
What decisions have to be made?
Only after answering those questions should you start automating.
Aaron made essentially the same point later in the conversation:
Make the business first. Map it. Then automate it.
I think this is where a lot of businesses are going to get AI wrong.
They see a new AI tool and immediately ask, "How can we use this?"
That's backwards.
Start with the business problem.
If your customer onboarding process is a mess, adding AI may simply give you an automated mess.
It's like installing a conveyor belt in a kitchen before deciding where the refrigerator goes.
Figure out the process first.
Then automate it.
3. Start With the Boring Stuff
When people talk about AI, they tend to gravitate toward the flashy applications.
But some of the best automation opportunities are painfully boring.
Aaron described the kind of work he looks for when working with small businesses:
Following up with leads
Responding to missed calls
Following up on unpaid invoices
Requesting customer reviews
Moving information between systems
Updating a CRM
Sending routine customer communications
These are exactly the things founders tend to hate doing.
Most people start a business because they're good at something.
The plumber wants to fix plumbing.
The photographer wants to take photographs.
The consultant wants to help clients.
Nobody starts a company because they have an uncontrollable passion for updating their CRM and sending their third invoice reminder.
But those little administrative tasks consume enormous amounts of time.
Aaron's basic question to a client is essentially:
What takes up your time all day that you don't want to do?
That's probably one of the best places to start looking for automation opportunities.
4. Speed Matters, Especially When Responding to Leads
Aaron gave a great example involving missed calls.
Imagine you're a plumber under somebody's kitchen sink when a prospective customer calls.
You can't answer.
The customer doesn't necessarily wait around. They call the next plumber.
An automated missed-call text can immediately respond:
"Thanks for calling. I'm tied up right now, but I'll get back to you shortly."
Now the conversation has started even though you haven't stopped working.
Aaron also described an interesting approach to website leads.
When someone fills out a form, send the normal confirmation immediately.
Then wait about three minutes.
After that, send a second, more conversational message using their first name and referencing their inquiry.
Why three minutes?
Because an immediate personalized response obviously feels automated.
Three minutes feels like someone may have actually seen the inquiry and responded.
The important lesson isn't really the three minutes.
It's that good automation should be designed around the customer's experience, not just the company's efficiency.
As Aaron put it, automation shouldn't necessarily feel like automation.
5. The Better the Information You Give AI, the Better the Automation Becomes
Another theme that came up repeatedly was what Aaron called building "brains."
For example, his company maintains different sources of knowledge about:
Competitors
Their own products
Current marketing campaigns
Company messaging
Previous content
Which content performed well
An AI system can then reference those sources when creating something new.
If the company wants an article comparing its product to a competitor, the AI doesn't simply rely on whatever happens to be inside the language model.
It retrieves information from those specific knowledge sources first.
In AI terminology, this is commonly called Retrieval-Augmented Generation, or RAG.
But you don't really need to remember the acronym.
Think about it like giving a new employee a filing cabinet.
If you tell somebody, "Write an email to one of our customers," they'll probably produce something generic.
If you give them your customer history, product information, brand guidelines, examples of your best emails, and previous customer conversations, they're much more likely to produce something useful.
AI works the same way.
Garbage in, garbage out still applies.
6. Some Decisions Still Need a Human
Linda brought an important enterprise perspective to the conversation.
At Huloop, the idea of the human in the loop is built directly into how they think about automation.
A computer may be perfectly capable of processing thousands of repetitive transactions.
But what happens when one transaction suddenly represents a million dollars?
Maybe that is where a human should take another look.
The same principle applies outside banking.
AI might:
Summarize a customer situation
Review large amounts of information
Identify an unusual pattern
Recommend a course of action
Prioritize which customers need attention
But there are circumstances where judgment, risk, compliance, or simply human empathy should determine the final decision.
The goal isn't necessarily to remove humans.
The goal is to remove the work humans don't need to be doing so they can spend more time on the work where human judgment actually matters.
That is an important distinction.
7. Stop Asking, "Can We Automate This?"
During the audience Q&A, Aaron made another distinction I liked.
Today, you can probably automate at least some portion of almost any business process.
So the question isn't:
Can this be automated?
The better question is:
Should this be automated?
There is a big difference.
A process with hundreds of exceptions may technically be automatable, but building and maintaining the automation could cost more than simply letting somebody handle it.
Bad data creates another problem.
If half of your customer information lives in your CRM and the other half lives inside your salesperson's head, AI doesn't magically know about the second half.
Linda pointed out that exception-heavy processes, poorly defined workflows, and incomplete information can dramatically increase the difficulty and cost of automation.
Technology makes a lot of things possible.
That doesn't automatically make them good business decisions.
8. Don't Get Distracted by the Tool of the Week
We talked about quite a few AI tools during the panel.
Nicholas mentioned tools including Blaze, Lovable, Apollo, and Lindy.
Aaron discussed Fathom, Zapier, Pipedream, Claude, and others.
The list could probably have continued for another hour.
But Nicholas made a point I agree with: you don't need to master every AI application that launches.
Different parts of his team use different tools depending on what they're trying to accomplish.
I think founders sometimes suffer from AI tool FOMO.
A new application launches on Monday.
Everybody on LinkedIn says it will change civilization by Tuesday.
By Friday, we're all supposed to reorganize our companies around it.
Don't do that.
Start with the problem.
Then find the tool.
Not the other way around.
9. AI Can Become Your Personal Operations Assistant
Some of my favorite examples from the panel weren't enterprise applications at all.
They were personal productivity systems.
Aaron built himself an automated daily task manager.
Every morning, it looks across his CRM, task application, email accounts, and even notes from his phone.
It then gives him a summary of what deserves his attention that day.
What's aging?
What's urgent?
What email looks particularly important?
What should he work on first?
Considering he described himself as having two businesses, two kids, three dogs, a wife, and three hobbies, I can understand why he built it.
Linda gave another good example: employee onboarding.
Hiring someone may mean creating accounts in payroll, HR, benefits, IT systems, and other platforms.
Instead of somebody manually walking down the same checklist every time a person is hired, much of that workflow can be automated.
Neither example sounds like science fiction.
That's exactly the point.
They just save time.
10. Businesses Also Need Rules for How AI Gets Used
Toward the end of the discussion, we talked about something I think more small businesses need to start considering: a responsible AI policy.
Which tools can employees use?
What customer information can be entered into them?
Are public models allowed to train on company information?
How is personally identifiable information handled?
Who reviews AI-generated work before it reaches a customer?
Linda recommended deciding those rules deliberately, configuring appropriate security settings, protecting customer information, and continuing to have employees review AI-generated work rather than blindly accepting it.
Large enterprises already spend a lot of time thinking about governance, security, and compliance.
Small businesses generally don't have compliance committees or an AI governance department.
That's actually an advantage because we can move faster.
But moving faster doesn't mean ignoring the risks.
You still need some rules.
Where I Recommend Starting
If you've been watching all of this AI automation activity and aren't sure where to begin, I wouldn't start by buying software.
I'd do this instead:
Find one repetitive task.
Look for something you or your team do repeatedly every day or every week.Map the process.
Write down what starts it, each step involved, the decisions that occur, and the desired outcome.Identify the annoying parts.
What involves copying information, sending reminders, checking systems, creating repetitive communications, or moving data?Decide where humans are actually necessary.
Keep people involved where judgment, relationships, risk, or empathy matter.Automate one small workflow.
Don't try to automate your entire company on Friday afternoon.Measure whether it actually helped.
Did you save time? Respond faster? Generate more appointments? Collect invoices faster? Improve customer service?Then build the next one.
That's how I think most businesses should approach AI automation.
Not as one massive AI transformation project.
Just keep removing friction.
One process at a time.
Final Thoughts
The biggest lesson I took away from this panel wasn't about ChatGPT, Claude, Zapier, Lindy, or any particular piece of software.
It was much simpler.
The purpose of automation isn't to automate everything. It's to give people more time to do the things that actually require people.
Talk to customers.
Build relationships.
Make difficult decisions.
Create new products.
Solve problems.
Think.
Those are usually the things that created the business in the first place.
If AI can take some of the repetitive junk off your plate so you can spend more time doing those things, that's a pretty good place to start.
The AI for Automation panel was part of the AI for Business series at Roseville Venture Lab, designed to help local business owners understand and apply artificial intelligence in practical ways. The program is sponsored by the City of Roseville.
